0:00

Digital Foundry Direct Weekly is

0:02

presented by Sega and Atlas'

0:04

Metaphor Rifantazio. IGN's 2024 Game

0:06

of the Year available now

0:08

on Xbox PlayStation and Steam

0:10

and currently there's 25% off

0:12

for a limited time. Racking

0:14

up over 50 awards, including

0:16

Best RPG, Best Narrative and

0:18

Best Art Direction, at the

0:20

Game Awards, 2024. From the

0:22

creators of persona 3, 4,

0:24

and 5, this is Atlas'

0:26

first fantasy JRP in the

0:28

AAA space. Step into a unique

0:30

world where you join a royal

0:33

tournament to claim the throne, but

0:35

here's the twist. The late king

0:37

declared the next monarch must be

0:39

elected by the people. It's not

0:41

just about strength, you've got to

0:43

earn their thrusts. your time with

0:45

a calendar-based system choosing how to

0:47

spend your days. Questing, dungeon diving

0:49

or boosting your royal virtues. With

0:52

a free demo offering around five

0:54

hours of gameplay you can dive

0:56

in now and carry your save

0:58

onto the full game. Metaphor Rifantazio

1:00

is out now on Xbox PlayStation and

1:02

Steam. Buy it now at 25% off

1:04

for a limited time. Welcome

1:10

back to Digital Foundry for this,

1:12

the 250th edition of DF28 Weekly,

1:14

our weekly show, where we discuss

1:16

the latest gaming and technology news.

1:18

Lots going on this week, lots

1:20

to wait for joining me on

1:22

the panel, first of all, Oliver

1:24

McKenzie. Yes, hi Rich, you're to

1:26

talk about the latest in Microsoft

1:28

Leakes, and big news, Sony games,

1:30

should be fun. Absolutely, and

1:32

of course, Anexpetalia. Yeah, another round

1:34

of Microsoft Leakes. It seems like

1:36

it's a very common thing that

1:38

happens at Microsoft. Anyway, let's go

1:41

straight to it. Okay, first new

1:43

topic of the week, we are

1:45

going to be talking about

1:47

Microsoft Leakes. Specifically, news of

1:49

what's been described as the Project

1:51

Kenon Xbox handheld. Essentially, think

1:53

of this as an OEM

1:55

handheld design put together by Microsoft

1:58

that it can licensing... out

2:00

to the likes of a Seuss Lenova

2:02

etc. and it does seem to be

2:04

that the first product will be with

2:06

a Seuss and the details on this

2:08

well it seems to be we're actually

2:10

going to get to see what an

2:12

Xbox shell looks like on a PC

2:14

in this in this case a handheld

2:17

PC. And obviously our man with the

2:19

most experience of handheld PCs is of

2:21

course a lot more expensive. So I'm

2:23

really interested in your opinions on this

2:25

and what form this will actually take.

2:27

And bearing in mind that Microsoft are

2:29

intent on producing their own handheld, what's

2:31

the point of this endeavor? Yeah, I

2:33

definitely have some questions about that latter

2:36

part. But the basic contours of this

2:38

seem to basically be like that mooted

2:40

Windows Xbox hybrid kind of interface device

2:42

running Windows looking like Xbox behaving like

2:44

Xbox that kind of thing that makes

2:46

a lot of sense from my perspective

2:48

It sounds kind of a lot like

2:50

steam big picture mode or steam less.

2:52

You basically have an end-to-end interface that

2:55

would work for most gaming tasks and

2:57

then you'd still have the Windows Desktop

2:59

presumably for anything that required like more

3:01

advanced tweaking or downloading or maybe alternative

3:03

stores, things like this. You'd have that

3:05

functionality presumably available through Windows still. But

3:07

for me right now the benchmark for

3:09

this kind of experience is the steam

3:11

big picture mode. Right now my desktop

3:14

PC which I use in TV configuration

3:16

is booted right into steam big picture

3:18

mode and that covers most use cases

3:20

I think with controller friendly games. They

3:22

still have various issues with like launchers

3:24

crashing or not being cooperative with controllers.

3:26

Any kind of settings work you need

3:28

in Windows obviously does not work with

3:30

just the steam big picture mode Anything

3:32

that involves the Xbox app or you

3:35

be soft app or EGS or Battle.

3:37

That's not so good I've also had

3:39

some issues with the controller interface with

3:41

the store which sometimes has responded well

3:43

to control or input there But most

3:45

those issues are fixed in steamless so

3:47

to me basically there would be like

3:49

two levels to this You know number

3:51

one can Microsoft be steam big? picture

3:54

within Windows because that is kind of

3:56

providing us kind of analogous experience here.

3:58

And number two, can they beat SteamOS?

4:00

I think they really need to be

4:02

able to like get in there with

4:04

Steam's big picture mode in terms of

4:06

delivering a smooth end-to-end experience with hopefully

4:08

not too many compromises, but certainly if

4:10

they were able to get it closer

4:13

to that SteamOS level where it really

4:15

is an integrated interface for all manner

4:17

of adjustment, that would certainly be much

4:19

better here. And of course, if you

4:21

need to remember, like, the timelines on

4:23

here are maybe not so favorable to

4:25

Microsoft. They're talking about this machine potentially

4:27

coming out in late 2025 or even

4:29

2026. That's a period where you're going

4:32

to have quite a few Steam West

4:34

devices, I presume, on the market. Obviously,

4:36

with Steam deck itself, various steam deck

4:38

itself, or maybe even a full third-party

4:40

generic version of Steam on any device

4:42

that seems to be in the picture

4:44

as well. So that seems like a

4:46

bit of a bit of a dangerous

4:48

combination combination combination for people for people.

4:50

a lot of this effort might be

4:53

a bit of a non-starter because obviously

4:55

the beauty of steamless that it's built

4:57

around the largest gaming platform on PC's

4:59

period and this maybe would not be

5:01

that hopefully would have some good integration

5:03

of steam that would really kind of

5:05

be a prerequisite I think for this

5:07

to work out well. So ultimately I

5:09

am I have somewhat mixed feelings about

5:12

coupling windows with this kind of interface

5:14

but I think you know Microsoft definitely

5:16

has hopefully has the capability to pull

5:18

it off here. And there are a

5:20

decent number of games that won't run

5:22

on Steam Deck or won't run well

5:24

on Steam Deck, going to translation issues

5:26

with ProCon, obviously you have anti-cheek problems,

5:28

problems with multiplayer games. This effort would

5:31

presumably sidestep all those issues. That is

5:33

a big, big win for Microsoft potentially

5:35

if the rest of the operating system

5:37

can be. contorted into a form that

5:39

would work with controllers here. But I

5:41

mean, ultimately, I think one final question

5:43

here would be just with respect to

5:45

this particular effort, because like you mentioned,

5:47

Rich, this is an Aesus exclusive effort,

5:49

or at least the Aesus is potentially

5:52

the first firm involved in such an

5:54

effort here. And I kind of wonder,

5:56

like, would it be the same Zen

5:58

processors we've seen so far? Would it

6:00

be the strict point APU? Would it

6:02

maybe even be the... and Z2 go

6:04

APU which is a Rembrandt based APU.

6:06

I don't know that does not sound

6:08

super exciting to me from a performance

6:11

perspective. Now if this was based on

6:13

a custom chip or something that would

6:15

be a lot more exciting and potentially

6:17

quite tantalizing. but I think probably this

6:19

initial effort will just be with that

6:21

kind of off the shelf part here.

6:23

And I think ultimately like what I'd

6:25

like to see from Windows handheld would

6:27

be something that's led by Microsoft that

6:30

maybe has a custom ship that has

6:32

an additional level of power in it

6:34

that can kind of push beyond this

6:36

spec that we've seen for the past

6:38

three years with every device in the

6:40

15 watt range kind of settling in

6:42

at plus or minus 20% steam deck

6:44

performance. That'd be an area of interest

6:46

for me as well. Wow, that's a

6:49

lot to go to. I'm going to

6:51

go to Alex first though before I

6:53

pick that apart. Yeah, for me. I'm

6:55

going to echo everything Oliver said. I

6:57

think there's a great, this could lead

6:59

to a good way to unify the

7:01

access to settings in Windows in a

7:03

way that is more controller friendly. Microsoft

7:05

has had historical problems with this with,

7:07

you know, you can go into Windows

7:10

and there's still things stuck in the

7:12

control panel. There's still things stuck in

7:14

settings. You type in the search bar,

7:16

something from the settings and it doesn't

7:18

show up. You have to go to

7:20

settings manually. It's just a mess, right.

7:22

Windows UI interactions to actually start working

7:24

again. I'd be grateful. But to the

7:26

point at hand, I think actually this

7:29

project canon is going to be a

7:31

bit not from the exact same way

7:33

it's done on the Sony side, but

7:35

I think it's almost like Microsoft's PS5

7:37

Pro. in the aspect of it is

7:39

testing the waters and going to be

7:41

used as a beta testing platform for

7:43

what they're doing with the next Xbox,

7:45

all the UI stuff, the sideways integration

7:48

with steam that has been already mentioned

7:50

by Phil Spencer in the past, as

7:52

well as just mentioned by Oliver, all

7:54

of these things that are going to

7:56

be teething points and points where you

7:58

need kind of like open. beta testing

8:00

from an audience, like the flighting

8:03

as they call it at Microsoft,

8:05

I think. This is what this

8:07

product I think is actually going

8:09

to be very useful for for

8:11

them. And it's also going to

8:13

see exactly how does this fit

8:15

into the legacy catalog of PC

8:17

gaming. Unlike the Xbox handheld, which

8:19

is probably going to come out

8:21

later than this, maybe 2027. This

8:23

one. will not have the Xbox integration

8:26

in terms of like you carry

8:28

your library with you right in

8:30

the way that we expect with

8:32

an Xbox branded product this one

8:34

is presumably just gonna have like

8:36

Xbox anywhere I think is it

8:38

is called like the the streaming

8:40

capability as well as any other

8:42

things you might have already had

8:44

on the Microsoft store including your

8:46

third-party things you know purchased through

8:48

steam and EGS and such. So

8:50

it's not going to be the

8:52

exact same thing as following the

8:54

PS5 route, but I think that's

8:56

exactly what this project Kenan is.

8:58

It is a testing grounds for the

9:00

further development of what this next Xbox

9:03

ecosystem is going to be, which

9:05

is going to be much more like

9:07

a PC and much more like the

9:09

Windows ecosystem we've seen in the past.

9:12

Just hopefully with a lot more intuitive

9:14

controls and also just, yeah, extractable performance

9:16

that is not going to make you

9:19

tear your hair out. I really

9:21

think if Microsoft is pushing this thing,

9:23

they need to do exactly what

9:25

Steam does, where there's like controller

9:27

profiles downloading with your games and

9:30

stuff. They need the equivalent of

9:32

Game Scope to run really well. The

9:34

Microsoft Gamebar is okay. But once again,

9:36

you know, it could be better, the

9:38

ability to like frame rate limit and

9:41

do all these things that you have

9:43

with GameCope Integrated FSR, a whole bunch

9:45

of other features there that are really,

9:47

really good. I think that's something

9:50

that Microsoft needs to bring over

9:52

to the PC space because currently

9:54

that's filled by third party tools

9:56

in the PC landscape, which are

9:58

all plugging into Windows. API calls,

10:00

you know, it feels like they

10:02

could do that on their own

10:05

and perhaps do it better in

10:07

a way that doesn't require the

10:09

mouse and keyboard to work. So

10:11

yeah, there's a lot that I

10:13

feel like this product could represent.

10:15

But like Oliver, I'm very curious

10:17

about the technological underpinnings of it.

10:20

If it is really performance-wise, a

10:22

lot like just this plus or

10:24

minus 20% steam deck that Oliver

10:26

just said, it doesn't actually sound

10:28

too interesting. It has to also

10:30

maybe have some good hardware oof

10:33

there to make it sound more

10:35

interesting as well as a sleep

10:37

button. Like a good sleep button.

10:39

Something steam has, the steam deck

10:41

is awesome for that aspect. If

10:43

you try sleeping games in Windows,

10:46

good luck. You know, like if

10:48

you've ever tried it. It's not

10:50

usually a good thing. So that

10:52

would be another thing that they

10:54

would have to work on. Windows

10:56

sleep functionality, which hasn't been touched

10:58

in a long time. Okay, so

11:01

where do we begin with this?

11:03

Well, from my perspective, let's first

11:05

of all talk about why they're

11:07

doing it when they've got their

11:09

own handheld coming in a couple

11:11

of years. And I think it's

11:14

simply because, first of all, they

11:16

need to solve the problem of

11:18

Windows being able to be used

11:20

as a console-style operating system. And,

11:22

yeah, part of the interface is

11:24

obviously a key issue for it,

11:27

but there's a lot more to

11:29

it, like, you know, system software

11:31

updates, that kind of thing. you

11:33

know, how are you going to

11:35

install those third party stores unless

11:37

you plug in a mouse and

11:40

keyboard and do it, you know,

11:42

the manual way. That's my main

11:44

issue with PC gaming on a

11:46

handheld at the moment is that

11:48

kind of like a USB hub

11:50

with a mouse and keyboard are

11:52

kind of like essential for setting

11:55

everything up. And then you're kind

11:57

of good to go. There's a

11:59

lot going on there. And as

12:01

you say, Alex, you look at

12:03

SteamOS and what it's doing in

12:05

terms of bridging the gap between

12:08

what a console is and what

12:10

a PC is. And Game Scope

12:12

is an excellent example of that

12:14

because it's doing a lot of

12:16

great PC style stuff that I

12:18

think high end users would probably

12:21

want on an Xbox. In theory

12:23

it should just be plug and

12:25

play right but you know when

12:27

you start verging it into the

12:29

PC area rather you've got to

12:31

have PC style features I think.

12:33

I think fundamentally the interface will

12:36

just look like an Xbox you

12:38

know like an Xbox series X.

12:40

fundamentally, with extra bits added on.

12:42

The concept of unifying your game

12:44

libraries, an early solution to that

12:46

is probably just simply some kind

12:49

of unified launcher. I don't foresee

12:51

Microsoft allowing you to buy steam

12:53

games from within their shell. That

12:55

would be kind of like counterintuitive,

12:57

I think. In terms of why

12:59

they're doing it, I think it

13:02

is simply that... Well, you know,

13:04

maybe I always sort of, I

13:06

was going to say this could

13:08

be a reaction to the Lenovo

13:10

steam-o-s device, right? The thing is,

13:12

though, the industry doesn't react that

13:14

quickly. Products don't just magically appear

13:17

out of thin air, with a

13:19

possible exception of like the 1070-T-I

13:21

like the day when Vega came

13:23

along. And it was great. and

13:25

video just sort of like stripped

13:27

back the 1080 a bit, called

13:30

it a 1070-DI, very easy to

13:32

do in that sense, but from

13:34

this, I don't think, you know,

13:36

these sort of, this sort of

13:38

strategy, this kind of ability to

13:40

put together that interface, associating with

13:43

whoever the partner is, Sus, or

13:45

whatever, there have been rumblings about

13:47

Microsoft providing OEM hardware for like...

13:49

you know, well over a year

13:51

at this point. So I think,

13:53

you know, it was always going

13:55

to happen, but I do think

13:58

that Steam Deck was a catalyst,

14:00

and I do think that Microsoft

14:02

saw it and thought, well, you

14:04

know, what if they start to

14:06

encroach into our Windows space? This

14:08

is problematic. Hardware-wise, let's talk about

14:11

that, could go one of two

14:13

ways. If it is an OEM

14:15

device that Microsoft is basically licensing

14:17

out, there is potential for some

14:19

customization, right? However, what if to

14:21

create custom silicon is not actually...

14:24

a cheap endeavor, it requires a

14:26

lot of money. And if AMB

14:28

already has off-the-shelf solutions that will

14:30

do the job, it's probably the

14:32

best way forward. Which one? Well,

14:34

tricky, right? You'd really want Strict's

14:36

point, which would be the latest

14:39

and greatest. But even then, you

14:41

know, what for what? It's not

14:43

a generational leap over steam deck.

14:45

You have to kind of, similar

14:47

to the Asuzvog ally and... Z1,

14:49

Z2, extreme. You're going to need

14:52

to throw more power at it

14:54

to get more performance, you know,

14:56

a lot more performance. I'm not

14:58

sure that's what Microsoft would want.

15:00

If it is that, was it

15:02

all over the Z2 Go? Z2

15:05

Go is remember. Yeah, which is

15:07

basically what, a 6800 U. Yeah.

15:09

Yeah, which is okay, it is

15:11

faster than steam deck, but again,

15:13

you need to throw more power

15:15

at it to get that extra

15:17

performance. 15 watt versus 15 watt

15:20

is extremely unimpressive. So yeah, I

15:22

mean, it's really, really interesting that

15:24

they're doing this, and I think

15:26

it is because they probably do

15:28

need a proving ground, as you

15:30

say, Alex. But my expectations are

15:33

kind of tempered as to what

15:35

it's going to be. I honestly

15:37

think that if you're putting out

15:39

an Xbox that doesn't run Xbox

15:41

games, you know, for example, the

15:43

360 back compartment stuff. I think

15:46

that's a problem, right? Because you

15:48

want your digital library to run

15:50

on your Xbox, really. Let's take

15:52

some supportive questions on this. This

15:54

one from Anctus the Cronode figure.

15:56

Hi DF Crew, we've all now

15:59

heard of... Xbox Ken and the

16:01

Barbariat. I think 2025 would be

16:03

too soon to release it for

16:05

Microsoft as there wouldn't be any

16:07

hardware advantage over other handhelds. I

16:09

mean, it would be great to

16:11

see FSR 4 working on the

16:14

device. What do you think, Giers,

16:16

exclamation point? Yeah, I mean, you'd love

16:18

to see FSR 4, either you'd need

16:20

a back ported version of FSR 4,

16:22

that runs on our DNA 3, or

16:25

you would need... a new APU that

16:27

runs our DNA for. And even

16:29

then, there are problems, I think,

16:31

in terms of its computational

16:34

cost. This one from rave

16:37

underscored Damos in all in block

16:39

capitals. Dear Digital Foundry

16:41

Fellows, I have two

16:43

questions, if that's allowed. I'll

16:46

allow that. Yeah, might as

16:48

well. According to Jess Corden,

16:50

Microsoft is repulsively working on

16:52

Windows updates that add an

16:54

Xbox software shell for PC

16:56

gaming, likely compatible with various

16:58

hardware. Conceivably, this shell

17:00

would work on existing devices

17:03

like the Rog-ALI2 thoughts. I'd say

17:05

yes. If they're putting out

17:07

an OEM Xbox, unless there's

17:09

some sort of... hardware differentiation.

17:11

I don't see any reason

17:13

why it wouldn't run on

17:15

pretty much any APU-based handheld

17:17

system. And I think that's

17:19

fine. And I think Microsoft

17:21

would actually welcome that. The

17:23

more people that are running

17:26

an Xbox-like experience on their

17:28

perspective, the better it's all

17:30

ties into this. This is an

17:32

Xbox beat that they're putting out

17:34

there. Perchance, which... I love the

17:36

way that people sort of adopt

17:38

like... Victoria. Four like language. Whenever

17:40

they address me in questions. But

17:42

chance which, there's a digital foundry

17:44

plan coverage of the Jesus Xbox

17:47

branded PC handheld rumored to arrive

17:49

later this year. Given the hardware

17:51

is likely the same or marginally

17:53

derivative of rogue ally hardware, would

17:55

the focus be on the software

17:57

user experience rather than just performance?

18:00

Absolutely, right? Because, you know, this is

18:02

the big problem Microsoft's got. It hasn't

18:04

got a handheld experience that works, really.

18:06

Windows is really not good. What's this

18:09

question here? I've got one more here

18:11

from Die Hard X86. Sounds like a

18:13

really great YouTube channel. It is. Anyway,

18:15

speaking on the OEM Collabor, Xbox branded

18:18

handheld, I'm a bit confused why Microsoft

18:20

would release another handheld skew instead of

18:22

just updating the current Rogalai slash Legion

18:24

Go. I mean, I don't see why

18:27

they can't do that. Why do you

18:29

think they are doing this? Windows Central

18:31

Jez Gordon mentioned it being a test

18:34

bench for their future hardware efforts, but

18:36

couldn't they get the same data from

18:38

currently available handhelds? Thanks. I think it's

18:40

all about this OEM program and whether

18:43

it's going to work and if it

18:45

is actually Microsoft hardware that's being licensed,

18:47

that's probably going to be one thing

18:49

they're going to do in future with

18:52

their quote-unquote actual handheld. And similarly, you

18:54

know, it doesn't really matter if you're

18:56

running this OS on other handhelds, the

18:58

more the merrier, right? Any final thoughts,

19:01

Oliver? Yeah, I mean, I think there

19:03

is probably some merit to having a

19:05

dedicated device with like a proper Xbox

19:08

button, you know, that's nice to have,

19:10

maybe a more pronounced grip, something like

19:12

that, inching more in that direction, inching

19:14

more in the direction of something that's

19:17

distinctively Xbox there. That could be part

19:19

of it. Also, it's just maybe easier

19:21

to launch in one device with one

19:23

ship set to start with. Maybe they've

19:26

got some special sauce in terms of

19:28

that ship set. you know, something new.

19:30

I guess that could be some, I

19:32

don't know. Maybe something, I don't know,

19:35

unlikely, but perhaps they have something interesting

19:37

from the hardware perspective that would justify

19:39

moving over. But I'd also presume that

19:42

this would be available for PCs to

19:44

use generally and handhelds to use in

19:46

particular as well, other handhelds. Yeah, yeah,

19:48

I agree on that. And yeah, they

19:51

have also S. Of course, which I

19:53

think would be an absolutely fine. to

19:55

have to a handheld based on it

19:57

showing on the Qualcomm, Snaptag and Ex

20:00

Elite surface. Yeah, some sort of all

20:02

in one style upscaler for any game.

20:04

There's certain value to that. I wonder

20:06

if it could be backported to existing

20:09

MD hardware that does have an MPU

20:11

in it. That's certainly an interesting thought.

20:13

Any final thoughts for you on this

20:16

one, Alex? Oh, I was just imagining

20:18

like a like a handheld with like

20:20

a pistol grip on it. You know,

20:22

it would fit really well into the

20:25

gamer aesthetic. Pistol grip, sight rails, hollow

20:27

sight on it. You could imagine what

20:29

it would look like. Okay, let's move

20:31

on quickly. Okay, our next new story

20:34

of the week. Well, it turns out

20:36

that Death Stranding, too, is going to

20:38

be releasing later this year, June, I

20:40

believe, and a pre-order trailer was released,

20:43

and holy crap, it looks really, really,

20:45

really good. Oliver, you were insisted that

20:47

this was discussed in this particular episode.

20:49

It does look absolutely amazing, right? Yeah.

20:52

Insistent is, perhaps, of the same thing,

20:54

but I do think we should definitely

20:56

talk about it this week. because I

20:59

thought it looked quite beautiful. First of

21:01

all, this trailer is apparently captured on

21:03

PS5. It's not totally clear, at least

21:05

the cinematic parts, because the specific wording

21:08

is the in-game cinematics are captured on

21:10

PS5, and that it's a mix of

21:12

gameplay and any cinematics that leaves... The

21:14

gameplay side of things is a little

21:17

bit up for debate and up for

21:19

question, but at least the cinematics are

21:21

captured on PS5 there presumably in real

21:23

time. And I think those cinematic scenes

21:26

look absolutely stunning, like I'm pretty confident.

21:28

These are some of the best directed,

21:30

best looking cutscenes in all of video

21:33

games as you might expect from Kojima.

21:35

The characters are using metahuman tech here

21:37

and I think they look much more

21:39

real than their counterparts in other decimal

21:42

engine games and the horizon games or

21:44

even in the characters from the first

21:46

game. The facial scans look really accurate.

21:48

The likeness of each actor is just

21:51

perfect in a way oftentimes. you see

21:53

in games there's like a little bit

21:55

of inconsistency between the way the actor

21:57

looks and the way their skin looks

22:00

and their hair looks and the their

22:02

portrayal in the game but here it

22:04

just looks perfect throughout. The lip sync

22:07

is very good the characters are quite

22:09

convincingly here. And there's obviously that scene

22:11

at the end as well with the

22:13

two characters kissing with Norman Redis and

22:16

Elias. They do, I think those are

22:18

the names, kissing, not their names, the

22:20

game. And you get that nose bending

22:22

thing and that reminded me a lot

22:25

of that same sequence with Nate and

22:27

Elena in uncharted four that was so

22:29

famous. So it's that nice little detail

22:31

here that they're throwing in there as like

22:34

a kind of a flex almost, that looks

22:36

pretty good. In general, in terms of the

22:38

technology, there were some things that. I'm

22:40

kind of wondering about maybe Alex

22:42

can provide his own insight here,

22:44

but I was thinking about the

22:46

lighting here and I'm not really

22:48

sure about RTGI like there are

22:50

some large-scale environments with pretty good

22:52

indirect diffuse lighting and earlier trailers

22:55

there were some moments where you'd

22:57

like open a hatch and light

22:59

would spill in quite beautifully into

23:01

the scene which seemed quite RTGI-ish

23:03

or potentially these moments of large-scale

23:05

destruction but in any case whatever

23:07

they're doing here looks hugely improved

23:09

over the first game it didn't have

23:11

that like rich physically accurate look as

23:13

much as we're seeing here i think

23:15

in a lot of scenarios that first

23:17

game looked a little bit more stark

23:20

didn't look as as nice as this

23:22

doesn't look as nice as this does

23:24

in that in other areas it's nice

23:26

as this does in that respect in

23:28

other areas it's a little bit less

23:30

in other areas it's a little bit

23:32

less there are less careful to not

23:34

position the camera in a way that

23:36

would reveal occlusion issues too much, but

23:38

at the edge of the screen you

23:40

see that kind of characteristic SSR occlusion

23:43

problem where it's kind of not appearing

23:45

there. But the kind of weird aspect

23:47

of it for me and an aspect

23:49

that I'm kind of throwing my hands

23:51

up and asking some questions about potentially

23:53

is image quality. Now image quality is

23:56

pretty good in this trailer. There are some

23:58

issues with like common pain points. the intersection

24:00

of hair and depth of field, stuff

24:02

like that. Not a big deal at

24:05

all. But I did get a couple

24:07

of pixel counts and they came in

24:09

close to 4K or at 4K, maybe

24:12

slightly under sometimes, but others quite close

24:14

to a 4K match. Those shots are

24:16

all in gameplay at 30 FPS and

24:19

again, the specific platform used to capture

24:21

gameplay was not delineated specifically. But certainly

24:23

I'm thinking like if it was using

24:26

RTGI on base PS5. maybe those numbers

24:28

would be lower, is my thinking, because

24:30

that's not typical we'd see a game

24:33

running at or near 4K with a

24:35

prominent ray trace lighting effect on PS5.

24:37

So that's kind of just throwing a

24:40

couple questions in there because we don't

24:42

know the specifics of what it's captured

24:44

on or if they're using the technique

24:46

in the first place or whatever, there's

24:49

some signs of it here and there,

24:51

but who knows? So yeah, I'm not

24:53

too sure about that or where it

24:56

quite points you, but I think one

24:58

thing you can be pretty confident about

25:00

is regardless of any techniques that may

25:03

or may not be being used there.

25:05

I think the game just looks tremendous.

25:07

I think it's going to be one

25:10

of the best looking current generation games

25:12

out when it arrives, I think on

25:14

June 25th, so it's coming up. I

25:17

see. Getting to your point at hand,

25:19

I, in terms of RTGI, it's really

25:21

hard to point out at this trailer

25:24

because so much of it is dedicated

25:26

to Cutscenes and in cutscenes you have.

25:28

Extremely way to just point lights, turn

25:31

on lights, use soft lights that don't

25:33

have a specular component, you know, there's

25:35

a lot of things you can do

25:38

in rasterization to in a cut in

25:40

a directed camera scenario like you were

25:42

just describing with the SSR where they

25:45

really try and avoid it. Like the

25:47

camera is always like here and the

25:49

camera's always here. It's the perfect place

25:52

for it to be. So they're doing

25:54

a lot there. And so they're doing

25:56

a lot there. since the trailer is

25:59

made up so like if I were

26:01

didn't want spoilers I would not watch

26:03

this by the way it's got so

26:05

much in it but uh yeah so

26:08

like I don't want to actually say

26:10

yes or no on that because so

26:12

much of it is based upon the

26:15

the trailers and also the landscapes of

26:17

a game like Death Stranding, the best

26:19

places that Raster GI works really really

26:22

well, even probe solutions, is rolling, undulating

26:24

hills on such. And the first game

26:26

was made up of a lot of

26:29

that. It wasn't nearly as well indirectly

26:31

let as I would say this was.

26:33

But we also saw improvements in indirect

26:36

lighting between Horizon Zero Dawn and Horizon

26:38

Forbidden West, you know. So I think

26:40

it is possible, even with a raster

26:43

GI solution, given the landscapes in this

26:45

game, to have a much more convincing

26:47

indirect lighting scenario without having to rely

26:50

on ray tracing. It really depends, you

26:52

know. But the resolution and the frame

26:54

rate presentation, I think, is interesting. The

26:57

first game, obviously on PS4 and PS4

26:59

Pro, 30 FPS title, and I would

27:01

be curious to see since we're seeing

27:04

this like this, I think you're seeing

27:06

if this is actually a 30 FPS

27:08

title. It is very, very good looking

27:11

in terms of like all this stuff

27:13

that they're showing here. We already saw,

27:15

especially since it's presented at Native 4K,

27:18

like your pixel counts show. We don't

27:20

know the actual platform PS5 could be

27:22

an umbrella term, but that would be

27:24

an interesting thing. It would allow them

27:27

to push certain things. much harder, especially

27:29

in terms of the way cutscenes are

27:31

directed. If you're not trying to achieve

27:34

a steady 60 FPS and cutscenes, you

27:36

can go crazy with things like how

27:38

close the depth of field is, how

27:41

expensive the depth of field is, things

27:43

like the hair, which is a really

27:45

big part of this. You know, having

27:48

multiple characters in a scene with complex

27:50

hair is expensive at 60 FPS, about

27:52

much more to about 30. You know,

27:55

I'm very... I'm actually very interested to

27:57

see what ends up coming out of

27:59

this. If it looks like this on

28:02

a base PS5 I would be astounded

28:04

on a lot of levels at 60

28:06

FPS. But at 30 FPS I think

28:09

it all seems a lot more realistic

28:11

and in which case I'd be curious.

28:13

about like resolution, DRS, anything like that.

28:16

And in the eventual, this is my

28:18

personal interest, first game was published side

28:20

separately, away from Sony on the PC

28:23

platform, and then kind of brought back

28:25

later with the directors cut. I would

28:27

be really curious to see what they've

28:30

been doing with Decima that might be.

28:32

goes beyond what they've shown in this

28:34

trailer. The first game didn't scale that

28:37

much higher on PC. Maybe this could.

28:39

Historically at least when Cojima was working

28:41

with Fox Engine, one of the things

28:43

that I really loved about MGS5 is

28:46

that they had the high setting that

28:48

they used on consoles primarily, but then

28:50

they had this very high setting which

28:53

pushed out like depth of field in

28:55

a really cool way, pushed SSR out

28:57

in a really good way. shadows and

29:00

stuff. I would like to see that

29:02

more because like the first game, like

29:04

on PC, it still had tons of

29:07

pop in on, you know, while it

29:09

was running at like hundreds of FBS

29:11

on like processors from back then. So

29:14

I would love to see if they

29:16

could push those things out further in

29:18

the eventuality of a PC release. But

29:21

yeah. Cool trailer still have no idea

29:23

what the hell's going on. I guess

29:25

that's fine for those who are into

29:28

this kind of thing. So yeah, everybody

29:30

seems very excited about some kind of

29:32

pseudo metal gear solid association. Like yeah,

29:35

the thing looks like metal gear and

29:37

the dude looks like snake. Yes, basically.

29:39

Yeah. The scalability thing is an interesting

29:42

question Alex because Sony now and you

29:44

know, Guilla Decima. does actually have a

29:46

reason to scale, that reason being PS5

29:49

pro. So you would hope that there

29:51

would be some sort of scalability built

29:53

in just to service the PlayStation 5

29:56

family. Similarly, if there are going to

29:58

be experiments happening or if indeed so.

30:00

is indeed planning its own handheld, they

30:02

would be looking to have some kind

30:05

of thoughts towards that in the future.

30:07

So yeah, maybe there will be some

30:09

sort of scalability to it for PC.

30:12

I'd hope so. Yeah, not really too

30:14

much more to say about this. Oliver,

30:16

you're saying that basically it's native 4K

30:19

on the areas you could discern, right?

30:21

So native 4K are very close during

30:23

the gameplay sections, like lots of fighting

30:26

and explosions, things like that you can

30:28

count. Interesting, yeah. I

30:30

mean, 30 FPS on pro would

30:32

solve a lot of those sort

30:34

of viability questions, right? Yeah, I

30:36

mean, I'd hope that they could

30:38

run into the higher

30:41

framer than PlayStation 5,

30:43

and definitely PlayStation 5

30:45

pro consoles, maybe use that

30:47

in-house gorilla, really impressive up sampling

30:49

there. Hopefully for this title, but I

30:51

don't know that it's present here. I

30:54

wasn't like it didn't look super like

30:56

that to me here It looked a

30:58

little bit sharper and maybe a little

31:00

bit noisier than that But that's that's

31:03

very preliminary given that we're looking at

31:05

just compressed YouTube footage But I wasn't

31:07

getting quite that impression, but perhaps

31:10

they will in the final title adopt that

31:12

who knows all of the Sony first party

31:14

games have had some sort of 60 FPS

31:16

support I believe yeah, so I'd say it's

31:18

probably going to happen for this

31:20

one. You'd think, but you know, 30

31:22

FPS would be fine as well

31:25

for, for, for presentational purposes and,

31:27

you know, certainly looks fantastic. Okay,

31:29

let's move on to the next

31:31

topic. Okay, so this week we

31:33

couldn't get a video of you

31:36

ready in time, but the Verizon

31:38

9, 9950 X3D has arrived, the

31:40

latest and greatest CPU from AMD,

31:42

and I have personal stakes in

31:44

this because I've bought one, and

31:47

here it is. Yeah,

31:50

it's time to upgrade my PC.

31:52

It's time to upgrade my

31:54

PC from the 13900K, which

31:56

has always been a bit

31:58

questioned for that. It used

32:00

to crash a lot, which

32:02

I thought was the issue

32:05

with the high voltage. The

32:07

Bios updates seem to improve

32:09

matters though. It's been relatively

32:12

stable. However, we're seeing big

32:14

performance increases and the thing

32:16

about this Verizon processor is

32:18

that It's the best of

32:21

both worlds, the best of

32:23

all worlds. You want a

32:25

high productivity-based processor for the

32:27

kind of work we do,

32:30

but at the same time,

32:32

you want the X3D cash

32:34

for the gaming side of

32:36

things. And with the previous

32:39

X3D, there's always been some

32:41

sort of. Gotcha, right, when

32:43

it comes to the 16

32:45

core parts. But Wills reviewed

32:48

it. He's put all the

32:50

benchmarks together. I've put some

32:52

notes together based on his

32:55

benchmarks and holy crap, I'm

32:57

glad I bought this thing.

32:59

It looks like a complete

33:01

beast. I just first of

33:04

all wanted to talk quickly

33:06

about video encoding. It seems

33:08

to be marginally faster than

33:10

the 9950X non-3D. which is

33:13

great, blows my 13900K out

33:15

of the water, which is

33:17

great, because video encoding is

33:19

one of the most CPU-intensive

33:22

tasks we do. But the

33:24

gaming side of things, the

33:26

9800X3D was kind of mind-blowing,

33:28

and so it is with

33:31

this. I'm just going to

33:33

go through some of the

33:35

games based on wheels benchmarks

33:38

here. Dragon's Dogmatoo. We've got

33:40

this custom run where we

33:42

just sort of circled around

33:44

the city. And it is

33:47

very CPU intensive. 32% faster

33:49

than the non-X3D product. Wow.

33:51

Which is nuts. This puts

33:53

it on par with the

33:56

9800X3D, 15% faster than a

33:58

14900K. There are a couple

34:00

of scenarios where this thing

34:02

is slightly slower than the

34:05

9800X3D. Starfield. 3D 4% faster,

34:07

but come on 4%. I

34:09

can trade that 4% for

34:11

8 more cores and 16

34:14

more threads for my production

34:16

work. And even so, they're

34:18

20% faster than the 9950X,

34:21

8% faster than the 14900K.

34:23

Mind-blowing. We've got our custom

34:25

run also here in Cyberpunk

34:27

2077 where we just circled

34:30

the cherry blossom marketplace and

34:32

this. This batters the CPU.

34:34

This is another example where

34:36

the X3D, the 8-core 1,

34:39

3% faster, so I don't

34:41

care, you know. 19% faster

34:43

than the non-3D model, 15%

34:45

faster than the 14900K. I

34:48

mean, only crap, this is

34:50

just astonishing. Alex, we have

34:52

to do this one, flight

34:54

simulator 2020. That run over

34:57

New York, which is basically

34:59

super super super CPU limited,

35:01

35% faster than the non-X3D

35:04

model. 35%. That puts it

35:06

on par with the 9800X3D,

35:08

32% faster than the 4900K.

35:10

There are certain games where

35:13

just having the 3D cash

35:15

just gives like a monumental

35:17

boost to performance. And I

35:19

think we can quite safely

35:22

say it's the ones that

35:24

aren't particularly cash-friendly, aren't particularly

35:26

optimal for a senior perspective.

35:28

And I also was quite

35:31

interested in wheels result for

35:33

Farcri-5 because that engine really

35:35

is still driven mostly by

35:37

a single core, but it

35:40

does seem to be the

35:42

case that again the X3D

35:44

makes a big difference. another

35:47

35% win over the non-X3D

35:49

model. The 9800x3D, the 8core

35:51

part, 3.5% faster, but again,

35:53

that's irrelevant. The 14900K comparison,

35:56

wow. 37.5% faster than a

35:58

14.900K, which in turn means

36:00

it's going to be like

36:02

wiping the floor with the

36:05

arrow lake parts. Yeah. Yeah.

36:07

Final one, which kind of

36:09

blew my mind a bit,

36:11

almost to the point where it's,

36:14

well, the first of all of

36:16

benchmarking is if something looks

36:18

wrong, you kind of think it

36:20

is wrong. So maybe I'd like

36:23

to retest this, but it actually

36:25

thinking about it sort of makes

36:27

sense. Boulder 3, Act 3, Alex.

36:29

We've got that custom

36:32

run through the city

36:34

of Boulder's Gate. 48%

36:36

faster than the non-X3D

36:39

model. The reason why

36:41

I actually think the

36:43

benchmark is sound is

36:46

because it is on

36:48

par with the 9800X3D.

36:50

43% faster than 4900K.

36:53

So, Oliver, sitting there

36:55

with your 14700K, A 13700K,

36:57

not even as good as

37:00

that. Yeah, you've been thinking

37:02

about upgrading for a while.

37:04

Yeah, they have. Well, I've

37:06

been contemplating upgrading different computers

37:08

at different points this year,

37:11

but I'm trying to. trying

37:13

to restrain myself to some

37:15

degree. Yeah. But this seems,

37:17

I mean, this seems really

37:20

compelling. It's within a margin

37:22

of error within any 100X3D,

37:24

which is obviously the fastest

37:26

gaming processor by some margin. And

37:28

it basically seems to me really

37:31

on par with 1950X, which was

37:33

already pretty much the fastest content

37:35

creation processor in the business. Yeah.

37:37

For almost anything. So, and they

37:39

also seem to have sorted out

37:41

all the CPU scheduling. that they

37:43

had issues that they had with

37:45

gaming tasks not consistently running on

37:47

the V cash CCD, that seems

37:49

sorted out as well. And they

37:51

also say you don't need to

37:53

reinstall windows between profit or swaps

37:55

necessarily, which is a nice boost

37:57

for ease of use and I'm

37:59

sure for. testing eventually as well.

38:01

So it seems like a really

38:03

strong product, best of all worlds,

38:05

almost any production relevant workload. It's

38:07

the fastest CPU available for Windows

38:10

users. So that's pretty pretty promising.

38:12

I would be very curious to

38:14

see a comparison between this and

38:16

something like the Apple M3 Ultra,

38:18

which I think is probably faster

38:20

in a lot of circumstances. That

38:22

would be interesting. Although I don't

38:24

know if you consider that consumer

38:27

level necessarily because Apple doesn't really

38:29

have a... super delineated pro and

38:31

consumer line necessarily, but that CPU

38:33

is obviously very expensive. So yeah,

38:35

I think this is a really

38:37

promising CPU. It's certainly something I

38:39

consider if I was going to

38:42

get a gaming computer upgrade, but

38:44

I think now I would probably

38:46

gravitate towards the 900x3D, just because

38:48

if you're doing testing on it,

38:50

you know 100% that it's going

38:52

to operate. within the parameters you

38:54

probably think it will. There's no

38:56

dual CED stuff, so I would

38:59

just like to eliminate that one

39:01

source of like testing uncertainty. I

39:03

know in the bios you can

39:05

disable the second CCD if you

39:07

want to, but that seems like.

39:09

But nobody's ever going to do

39:11

that really. If it was just

39:14

it was just for casual use,

39:16

then I think the 900X30 or

39:18

the 9950X3D rather is a really

39:20

compelling CPU, but if you're doing

39:22

like testing on it, I think

39:24

900X3D is just a bit more

39:26

reliable in that respect. Alex thoughts.

39:28

I'm just listening for the crickets

39:31

coming from Intel here and they

39:33

don't got anything man, got nothing.

39:35

It's crazy. The AMD dominance is

39:37

glorious. I really love the X3D

39:39

chips, like Rich, I switched from

39:41

a 12900K and for me there,

39:43

like the aspects of the 12900K

39:46

and that system that I really

39:48

like, but for me, actually the...

39:50

The E-cores were always the biggest

39:52

problem, kind of like the 7950

39:54

X-3D parts, etc., where you wouldn't,

39:56

with the Gremlin's... the system, you

39:58

were sure if you were getting

40:00

the proper performance you should be

40:03

getting due to the fact that

40:05

scheduling was maybe not prioritizing the

40:07

right cores. And that really pushed

40:09

me away from that system. There

40:11

were so many times where I

40:13

would go in a game and

40:15

I'd just be like, I'm losing

40:18

performance by having e-cores on. I'm

40:20

losing performance by having e-cores on.

40:22

And that is, if they've really

40:24

cleaned this up based on Will's

40:26

benches, like that they show off,

40:28

this minor is 3 to 4%

40:30

lower than 9x3D is like you

40:32

said, completely irrelevant. No one cares.

40:35

And if they really did clean

40:37

that up, then this is a

40:39

great multi-purpose processor because you're getting

40:41

that extra width when you really

40:43

need it for the task that

40:45

we do. I'd be very curious

40:47

to see how it scales also

40:49

in like crazy recording casts. 8k

40:52

60 was like super high quality

40:54

like we have the ability to

40:56

do that or even you know

40:58

like 4k 120. Yeah that's a

41:00

good point yeah yeah we'll get

41:02

on to the platform in a

41:04

second Alex yeah and and I

41:07

was very curious about all those

41:09

things also the boards that you

41:11

attach with it I'm always really

41:13

curious about like how many PCIE

41:15

lanes this thing would have because

41:17

I extensively use all of them.

41:19

So yeah that's all these kind

41:21

of things and I think this

41:24

sounds like an amazing product. What

41:26

is the retail MSRP? Oh well

41:28

this is where we start going

41:30

into you know into Hebe Dragons

41:32

territory. Because the, well, I understood

41:34

that this thing in the UK

41:36

costs £659, right? That'd be good.

41:39

Which is like 200 more than

41:41

the 9800X3D, which I think is

41:43

actually a good price. I ended

41:45

up paying 700 pounds from a

41:47

highly reputable supplier in the UK,

41:49

simply because all of the other

41:51

suppliers in the UK were charging

41:53

700 pounds for it. So, yes,

41:56

I was scalped, buyer retailer. And

41:58

that's unacceptable. And, you know, again,

42:00

you know, what's going on in

42:02

the PC hardware market is, it's

42:04

kind of like... Well, I guess

42:06

you could call it capitalism and

42:08

the laws of supply and demand.

42:11

But, you know, it doesn't seem

42:13

to happen with other areas of

42:15

consumer electronics, right, in terms of

42:17

anything that isn't a PC part

42:19

specifically. You know, if I want

42:21

to buy a Samsung S25 Ultfor,

42:23

I go to the Samsung website,

42:25

I see the MSRP there and

42:28

I buy one, you know, if

42:30

I want a new Mac. same

42:32

difference, right? If I want a

42:34

PlayStation, it's very rare that I

42:36

go to Amazon and see Amazon

42:38

trying to rip me off, you

42:40

know, what's going on here? This

42:43

has got to be sorted out.

42:45

It's completely nuts. Let's talk about

42:47

the platform, Alex, because yes, I

42:49

do want to actually work with

42:51

this thing. That's the key point

42:53

of it. It's a personal PC,

42:55

but I am going to be

42:57

using it for my workstation. And

43:00

yeah, there's a problem with PCIE

43:02

lanes that hasn't been solved since

43:04

the end of the high-end desktop

43:06

line. And the, I've got an

43:08

Aseuss X67E, X670E, pro art, create

43:10

a board. Oh, that's a good

43:12

one. Yeah, it's, it's got, basically,

43:15

the lanes can be quite intelligently

43:17

divided, so you have two Aetex

43:19

lanes, if you want them, across

43:21

two slots, and then a four

43:23

X lane that comes off the,

43:25

the, the chipset. So in theory

43:27

then. I can have my GPU

43:29

on one of the X8 slots,

43:32

and I can have the 8K

43:34

capture card on the next X8

43:36

slot, it leaves an X8 slot.

43:38

And then we can have a

43:40

4K card on the X4 slot.

43:42

So that that should work. I

43:44

think that's how you've got things

43:46

set up on your system, isn't

43:49

it? Yes, it is. It's an

43:51

older board. It's PCIEE4, and then

43:53

also the bottom can be for,

43:55

but also that limits. I don't

43:57

know what it is like. on

43:59

yours, but that limits how many

44:01

MVMEs I can throw in the

44:04

system? Yeah, I think I can

44:06

have a maximum of four. Okay,

44:08

see, I have a maximum of

44:10

two with the setup on mine.

44:12

That is unacceptable. But as a

44:14

result, they're very high capacity ones,

44:16

and then the rest is going

44:18

through SATA on my end for

44:21

like longer cold storage kind of

44:23

stuff. but not really cold, it's

44:25

warm. But yeah, that's what I'm

44:27

always serious about. Yeah, it's happened.

44:29

It's an SSD Saturday. They don't

44:31

get that hot anymore. But the,

44:33

yeah, that's the always thing for

44:36

my like my buying preferences for

44:38

these things, which really to look

44:40

for it, because like lately also,

44:42

I updated the bios on my

44:44

board and I've been struggling with

44:46

PCI since then a little bit

44:48

occasionally. Some gremlins that I'm trying

44:50

to chase. And if I just

44:53

had enough from the get go,

44:55

from the get- from the get-

44:57

from the get-go, I wouldn't- chasing

44:59

these things anyway. So that would

45:01

be my upgrade direction to ensure,

45:03

because I use like Rich, I

45:05

use a 49D as the GPU,

45:08

then depends on the capture card.

45:10

it depends for the like 4k

45:12

60 or 4k 121 depends on

45:14

what I'm capturing because I'll use

45:16

the one that has more flexibility

45:18

with HDR for example as well

45:20

as work it works better with

45:22

my PlayStation 5 pro and then

45:25

on the bottom of that I

45:27

have an analog capture card which

45:29

is 4x and definitely uses that

45:31

when I'm doing something like 1080P

45:33

and I want to use RGB

45:35

with that actually because low resolution

45:37

content You can't really be decimating

45:40

the chroma on it because it

45:42

looks really really bad. So when

45:44

you do your zooms and stuff

45:46

especially. Yeah, so you really don't

45:48

want to do that. Depending about

45:50

also what I'm recording, I want

45:52

more than 422 or 420. I

45:54

want 444 occasionally. And these are

45:57

things that I need all the

45:59

PCIEE for. And that's not a

46:01

guarantee. And your platform sounds great

46:03

though. Well, fingers cost it all.

46:05

works. Yes, that's all I can

46:07

really say. I'm just kind of

46:09

baffled on my main test system

46:12

for GPU reviewing. It uses a

46:14

9800x3D. High-end-a-soosboard. It's got two PC-I-E

46:16

slots. I mean if

46:18

I'm buying a high-end-a-sus-board

46:20

I'd kind of like more than

46:22

two PC-I-E slots. It's kind

46:25

of a bit bizarre right? you know,

46:27

that first memory training you're going to

46:29

have when you turn on the board.

46:31

Yeah, it is a little scary because

46:34

you're not sure what's going on. So

46:36

get used to that. And also get

46:38

used to, since this is your first

46:40

like, I think it's the first time

46:42

you'd be using a Verizon platform for

46:45

your main editing rig, right? Yeah. The

46:47

memory training as well as sending out

46:49

whether or not to use memory

46:51

context restore, I've had less luck

46:53

with it than other people. I

46:55

would say what is it first

46:57

of all? Okay so memory conduct

46:59

restore remembers essentially the settings that

47:02

the memory training does from like

47:04

the previous boots so it doesn't

47:06

have to do it to the

47:08

same degree but the problem is

47:10

I've had boot errors when that

47:13

thing starts it doesn't get to post

47:15

ever if I use it. That is

47:17

problematic but maybe you're luck

47:19

gear than me. Okay. Well, I'll

47:21

keep everyone posted there. GPU-wise, I

47:23

currently use a 4080, which I'm

47:26

quite happy with. The main reason

47:28

I use the 4080 is that

47:30

there's, you know, wasn't useful for

47:32

anything else. You know, it was

47:35

that weird kind of in-between a

47:37

GPU. Obviously, for most of our

47:39

sort of high-end testing, we use

47:42

the 1490. I guess I could use

47:44

a 4090 now, the 5090, but yeah,

47:46

I'll keep your posted. Anyway. That's going

47:48

to be fun to build. But

47:50

until then, let's move on.

47:52

Interesting news story this week.

47:54

It's been rumored for some time

47:57

that oblivion is actually going to

47:59

receive... a remake, a remaster perhaps,

48:01

and it's going to be running

48:03

on Unreal Engine 5. Lots to

48:05

talk about on this one, Alex.

48:07

I'm good. Digital Foundry Direct Weekly

48:09

is presented by Sega and Atlas'

48:11

Metaphor R. Fantasio, IGN's 2024 Game

48:13

of the Year, racking up over

48:15

50 awards including Best RPG, Best

48:18

Narrative and Best Art Direction, at

48:20

the Game Awards 2024. From the

48:22

creators of persona 3, 4, and

48:24

5. This is Atlas' first fantasy

48:26

JPRPG in the AAA space. Step

48:28

into a unique world where you

48:30

join a royal tournament to claim

48:32

the throne. but here's the twist.

48:34

The late king declared the next

48:36

monarch must be elected by the

48:38

people. It's not just about strength,

48:40

you've got to earn their thrust.

48:42

Manage your time with a calendar-based

48:44

system choosing how to spend your

48:46

days. Questing, dungeon diving or boosting

48:48

your royal virtues. With a free

48:51

demo offering around five hours of

48:53

gameplay you can dive in now

48:55

and carry your save on to

48:57

the full game. Metaphory Fantazio is

48:59

out now on Xbox PlayStation PlayStation

49:01

and Steam. Buy it now at

49:03

25% off for a limited time.

49:05

The venture to suggest it's going

49:07

to be one of these things

49:09

similar to the way Blue Point

49:11

handles Remasters or the way I

49:13

think the new Metal Gear Delta

49:15

works, where Unreal Engine is essentially

49:17

a wrapper around the original engine.

49:19

Possibly that would make sense. I

49:21

just can't imagine Unreal Engine 5

49:24

being able to do things that

49:26

the original Bethesda engine does. Yeah,

49:28

I'd agree with that. All the

49:30

things, they would have to essentially

49:32

reprogram everything into Unreal engine five.

49:34

And it seems like an absolute

49:36

waste of time when the game

49:38

play code is already there. And

49:40

you could almost like RTX remix

49:42

essentially replace the rendering. That is

49:44

much more doable. I'd be very

49:46

curious though what that CPU workload

49:48

looks like. I would be very

49:50

curious to what that looks like.

49:52

I think this is a really

49:54

cool idea because there's a really

49:57

great opportunity. here to fix aspects

49:59

of oblivion along with giving people

50:01

access to a game that, you

50:03

know, a lot of people are

50:05

younger and maybe didn't play it

50:07

on Xbox 360 or PS3 or

50:09

PC back in the day. And

50:11

this is a really great opportunity

50:13

to play it again and maybe

50:15

go back in and tweak some

50:17

things. On the visual side, I

50:19

think what they need to do...

50:21

is ensure that is very bloomy

50:23

and strange looking just like the

50:25

original. Is that a technical term?

50:27

That is a very technical term.

50:30

Bloomy and strange. I think that's,

50:32

there's a white paper on that

50:34

one. And the, they also, you

50:36

know, water looks, needs to look

50:38

great. It needs to be wonderfully

50:40

green. It's got to have those

50:42

teals, everything, you know, there's, there's

50:44

a certain look that I think

50:46

they abandoned after. That when they

50:48

moved to Skyram where it was

50:50

much more earthen toned and Oblivion

50:52

almost looks like a fantasy novel

50:54

cover that you'd pick up at

50:56

like a like a one dollar

50:58

store It's very very very particular

51:00

look that I hope they maintain

51:03

On the gameplay side I would

51:05

be really excited to see them

51:07

get it get rid of auto

51:09

leveling of the world alongside you

51:11

I always thought that took out

51:13

the entire RPG element of the

51:15

RPG it was like you would

51:17

you would get better but then

51:19

your enemies would get better so

51:21

you the sense of progression was

51:23

limited to abilities which just kind

51:25

of scaled in a way that

51:27

wasn't ever satisfying. People like to

51:29

farm in our refugees you couldn't

51:31

really farm in oblivion other than

51:34

getting a bunch of stuff. It's

51:36

also an opportunity to revisit some

51:38

of the the dungeons. There's a

51:40

lot that can be done here,

51:42

but there's only so much that

51:44

can be done at the same

51:46

time if they do end up

51:48

using the gameplay code from back

51:50

then. I'd be curious to see

51:52

if the game actually does run

51:54

above 60 FPS without causing big

51:56

issues. The original game has big

51:58

issues above 60 FPS. I'd be

52:00

curious to see also. If they

52:02

would keep in the junkie animations

52:04

from the original game, or if

52:07

that would be, that would be

52:09

a different thing entirely. I don't

52:11

know what is technically possible. This

52:13

would be the first time, one

52:15

of the first times we've seen

52:17

something scaled from this generation of

52:19

hardware in this manner. So there's

52:21

a whole lot of open questions

52:23

I've had, but apparently it's also

52:25

like imminent, like it might be

52:27

shadow dropped. Yeah, yeah, this is

52:29

the report that's apparently from noted

52:31

leeker, Nate's to hate. I say

52:33

leeker, basically he's report, this is

52:35

the thing that kind of I

52:37

find quite frustrating is, is totally

52:40

irrelevant, but he's not the leek,

52:42

he's reporting a leak. He hasn't.

52:44

He didn't leak it. He doesn't

52:46

work for Bethesda or whoever. But

52:48

yeah, anyway, there is an interesting

52:50

leak here saying that yes, there'll

52:52

be some marketing soon and then

52:54

the game will be available very,

52:56

very soon after, which I'm quite,

52:58

you know, I'm on board with.

53:00

There's also discussion on what's going

53:02

to be changing. Six-v-worked systems, according

53:04

to this video games conical report,

53:06

which includes stamina, sneaking, sneaking. Blocking,

53:08

aren't you? He hits the action

53:10

and HUD. Yeah, improvements to the

53:13

game's blocking system have reportedly been

53:15

inspired by souls like games, while

53:17

sneak icons are now highlighted in

53:19

the new system according to MP1ST.

53:21

So it does look as though

53:23

they are going to be making

53:25

some gameplay changes at least. It

53:27

seems to be that Virtuos is

53:29

working on it, which they've got

53:31

a bit of a spotty track

53:33

record. Just thinking back to that

53:35

45 FPS cap on the arch

53:37

of asylum. Yeah, yeah, choice, necessarily.

53:39

All right. Any thoughts on this

53:41

one, Oliver? This project has been

53:43

like rumored for so long now.

53:46

I think it initially leaked out

53:48

in the ACTA. lizard court documents

53:50

with the Microsoft acquisition. Way back

53:52

when that previewed the original design

53:54

for the next generation Xbox and

53:56

other things, but here it sounds

53:58

like the graphics are going to

54:00

be probably fully remade in UE5.

54:02

I have seen some fan efforts

54:04

that I've taken the original models

54:06

and environment and things like that.

54:08

from oblivion into unreelension five and

54:10

it actually looks pretty all right

54:12

there with the improved lighting and

54:14

things like that. But I'd presume

54:17

here it will be a more

54:19

substantial effort given the time and

54:21

amount of money they presumably poured

54:23

into this. But I think it

54:25

would probably still have some of

54:27

the same limitations as the original

54:29

just in terms of the way

54:31

the game is structured like I'd

54:33

presume the game is still going

54:35

to be segmented in the same

54:37

way for gameplay purposes like. having

54:39

the cities as separate areas, having

54:41

individual in shops, things like this,

54:43

buildings, yeah, caves, things like that

54:45

in separate areas. I'd hope though

54:47

that there are some gameplay systems

54:50

that could touch up, like Alex

54:52

said, that level scaling was kind

54:54

of awkward. And that was actually

54:56

a problem in oblivion. It wasn't

54:58

just like an issue in terms

55:00

of game player progression. It was

55:02

also an issue in terms of

55:04

having NPCs die very easily because

55:06

the NPCs wouldn't scale and level

55:08

in the same way that the

55:10

enemies did. So if you were

55:12

trying to complete some end game

55:14

missions, but you were already a

55:16

very high level, like you completed

55:18

a lot of the side content,

55:20

your companions would just get one

55:23

shot by these demons. It was

55:25

quite bad. So yeah, my own

55:27

personal experience with Oblivion, I got

55:29

it on day one on both

55:31

360 and PC. I was super

55:33

excited for that. Although I remember

55:35

playing it on PC and actually

55:37

being more. I'm excited with the

55:39

visuals, even though I had a

55:41

weaker PC than on my Xbox,

55:43

just because I think blown up

55:45

to a 40-inch TV, the game

55:47

did not look quite as appealing

55:49

to me as it did on

55:51

my PC. I also probably had

55:53

better anisotropic filtering setting on PC

55:56

at that time. There's still some

55:58

elements of it that look visually

56:00

pretty awkward, even though at the

56:02

time it looked pretty good, like

56:04

the billboard trees. the pretty limited

56:06

use of shadow maps. And yeah,

56:08

that bloom effect. I'm not so

56:10

keen on that. I love it.

56:12

I thought a huge fan of

56:14

it. And also like there are

56:16

some weird quirks that I don't

56:18

know if they're gonna overcome these

56:20

or address them in some way,

56:22

but like the creation engine terrain

56:24

system. was pretty limited as far

56:26

as I recall, like you had

56:29

these rolling hills and nothing more

56:31

kind of angular or overhanging than

56:33

that in terms of the terrain.

56:35

That was also an issue in

56:37

Fallout 3 and Fallout New Vegas

56:39

and I think it was overcome

56:41

to some degree by the time

56:43

Skyman came out it looked a

56:45

lot better in that respect at

56:47

least. There's also issues with the

56:49

VO like famously the male VO.

56:51

was mostly done by one voice

56:53

actor for the non-named NPCs, or

56:55

not the non-named NPCs, but just

56:57

the less prominent NPCs. You had

56:59

like board Patrick Stewart in there,

57:02

but you didn't have a whole

57:04

lot of variants between like the

57:06

male orks and the male, you

57:08

know, humans of different kinds and

57:10

things like this. That was not

57:12

so good. But yeah, there are

57:14

number of areas they could touch

57:16

on, but also there's a certain

57:18

charm to oblivion, and a certain

57:20

charm to how like broken the

57:22

radiant, the radiant AI could be

57:24

radiant AI could be. and certain

57:26

charm to the voices and charm

57:28

to board Patrick Stewart and charm

57:30

to like the way the faces

57:33

look so awkward. I don't know

57:35

what elements necessarily that you want

57:37

to improve in a really wholesale

57:39

way to get it up that

57:41

modern experience or if you want

57:43

to kind of leave it in

57:45

the middle a little bit and

57:47

have some of that kind of

57:49

rustic elements still in the mix

57:51

potentially for for players. I'm reminded

57:53

of Star Trek the animated series

57:55

Alex where basically Scotty and Nurse

57:57

Chapel did all the other voices.

57:59

They all sound like variations have

58:01

the same human. Okay well it

58:03

sounds as though we should be

58:06

finding out about this sooner other

58:08

than later if these rumors are

58:10

to believe so yeah I guess

58:12

we'll be on top of that

58:14

if and when it emerges. With

58:16

that let's move on. So I'll

58:18

find on used topic of... the

58:20

week and this was a report

58:22

that was sourced by Tom Warren

58:24

at the verge where a video

58:26

emerged of Sony's experimentations with AI

58:28

conversations with characters in games. We've

58:30

seen it all before with Envideo's

58:32

Ace. character. And Alex, I don't

58:34

know if you saw that video,

58:36

but it was basically exactly the

58:39

same, right? As Ace. Well, a

58:41

lot less polished, I would definitely

58:43

say. In terms of Asian fashion.

58:45

Yeah. Yeah. Yeah. Aloy had some

58:47

very interesting ticks going on there,

58:49

which didn't quite match her character.

58:51

No, didn't. I think this is

58:53

a... both a good and a

58:55

very bad thing at the same

58:57

time. It's both it's good in

58:59

the aspect that you can imagine

59:01

in a game that you can

59:03

have more dynamic branching conversational options

59:05

with non-quest like central non-main quest

59:07

at PCS. Presuming the actors sign

59:09

off. or maybe there's just fully

59:12

generated voices. You know, the big

59:14

deal, the big problem I have

59:16

with this is that they use

59:18

Aloy here and there's a very

59:20

specific person who voices Aloy and,

59:22

you know, their creative artistic work,

59:24

their licensing their voice for the

59:26

product, and, you know, there's just

59:28

like in the movie industry. There

59:30

has to be a way to

59:32

have your image, your voice, your

59:34

likeness be almost your own thing

59:36

that you're selling and you get

59:38

that money after the fact when

59:40

they do reruns of a show

59:42

or they re-release a show. And

59:45

that's something that needs to be

59:47

kind of thought out in the

59:49

system if we're going to be

59:51

using AI-generated likenesses and voices for

59:53

games. And that's the one area

59:55

that I'm... I think is really

59:57

morally gray and not good. But

59:59

in terms of what it could

1:00:01

mean from a gameplay perspective, it

1:00:03

could be enriching to an experience

1:00:05

if it is done well and

1:00:07

not awkward and there isn't tons

1:00:09

of delay and all these other

1:00:11

things like we saw with cloud

1:00:13

gaming where you can imagine experience

1:00:15

but it's a matter of getting

1:00:18

there with all these steps in

1:00:20

terms of latency. and believability. So

1:00:22

you're imagining a scenario there that

1:00:24

is completely odds with the video

1:00:26

that we saw. It is complete.

1:00:28

Yeah, that's, I was being very

1:00:30

generous. Did not look good at all.

1:00:32

Yeah, and also like, this is, this,

1:00:34

the character I would have used for,

1:00:37

it's really bizarre that they chose Aloy.

1:00:39

There are tons of other like

1:00:41

side characters in games that would probably

1:00:43

be way better fit. I don't know.

1:00:45

Right. What do you think? I'm very

1:00:47

interested in all of his take on

1:00:50

this bearing of ideas on our resident

1:00:52

master of AI. I'm very split on

1:00:54

this. I'm kind of split on this because

1:00:56

I think one issue just comes down to

1:00:58

the actual quality of the demo. It's kind

1:01:01

of low quality and a little

1:01:03

bit bizarre. It doesn't make sense

1:01:05

fundamentally within the context of the

1:01:07

games or anything related to the

1:01:09

games that Aloy would just be

1:01:11

sitting there addressing the player. or some

1:01:13

stranger in the game with basically a

1:01:15

Wikipedia synopsis of the game's plot. And

1:01:17

also in the context of a game,

1:01:20

the player character would never ask the

1:01:22

question, I have some questions about horizon.

1:01:24

Like the players in the game do

1:01:26

not think of their world as horizon.

1:01:28

It's just a marketing term for the game.

1:01:31

You know, I mean, that's so bizarre

1:01:33

to me. Also, the dialogue is not

1:01:35

written by, like, alloy dialogue would be,

1:01:37

the quality of the voice synthesis is

1:01:39

very low. It doesn't sound like the

1:01:41

actor at all, Ashley Birch, either. Obviously,

1:01:43

there are some issues, obviously, that Alex

1:01:45

has talked about with using a voice

1:01:47

clone of the actor. without some kind

1:01:49

of arrangement potentially but it doesn't sound

1:01:51

like her and that's very off-putting because

1:01:53

this is a distinctive character but it's

1:01:55

a very generic voice. The automated lip

1:01:57

sync is actually pretty good here but the

1:01:59

rest of the face... animation is otherwise very

1:02:01

static and a little bit awkward

1:02:03

like the eyes are not always

1:02:05

looking in the direction of the

1:02:08

player it looks a little strange

1:02:10

there. The tech stack they're using

1:02:12

which apparently is GPT4 whisper and

1:02:14

llama 3 in some capacity plus

1:02:16

something I'm sure to drive the

1:02:18

animations is not working at the

1:02:20

level that it really needs to.

1:02:22

And I think we have such

1:02:24

high expectations for the characterization of

1:02:27

game characters. Like if you look

1:02:29

at what We as humans see

1:02:31

as high quality plausible representations of

1:02:33

characters in games. We're looking at

1:02:35

titles like Untarted 4 or Hell

1:02:37

Blade 2. These titles that look

1:02:39

really really incredible have a lot

1:02:41

of very careful hand-done work or

1:02:44

these carefully curated cinematic experiences totally

1:02:46

like this. And even relative to

1:02:48

some like of the invidious ace

1:02:50

demos who've seen and stuff like

1:02:52

that, including at CS 2025, this

1:02:54

is not particularly close to that

1:02:56

like And also the video demos

1:02:58

aren't necessarily there in terms of

1:03:00

quality either. But on the other

1:03:03

hand, and this is where I'd

1:03:05

be more positive about it potentially,

1:03:07

is that I think there is

1:03:09

a role for technology like this

1:03:11

in games, especially games that are

1:03:13

built around this tech where you

1:03:15

can maybe see past some limitations

1:03:17

that are kind of inherent to

1:03:20

the tech in the moment. I

1:03:22

think about the game facade, which

1:03:24

is a game I'm thinking a

1:03:26

lot about recently, which is a

1:03:28

game from 20 years ago where

1:03:30

basically you could type in your

1:03:32

response to characters at a dinner

1:03:34

party who are having some kind

1:03:36

of marital dispute. That's amazing. Yeah,

1:03:39

and it had this kind of

1:03:41

hand-coded AI system that could kind

1:03:43

of do a rough kind of

1:03:45

sentiment analysis on what you were

1:03:47

saying or respond to keywords. It

1:03:49

was very, very crude. But I

1:03:51

imagine the game like that nowadays,

1:03:53

nowadays, using... a system like this

1:03:55

could actually be very compelling and

1:03:58

maybe could give you a much

1:04:00

more dynamic gameplay experience. So I'd

1:04:02

really try to think about the

1:04:04

kinds of games and applications that

1:04:06

would enable experiences above and beyond

1:04:08

where games are now in terms

1:04:10

of that dynamism, in terms of

1:04:12

responding directly to like intimate and

1:04:15

intricate player feedback. That's the kind

1:04:17

of thing that seems really cool

1:04:19

to me. But it's like a

1:04:21

replacement for a voice actor in

1:04:23

a scripted narrative game. I don't

1:04:25

think it'll get to that level

1:04:27

any time soon. State of the...

1:04:29

voice synthesis isn't not very good

1:04:31

in my opinion and you know

1:04:34

I think for dynamic reactions or

1:04:36

dynamic dialogue it could have a

1:04:38

role but I think you always

1:04:40

want to have good acting in

1:04:42

crystal clear sound quality and good

1:04:44

sound quality that matters so much

1:04:46

in the game so I wouldn't

1:04:48

I wouldn't worry about that set

1:04:50

of it too much maybe in

1:04:53

some news people will try it

1:04:55

for that purpose but I'd hope

1:04:57

not too much there I just

1:04:59

think this is the kind of

1:05:01

thing that could enable it experience

1:05:03

that probably we can't even contemplate

1:05:05

in the current moment. Like I

1:05:07

really think it could enable something

1:05:10

that's really quite exciting. But not

1:05:12

in the context of like a

1:05:14

super narrative oriented game. I think

1:05:16

those will always exist. I don't

1:05:18

think it has much of a

1:05:20

place there, but I also don't,

1:05:22

I'm not sure that this demo

1:05:24

is like super intricate, super carefully

1:05:26

put together. I think it's just

1:05:29

something there is showing off behind

1:05:31

closed doors. It's kind of a

1:05:33

cruitable sample, but I don't think

1:05:35

this indicates that. Lots of thoughts

1:05:37

about this, where do I begin?

1:05:39

I've played these ace demos that

1:05:41

Envideo has done, right? And ultimately,

1:05:43

you're being asked to be sort

1:05:46

of like a highly, you know,

1:05:48

it's your agency being injected into

1:05:50

the story. The problem is, whenever

1:05:52

I've had these demos, you know,

1:05:54

you kind of go to the

1:05:56

demo and it's like, well, where

1:05:58

do you begin? What should I

1:06:00

say? What's going on? It's actually

1:06:02

quite difficult to get to grips

1:06:05

with it. And you have to

1:06:07

be deeply immersed in the law,

1:06:09

in the concept, in order, you

1:06:11

almost have to be a role

1:06:13

player as such, an actual role

1:06:15

player. And it just causes fiction,

1:06:17

right? You know, you're just standing

1:06:19

there wondering what to say for

1:06:21

the most part. And then basically

1:06:24

whoever's guiding you through the demo

1:06:26

does exactly that. They say, well,

1:06:28

okay, maybe you should ask them

1:06:30

about this. If you just have

1:06:32

a dialogue tree similar to what

1:06:34

you have now, you just get

1:06:36

through it a lot quicker. You

1:06:38

get reminded of your place. the

1:06:41

story and how it should be

1:06:43

progressing. Giving too much agency to

1:06:45

the player actually makes things somewhat

1:06:47

worse. Also the quality of the

1:06:49

interaction is always this sort of

1:06:51

lag between asking a question or

1:06:53

saying something and getting a reply.

1:06:55

It's just completely immersion-breaking because it

1:06:57

doesn't happen in real life and

1:07:00

that's essentially what they're trying to

1:07:02

do here to make conversations. closer

1:07:04

to real life, but the immersion

1:07:06

aspect of it's just kind of

1:07:08

broken. And I've been thinking a

1:07:10

lot about what this reminds me

1:07:12

of, and I sort of... I

1:07:14

said it to you the other

1:07:17

day, Alice. It's like the old

1:07:19

Frank phone calls with the Arnold

1:07:21

Schwarzenegger soundboard. where, you know, you're

1:07:23

desperately trying to come up with

1:07:25

phrases that fit the circumstances to

1:07:27

sort of make things happen. And

1:07:29

yeah, that's complete with the lag

1:07:31

as well, the all-important lag. Yeah,

1:07:33

the awkward pause when they have

1:07:36

to search for the... For the

1:07:38

correct, aadi sound quote. That's exactly

1:07:40

what it reminded me of. So

1:07:42

I don't really see how this

1:07:44

is going to have any sort

1:07:46

of long-term impacts on narrative-driven games.

1:07:48

But we've got some interesting supporter

1:07:50

questions on it, that's for sure.

1:07:52

This one from Hunter, would you

1:07:55

be open to having Aloy as

1:07:57

a guest on the podcast? I

1:07:59

realize she may not be particularly

1:08:01

well-versed in current technology, but maybe

1:08:03

if she shuffs her stick into

1:08:05

a 50-70 TI and points her

1:08:07

focus at it, she'll have something

1:08:09

relevant to say. Oh boy. This

1:08:12

question, while interacting with chatpots, that's

1:08:14

exactly it, a chatbot, while interacting

1:08:16

with chatpots like alloys, I'd rather

1:08:18

see AI take things a step

1:08:20

further, bringing MPCs and the world's

1:08:22

life in more dynamic, immersive ways,

1:08:24

instead of just reducing them to

1:08:26

basic chatpots. Imagine taking full... or

1:08:28

breaking a fence with your sword

1:08:31

and an MPC reacts dynamically to

1:08:33

quite a fall there? Or what

1:08:35

did that fence ever do to

1:08:37

you? The AI-driven system could generate

1:08:39

endless natural interactions, creating immersion that

1:08:41

studios couldn't afford to fully voice

1:08:43

act. Wouldn't this be more a

1:08:45

more compelling use of AI than

1:08:47

just another chatbot? What are your

1:08:50

thoughts? That's an interesting one. That

1:08:52

is. I just think AI is

1:08:54

going to be used much more

1:08:56

generally for more, you know, to

1:08:58

make NPC's. and like crowd behaviors

1:09:00

are a lot more realistic and

1:09:02

and more context-driven because you know

1:09:04

I think that's probably one of

1:09:07

the big weaknesses of MPCs in

1:09:09

a lot of the open-world games

1:09:11

for example you can do some

1:09:13

crazy stuff and they don't give

1:09:15

a toss. You know there's only

1:09:17

very limited interactions to or reactions

1:09:19

rather to your to your inputs

1:09:21

and it's always a bit immersion

1:09:23

breaking more natural behaviors generated by

1:09:26

AI doesn't need to be sort

1:09:28

of deep, you don't need to

1:09:30

be holding like deep dive conversations

1:09:32

with alloy, it just needs to

1:09:34

be, you know, more realistic reactions

1:09:36

in the game world to your

1:09:38

inputs. I think that would be

1:09:40

far more transformative. Yeah, any final

1:09:43

thoughts, Oliver? Yeah, I mean, this

1:09:45

demo is just a bad... venue

1:09:47

for showing off the best of

1:09:49

AI I am sure they really

1:09:51

just had this behind closed doors

1:09:53

never intended certainly for it to

1:09:55

reach the public indeed they pulled

1:09:57

the videos using a copyright claim

1:09:59

pretty quickly there so I don't

1:10:02

think they intended this as a

1:10:04

finished product at all. You really

1:10:06

need NPCs to be much more

1:10:08

context aware and a character just

1:10:10

mod logging to the camera is

1:10:12

pretty boring, indeed reading off of

1:10:14

Wikipedia to the camera. Again, a

1:10:16

pretty boring experience, but I think

1:10:18

there is probably room for interesting

1:10:21

experiences, maybe indie games, but are

1:10:23

built around this tech, potentially like

1:10:25

a facade. a game like I

1:10:27

mentioned. That could be fun, but

1:10:29

in the interim, most these applications

1:10:31

are not that interesting when they're

1:10:33

just bolted onto existing game concepts

1:10:35

or given very simple monologues. It

1:10:38

needs to be something a little

1:10:40

bit more interesting that I feel.

1:10:42

I'll tell you what Alex, how

1:10:44

about this, an AI-driven Arnold Schwarzenegger

1:10:46

soundboard? Are we killed for it?

1:10:48

Yeah. This episode is brought to

1:10:50

you by Progressive Insurance. Do you

1:10:52

ever find yourself playing the budgeting

1:10:54

game? Shifting a little money here,

1:10:57

a little there, and hoping it

1:10:59

all works out? Well, with the

1:11:01

name your price tool from Progressive,

1:11:03

you can be a better budgeter

1:11:05

and potentially lower your insurance bill

1:11:07

too. You tell Progressive what you

1:11:09

want to pay for car insurance,

1:11:11

and they'll help you find options

1:11:14

within your budget. Try it today

1:11:16

at progressive.com. Progressive Casual Casualty Insurance

1:11:18

Company and Affiliates. Price and Coverage

1:11:20

Match Limited by state law. Not

1:11:22

available in all states. With that,

1:11:24

let's move on. Okay, let's move

1:11:26

on to Support a Q&A. Every

1:11:28

week I post on our paper

1:11:30

and asking our supporters to, you

1:11:33

know, basically pose their questions and

1:11:35

respond to our news topics. And

1:11:37

yes, I choose the best of

1:11:39

other ones we're best equipped to

1:11:41

answer. Gonna start off with a

1:11:43

couple of questions which, yeah, we're

1:11:45

gonna sort of group them together.

1:11:47

Basically, there was quite a big

1:11:49

reaction to our story last week.

1:11:52

where essentially Mark Serti asked, answered,

1:11:54

rather, a bunch of questions we

1:11:56

had about Project Amethyst and the

1:11:58

relationship between PSSR and FSR4. The

1:12:00

upshot being that next year we

1:12:02

should expect to see a variation,

1:12:04

a re-implementation of FSR4, for PlayStation

1:12:06

5 Pro, and that will be

1:12:09

the next evolution of PSSR. Lots

1:12:11

of people basically saying, well, you

1:12:13

know, PSSR's dead. FSR 4 was

1:12:15

taking over, it's a disaster for

1:12:17

Sony, quite baffling reactions I have

1:12:19

to say. Let's talk about these

1:12:21

questions though. Bespoke exclamation point asks,

1:12:23

crazy how even supposedly respectable journalists,

1:12:25

like a not-to-be-named, the verge reporter,

1:12:28

tried twisting what Mark Surdi said

1:12:30

about PSR, into quote-unquote, it fails,

1:12:32

and now we are switching to

1:12:34

FSR-4. The way I see it,

1:12:36

FSR3 equals DLS-3 equals DLS-1. FSR

1:12:38

4, it's better than DLS3, PSSR2

1:12:40

equals DLS3. I'm not sure that's

1:12:42

my time. Hierarchy is interesting. It's

1:12:44

an interesting take. I think that

1:12:47

by the time we get DLS5,

1:12:49

FSR5, 6, BSS34, the super resolution

1:12:51

part of these upscalers will be

1:12:53

indistinguishable from native. No real question.

1:12:55

Don't know why I bothered them.

1:12:57

Just making you an observation. But

1:12:59

can we give a heck yes

1:13:01

for Oliver Seemingly moving above ground?

1:13:04

You're always above ground, Oliver. I

1:13:06

don't think that's... I think so.

1:13:08

You just record at nights, don't

1:13:10

you? That's the bottom line. I

1:13:12

do. It's what, 2 AM here?

1:13:14

3 AM here, quite late. Oh

1:13:16

my God. Anyway, we got this

1:13:18

question and it is actually a

1:13:20

question from Dajarco in brackets Dan.

1:13:23

Happy half not, happy half fortnight

1:13:25

gents, exploration point, been reading Sereni's

1:13:27

comments closely and comparing them to

1:13:29

the stated figures for the RDLA4

1:13:31

GPUs. It seems that in order

1:13:33

to compare RDLA4 GPUs correctly with

1:13:35

the pro, we was determined that

1:13:37

the quote unquote dense tops for

1:13:40

int eight are what's the dense

1:13:42

tops are rather. According to the

1:13:44

9,000 series wiki page and a

1:13:46

few other sorts. 90-70 has 289

1:13:48

in-8 dense tops and the 90-70

1:13:50

XT has 389 in-8-dense tops. It

1:13:52

would seem that all the AMD

1:13:54

marketing for the new GPUs uses

1:13:56

the comically inflated in-4 sparse figures

1:13:59

with certainly also stating that sparsity

1:14:01

is not especially important for the

1:14:03

kind of models that they are

1:14:05

using for PSSR, it would seem

1:14:07

as though the PS5 pro sits

1:14:09

between the new AMD mid-range cards

1:14:11

in terms of dense in eight

1:14:13

tops performance. This is getting dense,

1:14:15

isn't it? It also seems conceivable

1:14:18

that FSR 4 would need to

1:14:20

be performing on the rumored lower

1:14:22

stack cards, i.e. the 1950-slifty slash

1:14:24

1960. In light of all of

1:14:26

this, all of this. Have we

1:14:28

underestimated PS5 Pro or have Sony

1:14:30

pissed on their chips? Okay,

1:14:33

so let's talk about first

1:14:35

of all, I mean, first

1:14:37

of all, we've got this

1:14:39

question of timelines about the

1:14:42

relationship between sodium AMD and

1:14:44

PSR and FSR4. Now, correct

1:14:46

me if I'm wrong, Alex.

1:14:48

Right. PSSR has been in

1:14:51

development with Sony for

1:14:53

like years. Right. And

1:14:56

FSR4 was remarkably,

1:14:58

some might say, seemingly

1:15:01

done in around

1:15:03

nine months, or

1:15:05

certainly in a

1:15:07

much shorter time

1:15:09

period than PSSR.

1:15:11

So the narrative

1:15:14

seems to be from

1:15:16

the Mark Surney side

1:15:19

is that FSSR

1:15:21

research over those years,

1:15:23

probably had an accelerating impact

1:15:25

on FSR 4 development because

1:15:28

it does seem quite unlikely

1:15:30

that AMD would just produce

1:15:32

this remarkably good upscaler so

1:15:34

quickly and out of nowhere.

1:15:36

Well, I mean, yes, I mean,

1:15:38

I imagine there is a back

1:15:40

and forth, but obviously AMD's been

1:15:42

hiring a lot of people in

1:15:44

the interim time to beef up

1:15:47

their machine learning capacity, just like

1:15:49

they hired a lot of people

1:15:51

to beef up the rate tracing

1:15:53

capacity of their business, like in

1:15:55

between after our D&A3 came out,

1:15:57

it's like right around then. And

1:15:59

I. You can just follow the

1:16:01

LinkedIn and get a sense of

1:16:03

that. And you can also see

1:16:05

that on the ML front as

1:16:07

well too. So I think it's

1:16:10

like that kind of combination of

1:16:12

things. But I think the timeline

1:16:14

is the most important part. you

1:16:16

know like you said they had

1:16:18

a council that needed to come

1:16:20

out a certain time to be

1:16:22

a bridging gap in terms of

1:16:24

time between PS5 and PS6 just

1:16:26

like the PS4 pro was and

1:16:28

they needed to have something ready

1:16:31

even if AMD didn't have anything

1:16:33

ready for them so they I

1:16:35

think they rolled a lot of

1:16:37

PSSR on their own initially and

1:16:39

then maybe that some of that

1:16:41

fed back into FSR forward development.

1:16:43

The only one thing I in

1:16:45

regards to the second question that

1:16:47

I will just like contest in

1:16:49

general is we already know from

1:16:52

at least a invidious side of

1:16:54

things that they are for that

1:16:56

transformer model they are no longer

1:16:58

using integer math for that it

1:17:00

is floating point I think it's

1:17:02

relied on their new FP8 performance

1:17:04

primarily unlocked with A to Lovelace

1:17:06

and Blackwell GPUs, so they scale

1:17:08

really well with the transformer model,

1:17:11

whereas when you go down to

1:17:13

the lower models, especially for the

1:17:15

biggest ones, like Ray Reconstruction, which

1:17:17

is huge, it performs a lot

1:17:19

worse on those GPUs. This is

1:17:21

the one thing that we don't

1:17:23

really know about with Peace5 Pro

1:17:25

and Art DNA4, is like... One,

1:17:27

Sparsity may not have anything to

1:17:29

do with it, but in just

1:17:32

general, the question for me is,

1:17:34

like, does PS5, bro, even sports,

1:17:36

Sparsity, who knows, older, invidious, you

1:17:38

be used, didn't, right? Another thing

1:17:40

is, like, is FSR 4, reliant

1:17:42

on this integer 8 masks to

1:17:44

do the work that it needs

1:17:46

to do? And we don't know

1:17:48

that either. So it's a lot

1:17:50

of guessing. I

1:17:53

don't really know, but the the

1:17:55

the porting period to me seems

1:17:57

really long for something. you know,

1:17:59

that arguably if it was in

1:18:01

between our DNA, if it's performance,

1:18:03

that was the most important part

1:18:06

is in between the levels of

1:18:08

our DNA there, the 1970 and

1:18:10

the 1970XT. Well, then it seems

1:18:12

like it would be a simple

1:18:14

porting process, but it's longer. And

1:18:17

I think it may have to

1:18:19

do with the specific math that

1:18:21

they use for it. than actually

1:18:23

the limiting aspect of performance, something

1:18:25

that Oliver also talked to with

1:18:27

a completely different cash structure and

1:18:30

everything, and maybe it needs a

1:18:32

lot more finesting to get in

1:18:34

that than an R-D-N-A-4 G-G-U, even

1:18:36

one that has less theoretical ML

1:18:38

performance, because maybe actually that is

1:18:41

the limiting aspect of performance, something

1:18:43

that Oliver also talked to with

1:18:45

Mark Surney and his interview with

1:18:47

then. So, I don't know, I

1:18:49

just think the porting period after

1:18:52

the fact that they say in

1:18:54

the beginning of 2025, which is

1:18:56

where we now, you'll see this

1:18:58

in titles first in 2026. It's

1:19:00

a pretty long time, actually. So

1:19:02

it's not, it shouldn't then be

1:19:05

an easy process. And the question

1:19:07

is, why is it not an

1:19:09

easy process? Yeah. from my understanding

1:19:11

like PSSR was really built around

1:19:13

the capabilities of the PS5 pro

1:19:16

to be as like fused as

1:19:18

fully fused as possible to stay

1:19:20

on the on-die memory there so

1:19:22

that was like a significant issue

1:19:24

for them and I think when

1:19:26

you look at FSR 4 it's

1:19:29

not it's not even clear if

1:19:31

FSR 4 is fully a CNN

1:19:33

model there's some rumblings that it

1:19:35

might have a transformer component and

1:19:37

getting that all accelerated in a

1:19:40

performance manner on a PS5 pro

1:19:42

which has this very bizarre hybrid

1:19:44

architecture that's based in our DNA

1:19:46

too, but has the RT performance

1:19:48

from our DNA 4, has moments

1:19:51

from our DNA 3 as well.

1:19:53

That's obviously a very complicated situation

1:19:55

there. There are a lot of

1:19:57

things we don't know ultimately, but

1:19:59

I do think that the time

1:20:01

period that is being suggested here

1:20:04

in terms of time to implement.

1:20:06

FSR4 is neural. network for PSSR

1:20:08

certainly does suggest that they have

1:20:10

a lot of work left to

1:20:12

do and indeed they're kind of

1:20:15

tentative about it like it's their

1:20:17

aim necessarily but not necessarily something

1:20:19

they're promising to end users. Yeah

1:20:21

I mean you know just looking

1:20:23

at the original announcement from AMD

1:20:25

it was excited for the co-development

1:20:28

with Sony interactive entertainment on the

1:20:30

models used for the FSR upscaler

1:20:32

so it does sound as a

1:20:34

widely you know, a wide collaboration

1:20:36

as part of Amethyst, you know,

1:20:39

it seems pretty straightforward, right? Yeah,

1:20:41

in terms of this scalability, I

1:20:43

mean, the thing that struck me,

1:20:45

Alex, was that it is quite

1:20:47

a heavy computational load, FSR 4.

1:20:50

I mean, obviously there is an

1:20:52

acceleration element to it because you're

1:20:54

operating from a lower resolution, right?

1:20:56

But in terms of the question

1:20:58

being asked here about, you know,

1:21:00

the 1950 and 1960, well, the

1:21:03

simple solution to that is that

1:21:05

you're likely to be upscaling to

1:21:07

like 1080P or 1440P on those

1:21:09

guards, right, which means that the

1:21:11

computational cost drops way down on

1:21:14

that. So it shouldn't be a

1:21:16

problem. Where things become more interesting

1:21:18

from my perspective is how this

1:21:20

would work on a prospective handheld

1:21:22

implementation or mobile IGBU implementation, let's

1:21:24

say, for an RDA4-based GPU, which,

1:21:27

you know, I think that's one

1:21:29

of the key objectives that AMD

1:21:31

has with FSR4. Maybe it will

1:21:33

be a different model, you know,

1:21:35

a lighter model. We've seen it

1:21:38

work before. Who knows, but that's

1:21:40

kind of the situation there. But

1:21:42

the top situation there, I mean,

1:21:44

Marx only did seem to indicate

1:21:46

that it's broadly comparable with these

1:21:49

new cards, but that's quite interesting

1:21:51

because these cards are considerably more

1:21:53

capable than the PS5 in all

1:21:55

other aspects. So how that... carries

1:21:57

over to PlayStation. Kind of need

1:21:59

a bit more detail there, I

1:22:02

think. Okay, let's move on to

1:22:04

the next question. This one comes

1:22:06

from King Penumbra. Could it be

1:22:08

that Envideo put out the 5,000

1:22:10

series knowing the performance uplift wasn't

1:22:13

amazing, but banking on the next

1:22:15

generation to be a larger node

1:22:17

shrink, sort of how 20 series

1:22:19

to 30 series... Okay, uplift was

1:22:21

followed by the node shrink on

1:22:23

40 series. Well, they would have

1:22:26

known that the performance uplift wasn't

1:22:28

going to be amazing because they're

1:22:30

on the same node. That fundamentally

1:22:32

limits what you can do because

1:22:34

you can't, you know, squeeze more

1:22:37

transistors on, you can't increase the

1:22:39

amount of logic by any meaningful

1:22:41

degree. So they would have known

1:22:43

that, you know, they were going

1:22:45

to be quite limited in that

1:22:48

regard. quite interesting here how the

1:22:50

20 series to 30 series okay

1:22:52

uplift. Sorry, the uplift is described

1:22:54

as being okay when it was

1:22:56

actually pretty awesome. And it was

1:22:58

done on what was effectively a

1:23:01

10 nanometer Samsung process that was

1:23:03

kind of refined and rebranded to

1:23:05

8th nanometer. But yeah, I think

1:23:07

we can safely assume that there

1:23:09

will be a no change for

1:23:12

60 series and there will be

1:23:14

a decent performance uplift there. 50-series

1:23:16

is just being kind of a

1:23:18

bit weird and an anemic from

1:23:20

a hardware perspective, right? Alex. Yeah,

1:23:22

it's it's I don't want to

1:23:25

say it's the most disappointing thing

1:23:27

that invidious put out in a

1:23:29

while. But in terms of on

1:23:31

the software side, I think it's

1:23:33

been one of the more interesting

1:23:36

ones they've had in the long

1:23:38

time. It's just the hardware, the

1:23:40

prices, the prices, and all the

1:23:42

other like. I haven't had driver

1:23:44

issues. I was just complaining to

1:23:47

Rich before we came here. I

1:23:49

actually was experiencing also the black

1:23:51

screen issue on the RTX 5080,

1:23:53

but I didn't know what it

1:23:55

was. I thought I did something

1:23:57

wrong. But it turns out that's

1:24:00

widespread issue. You know, like all

1:24:02

these things that just compound into

1:24:04

it. Regarding the question though, I

1:24:06

don't think it's like some sort

1:24:09

of slight of hand to like

1:24:11

make the next series look better.

1:24:13

I just think it was the

1:24:15

technological financial reality of the time

1:24:18

that the process is not ready

1:24:20

for them at this point within

1:24:22

the schedule of all the other

1:24:25

clients who want to use TSMC's

1:24:27

work. So this is what we

1:24:29

get. The question is, did it deserve

1:24:31

like a full RTX 1000, you

1:24:33

know, upgrade and naming? You know,

1:24:35

should they have pushed, if they

1:24:38

knew that the hardware was going

1:24:40

to be most similar, should they

1:24:42

have pushed also in other directions

1:24:45

like memory, like memory, like, they

1:24:47

are pushing in memory speed, but

1:24:49

they're not pushing in memory capacity.

1:24:51

So I feel like they could

1:24:54

have made this a more interesting

1:24:56

lineup of GPUs just by pushing

1:24:58

the hardware in also appreciable

1:25:00

directions that didn't require a

1:25:03

new process and they didn't

1:25:05

really do that. Mm-hmm. Well,

1:25:07

if I... Yeah, I don't really think they

1:25:09

banked on anything. I think that's...

1:25:11

Not quite correct. I didn't you

1:25:14

just have a design that doesn't

1:25:16

yield huge improvements for gaming. Obviously,

1:25:18

Blackwell is meant for different audiences

1:25:20

here and the results in terms

1:25:22

of efficiency for gaming aren't amazing

1:25:24

or seeing similar to eye sizes,

1:25:26

similar performance across the board. In

1:25:28

some cases, actually less efficient designs

1:25:30

in terms of wattage per frame

1:25:32

rate, wattage per FPS there. I

1:25:34

would expect the 60 series to

1:25:36

have a larger uplift because of

1:25:38

that 3-navator shrink-trink presumably. Hopefully 3

1:25:40

nanometers is a mature process by that

1:25:42

point. I'd imagine they'd have like some

1:25:45

3 n process, much the same as

1:25:47

they have 4 n now in terms

1:25:49

of a customized and video process. But

1:25:51

I also know that some of the

1:25:54

early 3 nanometer products weren't that great,

1:25:56

like my iPhone 15 pro here had

1:25:58

quite poor performance relative to that. expectations

1:26:00

was not actually that much faster

1:26:02

than the previous generation product. So

1:26:05

yeah, I'd hope that that would

1:26:07

be a situation where you're looking

1:26:09

at a more typical generational uplift,

1:26:11

but I wouldn't expect it to

1:26:13

be as large as something like

1:26:15

going from the 30 series cards

1:26:17

to the 40 series cards, where

1:26:19

effectively we're getting two no jumps

1:26:21

all at once, going straight to

1:26:23

the five nano euro and four

1:26:25

nanometer class hardware there. That was

1:26:27

such a huge upgrade. I don't

1:26:29

think we're going to see anything

1:26:31

like that really here. Yeah, I

1:26:33

mean I'm just going to say

1:26:35

what I've been saying in the

1:26:37

past, which is to say that

1:26:39

you're going to be seeing a

1:26:41

lot more software driving quote-unquote performance

1:26:43

improvements. It's just the way it

1:26:45

is as the cost for transistor

1:26:47

is not, you know, scaling to

1:26:49

anything like this sort of thing

1:26:51

we saw previously. These big generational

1:26:53

uplifts are getting more and more

1:26:56

expensive. So of course we're going

1:26:58

to be seeing this balance between

1:27:00

software and hardware and hardware. change.

1:27:02

I think I said that in

1:27:04

the 50-90 review. The problem is

1:27:06

that if you're going to be

1:27:08

doing this via software you need

1:27:10

to have every game support those

1:27:12

software innovations. We've kind of got

1:27:14

there I think to a certain

1:27:16

extent with DLS super resolution. You

1:27:18

know if a game ships now

1:27:20

without DLS that's a problem, right?

1:27:22

We haven't got there with frame

1:27:24

generation. And we certainly haven't got

1:27:26

there with Ray reconstruction, which I

1:27:28

think is crucially important as we

1:27:30

move into the RT era more

1:27:32

dramatically. So yeah, there's a lot

1:27:34

of work to be done in

1:27:36

terms of implementation of software, but

1:27:38

I do think that that's basically

1:27:40

the direction of travel, and that's

1:27:42

certainly what you're going to be

1:27:44

seeing on consoles, which is why

1:27:47

we've got stuff like Project Amethyst

1:27:49

happening in the first place. and

1:27:51

you know Microsoft have their own

1:27:53

AI endeavors happening behind the scenes

1:27:55

which we on the gaming side

1:27:57

all we've seen. so far as

1:27:59

auto SR really, I guess auto

1:28:01

HDI to a different degree. But

1:28:03

you know, that's basically the direction

1:28:05

of travel because the concept of

1:28:07

just scaling up GPUs is kind

1:28:09

of like, you can do it,

1:28:11

but you're going to have to

1:28:13

pay through the nose to enjoy

1:28:15

that to get that extra traditional

1:28:17

performance. Yeah, but. I would expect

1:28:19

to see a new node for

1:28:21

the next generation. And you know,

1:28:23

there's a lot happening with actual

1:28:25

competition seemingly, you know, Intel seemingly

1:28:27

back in the game with their

1:28:29

next process. And, you know, just

1:28:31

the concepts that you don't necessarily

1:28:33

need to go to TSMC to

1:28:35

fabricate your chips. That's probably going

1:28:38

to be a good thing. Yeah.

1:28:40

But who knows? I think it

1:28:42

has been a bit of a...

1:28:44

holding pattern for this particular generation,

1:28:46

but I think absolutely we should

1:28:48

consider, you know, there could be

1:28:50

a possibility that GPU, gen-on-gen uplifts,

1:28:52

would be more like CPU, gen-on-gen

1:28:54

uplifts. I think that's a very

1:28:56

real possibility, and it's ghetto upset

1:28:58

a lot of people, but ultimately,

1:29:00

you kind of need competition to

1:29:02

drive innovation, but if both AMD

1:29:04

and Envideo are kind of on

1:29:06

the same track, that kind of

1:29:08

says that there's an external factor

1:29:10

affecting affecting it affecting it. And

1:29:12

it probably is the fabrication side

1:29:14

of things. Let's move on to

1:29:16

the next question. This one from

1:29:18

Octalima. Something that's been floating around

1:29:20

my mind lately is given the

1:29:22

vast range of graphics games offer,

1:29:24

it's entirely possible that platforms like

1:29:26

the PS4 simply never go away.

1:29:29

Nothing would stop a lightweight indie

1:29:31

game, especially a 2D1, from running

1:29:33

on it in 10 years or

1:29:35

so. If the steam deck is

1:29:37

the bar for current gen, in

1:29:39

theory last gen ports should always

1:29:41

be possible, at least for this

1:29:43

generation, no. Well, actually, actually, I'm

1:29:45

going to go to where you

1:29:47

are on this one, because still

1:29:49

in development is your too big

1:29:51

for steam. 2025 update. I think

1:29:53

he's kind of right in the

1:29:55

way in the sense that you

1:29:57

know there are always going to

1:29:59

be lightweight games that can run

1:30:01

on older GPUs. There's nothing new

1:30:03

in that right but there are

1:30:05

definitely new games that don't run

1:30:07

particularly well on the steam deck

1:30:09

and in terms of this PlayStation

1:30:11

4 comment I think it's simply

1:30:13

the case that people just have

1:30:15

moved on from the PlayStation 4

1:30:17

to the PS5 or whatever so

1:30:20

there's no business case for making

1:30:22

those games. I think people have

1:30:24

moved on and I think ultimately

1:30:26

like the age of the hardware

1:30:28

is just a concern whether that's

1:30:30

a failure concern or whether that's

1:30:32

just people throwing out that hardware.

1:30:34

Obviously if you have a PS5

1:30:36

there's very little reason to have

1:30:38

a PS4 right and as far

1:30:40

as I understand it most PS4

1:30:42

players aren't buying like the hottest

1:30:44

new indie titles, they're playing Fortnite,

1:30:46

they're watching Netflix, they're playing war

1:30:48

zone, they're doing these kinds of

1:30:50

tasks, they tend to not be

1:30:52

as invested in gaming. as perhaps

1:30:54

people who have newer console hardware

1:30:56

and newer PC hardware. And also

1:30:58

at some point, who knows when

1:31:00

this will happen, but Sony may

1:31:02

just flat out stop you from

1:31:04

shipping games on PS4 on that

1:31:06

storefront. Right now on the PS3

1:31:08

Invita storefronts, you can access games,

1:31:11

you can download them, you can

1:31:13

buy them, but no developer publishers

1:31:15

been able to add new games

1:31:17

to those storefronts in years. I

1:31:19

think like five or six years