Support NeoGAF

[Playboi Carti]

I still think its pointless to release a PS5 pro but I now realize if money can be made Sony will do it even if it doesn't make sense.

Do you have any arguments to back up the "it doesn't make sense" narrative?. No point in calling a future Pro console pointless when you have plenty of evidence of why it would be objectively speaking, far from pointless. Starting by how PSVR2 games would greatly benefit of having a console with a beefier GPU to mitigate the use of reprojection and achieve higher native framerates/higher res, and that's just one reason why, I could go on and on. Calling it pointless cuz it's not a product that's justifiable for you doesn't make the product pointless pal.

 

[Mr.Phoenix]

I think that site is doing the 7700XT as having 4 Shader Engines, with 16CUs (8WGP) per Shader Engine.
64CUs, with 10CUs (5WGP) disabled.

I would assume this configuration is incorrect.

This is the correct specs for the full 60CUs chip.
AMD's RDNA 3 Graphics

Navi32​

• gfx1101 (Wheat Nas)
• Chiplet - 1x GCD + 4x MCD (0-hi)
• 30 WGP (60 legacy CUs, 7680 ALUs)
• 3 Shader Engines / 6 Shader Arrays
• Infinity Cache 64MB (0-hi)
• 256-bit GDDR6
• GCD on TSMC N5, ~200 mm²
• MCD on TSMC N6, ~37.5 mm²

Which lines up with the actual chip.

Which should be haft the Shader Engines of Navi 31.

Spot on. The GCD also lines up with what would be an expected size for the PS5pro without the CPU clusters. I wonder how much mm2 an 8-core zen 4? cluster would be and how much cache that allows them to add to the system. Because if I recall, the infinity cache in RDNA3 is in the MCDs. And the PS5pro would be a monolithic die.

I mean it's all fun speculation for now but everything I've been hearing for the past year indicates that a 2x increase in GPU perf is the target. If it's only a 7700XT/3080 level, then that is nowhere near a 2x increase from the current PS5 GPU (~RX6700/6650XT). For that reason, I'll think we will see something at least closer to a RX6900XT although I can see how the 7700XT as a model makes sense.

You are kinda contradicting yourself.

Also, you're getting too caught up in the TFLOPs which we should have been learned our lesson on that one by now. The 2x increase target is likely not solely TFLOP theoretical power, but actual game performance (e.g. can the PRO take the 4K/30fps output in Uncharted Legacy collection and now run it at 4K/60fps comfortably?). RDNA3 does have some improvements in efficiency and optimization over RDNA 2 that do not show in the theoretical TFLOP calculation. Also, Rich from DF made a good point in their weekly podcast this week that while we haven't seen the dual issue instructions in the ALU for RDNA3 (or NV's Ampere/Ada for that matter) amount for much gain on the PC side, a dedicated console could potentially take better advantage of that which would boost the computational power quite a bit. Not to mention the other likely console specific optimizations Sony will have plus any pieces of RDNA 4 they opt to put in and yeah it's entirely plausible that even a 54 CU part with a high enough clock (i.e. 2.5Ghz or more) could double the actual game perf of the base PS5.

And this part is why.

Yes, 54CU@ ~2400Mhz, is ony 16.5TF. So yes, nowhere near a 2xTF compared to the PS5. But if this is RDNA3+ we are talking about, then that is technically a 33TF GPU because of the whole dual issue thingy and at least that's how AMD counts RDNA3 TFs. Then there is also the fact that it wouldn't just have more RT cores compared to the og PS5, but AMD gen 3 RTcores, which on their own should yield over 2x the performance of gen 1 RT cores found in the OG PS5. Mostly in part due to BVH acceleration which the gen one RT doesn't have. And chances are there would be something else Sony could add in their chip, maybe something that allows them do AI upsampling. In the same way they built in CBR hardware into the PS4pro. And they can ee just double the available cache on the APU which would do wonders for the CPU besides just having a cock boost too.

The point is, outside just CU and clock math, there are lots of other areas the PS5pro that could be improved over the og PS5.

 

[anothertech]

But I am guessing....

18-24 TF's
20gb Ram
4ghz 8 core/16 threads CPU
1-1.5 TB storage
7750ish XT GPU

$549

Wait till gpu shit hits next year.

Ppl gon be like, '50 tf is low end' yadda yadda lolol

 

[Physiognomonics]

I mean it's all fun speculation for now but everything I've been hearing for the past year indicates that a 2x increase in GPU perf is the target. If it's only a 7700XT/3080 level, then that is nowhere near a 2x increase from the current PS5 GPU (~RX6700/6650XT). For that reason, I'll think we will see something at least closer to a RX6900XT although I can see how the 7700XT as a model makes sense.

Also, you're getting too caught up in the TFLOPs which we should have been learned our lesson on that one by now. The 2x increase target is likely not solely TFLOP theoretical power, but actual game performance (e.g. can the PRO take the 4K/30fps output in Uncharted Legacy collection and now run it at 4K/60fps comfortably?). RDNA3 does have some improvements in efficiency and optimization over RDNA 2 that do not show in the theoretical TFLOP calculation. Also, Rich from DF made a good point in their weekly podcast this week that while we haven't seen the dual issue instructions in the ALU for RDNA3 (or NV's Ampere/Ada for that matter) amount for much gain on the PC side, a dedicated console could potentially take better advantage of that which would boost the computational power quite a bit. Not to mention the other likely console specific optimizations Sony will have plus any pieces of RDNA 4 they opt to put in and yeah it's entirely plausible that even a 54 CU part with a high enough clock (i.e. 2.5Ghz or more) could double the actual game perf of the base PS5.

Could be close to 2x. It depends of how high they are going to clock it.

 

[TheRedRiders]

Lol the more I look back at the sentiments surrounding PS4 Pro launch at the time, the more I realize we are literally just repeating history.

PS4 Pro Represents Nearly One in Every Five PS4s Sold - IGN

Sony execs talk about the success of the PS4 Pro and what has led to the system's strong sales.

www.ign.com

I like this Jim Ryan quote in particular.

I made the same argument months ago in a different thread, even plucking the same Jim Ryan quote.

This is just false, you’re over-generalising the “average consumer”. Your fast-food analogy works both ways as well, the mobile phone industry has pretty much the same “fast-food” model, pumping out new products with new gimmicks every year and people eat it up like crazy. But let’s get back to the consoles, it’s true average consumers want a pug and play experience with games, however the idea of Pro console doesn’t contradict this. In fact the PS4 Pro did an excellent job in streamlining the graphical settings and options for most gamers. Performance and Fidelity modes, PS5 Pro will likely do something similar.

Let’s also look at the hard data, PS4 Pro sold very well, in fact even Sony were surprised judging by their comments.

"We were confident about the product but have really been taken aback by how well it's done," Ryan said. "Almost one in five PlayStations sold since that launch in November has been a Pro. That's significantly ahead of our expectations. We're feeling pretty good about that."

In fact here’s Sony pretty much debunking your argument.

"Sometimes I think we can be guilty of ascribing too much rationality to gamers," Ryan explained. "People just want the best. Maybe they just want to future proof? I think we see the same thing from Apple customers too — there are people that want the best that you can buy."

 

[ergem]

Lol the more I look back at the sentiments surrounding PS4 Pro launch at the time, the more I realize we are literally just repeating history.

PS4 Pro Represents Nearly One in Every Five PS4s Sold - IGN

Sony execs talk about the success of the PS4 Pro and what has led to the system's strong sales.

www.ign.com

I like this Jim Ryan quote in particular.

This is why I keep saying PS4 Pro sold about 10% or at least less than 20% compared to the base PS4. There was already a sizable amount of PS4 sold when PS4 Pro was released and even then PS4 Pro was only selling 1:5.

Those saying that the Pro version will affect them because their base PS5 will somehow be an afterthought couldn't be more wrong. The base version will be the choice of the overwhelming majority of gamers and it will continue to be the target specs.

 

[MatrixReloaded101]

Never mind, no point in repeating what's already been said.

 

[Mobilemofo]

I wonder what the average age of the playstation gamer is now? Would be interesting to find how the market has changed.

 

[mansoor1980]

regarding cooling and heat , there was a ton of false info spread online before ps5 launch , like this one

 

[TheRedRiders]

regarding cooling and heat , there was a ton of false info spread online before ps5 launch , like this one

Jesus Christ

 

[welshrat]

I wonder what the average age of the playstation gamer is now? Would be interesting to find how the market has changed.

Would be interesting. I know myself and mates are 46 - 47 and my wife joined the party in the last few years at 45 so although we are not the majority I do know that a lot will be in between 35 and 50 and will have a fair amount of disposable income to throw at a pro. I know myself and good friend will pre-order day 1 and I imagine there are a lot like us.

 

[GHG]

regarding cooling and heat , there was a ton of false info spread online before ps5 launch , like this one

That last paragraph.

Imagine sending out all the FUD merchants with marching orders only for it to have little to no effect.

 

[genesisblock]

regarding cooling and heat , there was a ton of false info spread online before ps5 launch , like this one

Oh my god, I remember this. Love good ol' FUD.

 

[HeisenbergFX4]

regarding cooling and heat , there was a ton of false info spread online before ps5 launch , like this one

I bet this started because of the reported devkits overheating and shutting down and people just ran with it

 

[MasterCornholio]

I bet this started because of the reported devkits overheating and shutting down and people just ran with it

From reading that comment it appears the individual was just making things up. Taking one rumor and turning it into fake insider knowledge to scare future PS5 consumers.

 

[mansoor1980]

I bet this started because of the reported devkits overheating and shutting down and people just ran with it

you mean rumored right?
or were there legit reports?

 

[DeepEnigma]

From reading that comment it appears the individual was just making things up. Taking one rumor and turning it into fake insider knowledge to scare future PS5 consumers.

Classic FUD.

 

[Rocco Schiavone]

I bet this started because of the reported devkits overheating and shutting down and people just ran with it

Most likely the devkits runned fine just like the retail hardware(another bullshit rumor). I remember this bullshit back then. Hillarious.

 

[welshrat]

I imagine the FUD will be 100x worse if Sony are the only ones with a mid gen refresh

 

[TheRedRiders]

you mean rumored right?
or were there legit reports?

I think there was some very early reports that a small number of PS5 dev kits were overheating, but it's likely it was resolved very quickly as no more reports surfaced and the PS5 is running whisper quiet and cool. The proofs in the pudding.

It is evident that certain individuals were greatly exaggerating those early reports, injecting their own nonsense and hearsay like the twat in that post. All of it just aged so pathetically.

 

[Horseganador]

The FUD against PS5 was wild on all fronts.

 

[damidu]

regarding cooling and heat , there was a ton of false info spread online before ps5 launch , like this one

disgusting, pretty much lying through the teeth. though expected of these rats, happens at every gen start.
probably organized by ms execs.

they know their box can’t compete with just being fine, they have to drag sony down somehow too.

 

[Tqaulity]

Spot on. The GCD also lines up with what would be an expected size for the PS5pro without the CPU clusters. I wonder how much mm2 an 8-core zen 4? cluster would be and how much cache that allows them to add to the system. Because if I recall, the infinity cache in RDNA3 is in the MCDs. And the PS5pro would be a monolithic die.


You are kinda contradicting yourself.

And this part is why.

Yes, 54CU@ ~2400Mhz, is ony 16.5TF. So yes, nowhere near a 2xTF compared to the PS5. But if this is RDNA3+ we are talking about, then that is technically a 33TF GPU because of the whole dual issue thingy and at least that's how AMD counts RDNA3 TFs. Then there is also the fact that it wouldn't just have more RT cores compared to the og PS5, but AMD gen 3 RTcores, which on their own should yield over 2x the performance of gen 1 RT cores found in the OG PS5. Mostly in part due to BVH acceleration which the gen one RT doesn't have. And chances are there would be something else Sony could add in their chip, maybe something that allows them do AI upsampling. In the same way they built in CBR hardware into the PS4pro. And they can ee just double the available cache on the APU which would do wonders for the CPU besides just having a cock boost too.

The point is, outside just CU and clock math, there are lots of other areas the PS5pro that could be improved over the og PS5.

Could be close to 2x. It depends of how high they are going to clock it.

I'd just point out that when talking about an RX7700XT vs an RX7800, we're likely pulling straws since both cards are expected to be pretty close to each in performance (roughly around 10% of each other). Again, I'm not even going mention TFLOPs as it really is meaningless in this context but we agree that there is a lot Sony can do to boost perf beyond just the PC equivalent parts.

After looking at this some more, I think the 6900XT as the broad target makes the most sense in terms of what is realistically feasible to get the most meaningful upgrade in a console form factor by end of 2024. Estimated perf for the 7700XT is within 5% of the 6900XT and the 7800 is estimated to be within 5% above it so they are all in the ballpark. Initially, I wasn't sure a 6900XT would be a big enough jump in perf to justify the Pro but real world game performance shows the 6900XT in 2023 being an even match for the RTX 3090 and slightly faster than an RTX 4070 (Source). That's pretty good actually and while not 2x the PS5 in raw perf, it's relatively close and like we already said, the additional customizations along with RDNA3 architecture and console efficiencies could push it closer to that 2x perf number. I always speculated that the "3090" was the ceiling of what we could realistically expect in a console for a Pro and these rumored specs actually show that being feasible (depending on active CU count and clock speeds of course).

Let's not act too jaded here...that much power in a dedicated gaming machine would be AWESOME! Remember, console games almost never run at ultra settings for good reason...it's extremely wasteful with minimal ROI. So while 7900 XTX and 4090 are pushing games to 200+ fps, that's not necessary for a living room console where 4K/120 is the max (and millions of gamers are still probably 1080/60 or 4K/60). Something like a 6900XT is running most games of the last 5 years or so at roughly 100fps average at 4K ultra settings (Source). Turning that down to High settings (more likely for console) show perf increase to >160fps average at 4K for the same set of games (Source). Of course, there are heavy outliers like TLOU Pt1 or Callisto Protocol running in the 50s and who knows what the full crop of "next gen" games and UE5 games would run like (we already see some games pushing 4090 to not even hit a steady 60fps without reconstruction). But a 6900 XT level of perf should be broadly capable of doing 4K/60 in most titles (raster) while delivering the 2.5x ray tracing perf due to RDNA 3+ which would be exciting indeed!

 

[mckmas8808]

Side by side specs

I'd buy this day one. Sony could probably make this console for an honest $600 at retail.

 

[SlimySnake]

My favorite part about the overheating rumors was that the PS5 actually ended up using RDNA 2.0 clocks that were rather conservative seeing as how some GPUs went above 2.5 Ghz. 2.2 Ghz is actually well below what the 6600xt uses.

It was the XSX GPU that used lower clocks that probably holds it back in some of the games where we've seen the PS5 perform better despite having fewer tflops. so even if it did overheat some devkits, it wasn't exactly something truly insane being done by Cerny to close the gap with the XSX.

If anything, MS shouldve probably stuck with 40 CUs and built a 6700xt equivalent with high clocks instead of this 52 CU slow and wide monstrosity. The 6700xt never loses any benchmarks to the 6600xt, no matter what game. That 11-11.5 GPU might not look good on paper but it wouldve definitely performed better and more consistently in every single game.

 

[asdasdasdbb]

I'd buy this day one. Sony could probably make this console for an honest $600 at retail.

I don't even think AMD supports GDDR6X.

It's probably something like this:

8 Zen 4c cores, max 4 Ghz
60 RDNA3/4 CUs, max 2.5 Ghz
16 GB GDDR6 @ 18 Gbps
1.65 TB drive
No Infinity Cache or Disc Drive
~$599

 

[Skifi28]

1.65 TB drive

I keep reading this, but it seems a tad unrealistic. If there's one relatively unimportant component where they could cut costs to hit their target, this is it. I also image that those upgrading to a pro will most likely be the "hardcore" users which are even more likely to have expanded their storage with a third party SSD they will just carry over. Just seems like a big waste of money that could have gone to another component or be used to reduce the retail price.

 

[Codeblew]

I keep reading this, but it seems a tad unrealistic. If there's one relatively unimportant component where they could cut costs to hit their target, this is it. I also image that those upgrading to a pro will most likely be the "hardcore" users which are even more likely to have expanded their storage with a third party SSD they will just carry over. Just seems like a big waste of money that could have gone to another component or be used to reduce the retail price.

It may be worth it just for the marketing bullet point if it doesn't cost a whole lot extra. It may be a lot cheaper that we think with die shrinks and buying the chips in mass quantity.

I am more surprised he thinks it will be 60 CU's instead of 54. Which implies none are disabled for better yields.

 

[Loxus]

It may be worth it just for the marketing bullet point if it doesn't cost a whole lot extra. It may be a lot cheaper that we think with die shrinks and buying the chips in mass quantity.

I am more surprised he thinks it will be 60 CU's instead of 54. Which implies none are disabled for better yields.

Yea, many people that want 72CUs, don't know it's 80CUs with 8CUs disabled.

So if it's 60CUs, following both 36CUs and 72CUs. It's 54CUs after disabling 6CUs.

 

[Sethbacca]

I keep reading this, but it seems a tad unrealistic. If there's one relatively unimportant component where they could cut costs to hit their target, this is it. I also image that those upgrading to a pro will most likely be the "hardcore" users which are even more likely to have expanded their storage with a third party SSD they will just carry over. Just seems like a big waste of money that could have gone to another component or be used to reduce the retail price.

Especially if we're talking late next year. I just bought a 2tb nvme for $80 to upgrade from my 1tb that I bought at ps5 release for around $225. We may see 4tb drives drop down into the $150 range by late next year at the rate costs are dropping. Personally I'd rather have the cost of the PS5 reflect better cpu/gpu etc hardware over the storage.

 

[Ronin_7]

My guess is 2500-2700Mhz GPU.

No way they clock it lower than this.

Edit: Cerny loves higher clocks, he'll probably try 3000Mhz.

 

[Mr.Phoenix]

I don't even think AMD supports GDDR6X.

It's probably something like this:

8 Zen 4c cores, max 4 Ghz
60 RDNA3/4 CUs, max 2.5 Ghz
16 GB GDDR6 @ 18 Gbps
1.65 TB drive
No Infinity Cache or Disc Drive
~$599

I don't know why we are still calling it 60CUs. We might as well start calling it 54CU. Because if Sony/Cerny is following what they have been doing thus far, 18/20CU (active/total) basically represents 1 GPU unit so to speak. PS4 has 1 unit. Ps4pro has 2 units. PS5 has 2 units. PS5pro will either have 3 units 54/60CU (most likely) of 4 units 72/80CU.

I really do not see them doubling the Storage size in the PS5 either. Just absolutely no reason to do so. They gave us of the shelf SSD support. And the kinda people buying a Pro console, are more than likely going to buy a bigger SSD anyways.

They don't have to put in an infinity cache. And that shit is named as such because of how big it is. All they have to do is just put in more cache for the CPU or even a pool of it that can be shared between the CPU and the GPU. Increasing it to 16MB would be over twice what the current PS5 has. And 32MB would be even better. Doesnt have to be in the 96-128MB we are seeing on the PC side of things.

And that price... $599. You guys keep throwing around prices like its nothing lol. It's Nvidia's fault. Anyways, $499 is plenty enough. I can see it being $499 without the disc drive while the PS5slim would be $399 without the disc drive. Can also see the disc drive add-on being $100. Think of the PS5pro like this, its the PS5, but the only thing that cost more to make is the APU, and maybe the RAM cost a little more being that you are using faster chips (18Gbs instead of 14Gbs). Everything else costs the same or less. The disc drive they are axing makes up for the increased APU cost and RAM cost..

It may be worth it just for the marketing bullet point if it doesn't cost a whole lot extra. It may be a lot cheaper that we think with die shrinks and buying the chips in mass quantity.

I am more surprised he thinks it will be 60 CU's instead of 54. Which implies none are disabled for better yields.

No its not. And even if it doesn't cost a lot, it still kinda does. And yes it doesn't cost a lot. A good way to guess its price, if you don't wanna just go to the Nand flash exchange (yes its a real thing), is just look at the cost of an SSD on Amazon and half its price. And you kinda have a per GB cost. eg. 1TB NvME4 WD SSD on Amazon cost $60 right now. So that's $0.06/GB. At that cost, the PS5 SSD cost $49.5. But PS5 SSD doesn't have OEM and retail markup, so it probably cost at least 25%-50% less than that. So let's just say it cost $35. Doubling it would be adding another $35 to the BOM. May not sound like a lot, until you start multiplying that by 10-20M.

What he says doesn't mean he thinks it would be 60 active CU. He is just saying what he heard. But if we use some sense, knowing how Sony and AMD do stuff... we can easily fill in the gaps. There s no way a console APU is made without disabling some CUs. None at all. Sonys narrow and fast approach, means they are not exceeding 10 WGs per SE. Identical to what they used in the PS4, PS4pro and PS5.And this is said to have 3 SEs. That means 30 WGs. Disabeone for each SE, and you end up with a total of 27 WGs aka... 54CU.

We even already know the exact GPU this APU is based on more or less. That would have the exact same SE layout.

 

[Codeblew]

No its not. And even if it doesn't cost a lot, it still kinda does. And yes it doesn't cost a lot. A good way to guess its price, if you don't wanna just go to the Nand flash exchange (yes its a real thing), is just look at the cost of an SSD on Amazon and half its price. And you kinda have a per GB cost. eg. 1TB NvME4 WD SSD on Amazon cost $60 right now. So that's $0.06/GB. At that cost, the PS5 SSD cost $49.5. But PS5 SSD doesn't have OEM and retail markup, so it probably cost at least 25%-50% less than that. So let's just say it cost $35. Doubling it would be adding another $35 to the BOM. May not sound like a lot, until you start multiplying that by 10-20M.

What he says doesn't mean he thinks it would be 60 active CU. He is just saying what he heard. But if we use some sense, knowing how Sony and AMD do stuff... we can easily fill in the gaps. There s no way a console APU is made without disabling some CUs. None at all. Sonys narrow and fast approach, means they are not exceeding 10 WGs per SE. Identical to what they used in the PS4, PS4pro and PS5.And this is said to have 3 SEs. That means 30 WGs. Disabeone for each SE, and you end up with a total of 27 WGs aka... 54CU.

We even already know the exact GPU this APU is based on more or less. That would have the exact same SE layout.

I would bet money that they double SSD size but really don't care personally. As far as the CU's go, why do you sound like you are lecturing me? Do you just naturally sound condescending? I know how CU's work...

 

[Ivan]

This is so dumb. Just really boring discussion that doesnt add anything to the topic. Just one bad take after another.

So true... They get way too much attention.

 

[ChiefDada]

I don't know why we are still calling it 60CUs. We might as well start calling it 54CU. Because if Sony/Cerny is following what they have been doing thus far, 18/20CU (active/total) basically represents 1 GPU unit so to speak. PS4 has 1 unit. Ps4pro has 2 units. PS5 has 2 units. PS5pro will either have 3 units 54/60CU (most likely) of 4 units 72/80CU.

I understand what you and others are saying about why having all CUs enabled isn't possible, but then I still don't understand the Kepler tweet suggesting otherwise. What else would be the point of that tweet?

 

[Mr.Phoenix]

I would bet money that they double SSD size but really don't care personally. As far as the CU's go, why do you sound like you are lecturing me? Do you just naturally sound condescending? I know how CU's work...

Wasnt my intention. Sorry if I came off that way.

However, this is a forum, and more of then than not, it's best to work from the assumption that whoever you are talking to might not know what you are talking about and give all the information necessary. Because even if you know, some people might not.

I personally feel, that to instead give a response like `that's not how CUs work` and leave it at that, would actually be condescending.

I understand what you and others are saying about why having all CUs enabled isn't possible, but then I still don't understand the Kepler tweet suggesting otherwise. What else would be the point of that tweet?

link to the tweet? Or just regurgitate what was said.

 

[Mobilemofo]

Would be interesting. I know myself and mates are 46 - 47 and my wife joined the party in the last few years at 45 so although we are not the majority I do know that a lot will be in between 35 and 50 and will have a fair amount of disposable income to throw at a pro. I know myself and good friend will pre-order day 1 and I imagine there are a lot like us.

Same here. 46 now, tho my mates are not gamers, but old school pissheads. I don't drink at all. No kids so a fair bit of disposable income and time. I'd rather enjoy a good game than movie etc.

 

[ChiefDada]

link to the tweet? Or just regurgitate what was said.

 

[Mr.Phoenix]

Ok.. well, I don't know what to make of that. He's not saying Sony is gonna use 60/60. Just saying nothing is stopping them. To which the argument would be, yields. Which in turn affects cost. Precedent and conventional wisdom suggest that WGs would be disabled to improve yields. But sure, nothing stops them from not doing that if they are willing to throw away more chips from each wafer inadvertently making each chip more expensive. All for what? Just under 2TF extra?

Ah well, guess it comes down to what their yields are like, and/or how much they plan on pricing the pro at.

 

[shamoomoo]

Interesting? Is it because 11 is an uneven number and they would have to increase the WGP to 12 dual CUs per shader engine so 24+24+24 instead of 20+20+20.

 

[Loxus]

In a way, he seems to be contradicting himself.


Sony's standard design has been to disable CUs for yields. 18/20, 36/40.
Sony has never had all CUs enabled before. So following what Sony did in the pass, common sense would tell you it's 54/60.

 

[ChiefDada]

In a way, he seems to be contradicting himself.


Sony's standard design has been to disable CUs for yields. 18/20, 36/40.
Sony has never had all CUs enabled before. So following what Sony did in the pass, common sense would tell you it's 54/60.

Yeah I'm just going to assume he got mixed up a bit. 54/60 makes sense.

 

[HeisenbergFX4]

you mean rumored right?
or were there legit reports?

The way I understand it the very first DKs ran hot and very loud but likely no overheating issues

 

[Mr.Phoenix]

Interesting? Is it because 11 is an uneven number and they would have to increase the WGP to 12 dual CUs per shader engine so 24+24+24 instead of 20+20+20.

For starters, its because AMD never designed their architecture to do that. Even AMD avoids using as much as 10WGs per SE in most cases. Eg. Their flagship GPU, 7900XTX has a total of 96CU. In a 6 SE config with each SE having 8WGs each. And the 7900XT? Its the 7900XTX but with one WG disabled in each SE. So it has 84CU.

Technically, they could have made a GPU with 120CU!!! if they went with the PS5 SE config.Or even bigger if they went with the XSX config.

Anyways, You have the l2 cache which sits on the bridge with the geometry engine and GPU command processor, this bridge is where all the SEs are connected to. Then within the SEs, you have the l1 cache, render backend, rasterizer...etc that feed the individual WGs. Some of these things are usually the same across AMD GPUs. So imagine the PS5 having 2MB l2 cache per SE of 18CU each and the XSX having 2.5MB for 26CU. PS5 having 64ROP for 36CU and the Series X having the same 64ROP for 52CU. Or the PS5 l2 cache even being faster because its clock is tied to the GPU clock.

The point is, just like with CPU cores, at some point, it's better to add more SEs, than to make each individual SE bigger. Or not you end up with something like the XSX, which can claim 12TF... but doesn't always perform as such.

 

[Bo_Hazem]

You guys are entertaining. You give me vibes of the legendary speculation thread.

 

[ChiefDada]

You guys are entertaining. You give me vibes of the legendary speculation thread.

I feel the same all this hype is getting me amped and I'm especially looking forward to what they're cooking for RT. While it's always exciting to look forward into the future, I personally love looking back in the past every now and then for gut checks and recalibrating expectations. I came across one of VFXVeteran post from just a year and a half ago. I know we're all speculating about PS5 Pro and nothing is certain, but what do we make of his prediction? Does anyone believe we'll be lucky for the PS6 to reach 3090 level performance lol?

I'm not saying that at all. I'm saying that even with RT on the level of the CGI requires enormous bandwidth which no GPU has today. And no, the PS6 won't be close either. The consoles lag behind the PC GPUs by a significant margin, so we should be looking at GPUs from Nvidia/AMD instead of consoles. I'm looking for VRAM that's at least 64G and bandwidth that's plentiful where these games can be run easily at native 4k/60FPS with all the options turned up high and no reconstruction techniques. We won't be there with the next iteration of the consoles for sure. We'd be lucky to get a PS6 to have the capability of a 3090 today. And a 3090 is way off from CGI (with it's bandwidth) requiring DLSS to even run at a reasonable FPS.

 

[deriks]

Trinity?!

So many Oppenheimer jokes

 

[SolidQ]

'd say the PS5 Pro is basically a 7700XT.

 

[asdasdasdbb]

I keep reading this, but it seems a tad unrealistic. If there's one relatively unimportant component where they could cut costs to hit their target, this is it. I also image that those upgrading to a pro will most likely be the "hardcore" users which are even more likely to have expanded their storage with a third party SSD they will just carry over. Just seems like a big waste of money that could have gone to another component or be used to reduce the retail price.

I think it will be in the Slim too.

My thinking is that the marginal cost for the double capacity chips isn't that much, and it might be better for sourcing long term.

 

[ChiefDada]

I'd say the PS5 Pro is basically a 7700XT

On paper maybe. But in practice have we ever seen a 6700xt outperform PS5? Nothing comes to mind for me.