• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

PS5 Pro Specs Leak are Real, Releasing Holiday 2024(Insider Gaming)

Heisenberg007

Gold Journalism
I really thought this leak was fake because it sounded too good to be true and way too powerful.

Hot damn if this is indeed true. I wonder how expensive this system will be.

Can't wait to see PSSR in action! 🔥
 

saintjules

Member
I still don't think the PS5 Pro is real.

I am starting to think the Pro does not exist and this thread was all about:

myu7Ftm.png

I’m not convinced we are getting a ps5 pro this year, as there was nothing in insomniac leaks, maybe next year.

I'm starting to think the pro doesn't even exist 😔

4MtD7Yk.gif
 

Bojji

Member
Completely ignoring the large improvements to architecture for RT and ML/AI SS….and closed box nature will punch above its paper weight compared to PC specs

This is pure raw power of this GPU, of course in RT it will be much faster than 7700XT but only some games have RT. New upscaling will change image quality but nothing else, rendering game in 1440p and upscaling to 4k will still "cost" the same (maybe slighty less compared to FSR2 thanks to dedicated hardware).

Ai stuff is mystery, in PC space tensor cores are used only for upscaling so far.
 

ChiefDada

Gold Member
Completely ignoring the large improvements to architecture for RT and ML/AI SS….and closed box nature will punch above its paper weight compared to PC specs

I mean, it's like I'm in the twilight zone! Insane how many people are so focused on pure raster when that's not even the point of the console. Imagine discussing nvidia gpu and not focus on DLSS, RT cores. If that were the case, every PC gamer would favor AMD for raster performance edge. Literally the same thing happening here.
 

SlimySnake

Flashless at the Golden Globes
I swear it's like some ignore specs and how these devs maximize these consoles. This is a sizeable jump. Don't let anyone make you think it's not.
Its a sizeable jump in ray tracing and machine learning hardware. But 1.45x increase in standard rasterization is NOT a sizeable jump when you compare PS4 Pro which was around 2.23x over PS4 Pro and Xbox One X which was 4x over the Xbox One.

Just look at the biggest games last year. Baldurs Gate 3, RE4, SPiderman 2, FF16, Alan Wake 2, Star Wars and Avatar. Only Avatar, Spdierman 2 and Star Wars will gain any major performance wins.

This year, FF7 will get the meager 45% increase in pixels. Dragons Dogma 2 might have ray tracing on consoles so it might see a substantial boost. Rise of ronin? not so much. Nothing big for TLOU remake, GOW ragnorak, HFW, Demon Souls or Returnal. You better hope Sony studios are going all in on ray tracing in the future because otherwise the jump is very mediocre in comparison to last gen's mid gen consoles.
 

LiquidMetal14

hide your water-based mammals
Its a sizeable jump in ray tracing and machine learning hardware. But 1.45x increase in standard rasterization is NOT a sizeable jump when you compare PS4 Pro which was around 2.23x over PS4 Pro and Xbox One X which was 4x over the Xbox One.

Just look at the biggest games last year. Baldurs Gate 3, RE4, SPiderman 2, FF16, Alan Wake 2, Star Wars and Avatar. Only Avatar, Spdierman 2 and Star Wars will gain any major performance wins.

This year, FF7 will get the meager 45% increase in pixels. Dragons Dogma 2 might have ray tracing on consoles so it might see a substantial boost. Rise of ronin? not so much. Nothing big for TLOU remake, GOW ragnorak, HFW, Demon Souls or Returnal. You better hope Sony studios are going all in on ray tracing in the future because otherwise the jump is very mediocre in comparison to last gen's mid gen consoles.
I change my mind and go to bed
 

Quantum253

Member
why are we assuming 60 CUs again? 54 CUs at 2.4 Ghz gets us 16.58 and 33.7 tflops. And it wont break BC.
I mentioned in another thread that
around the 12 min mark he mentions 56CU @2.2ghz (10.3TFLOPs to 33.5 FP32). I think that's how he gets the 42% increase
 

SlimySnake

Flashless at the Golden Globes
I don't know man.
You and me are trying to figure this out.

Which is it Kepler?


2450GHz has to be 54CUs.

64 × 54 CUs × 2 × 2 × 2425 = 33.5 TF.

likely the latter. no way in hell do they reduce the clocks. it will cause all kinds of issues.

unless they can increase the clocks in BC mode.

I mean, it's like I'm in the twilight zone! Insane how many people are so focused on pure raster when that's not even the point of the console. Imagine discussing nvidia gpu and not focus on DLSS, RT cores. If that were the case, every PC gamer would favor AMD for raster performance edge. Literally the same thing happening here.
Because we have Nvidia GPUs to compare against. We know exactly how this thing will perform. You are getting a 3070 best case scenario with nvidia quality ray tracing and dlss, not a 3080. 3080 is what you would need to get 2x more performance. I have that card and i dont even get that in non-rt games because in raster its only 60-70% more powerful than the PS5.

If you wanted a 2x upgrade, this aint it. In RT or non-RT games. It's an ok boost with some forward thinking features, but 45% is low and will affect half of the games.
 

Killer8

Member
I feel like the rasterization numbers are a bit of a red herring. Everything is moving towards ray tracing and AI image reconstruction nowadays. Seeing the ray tracing numbers is pretty exciting and should hopefully allow more games to run with the full suite of RT features instead of just one or two. If PSSR is better than FSR2 and potentially up there with AI solutions like DLSS (we'll see), it would automatically make the unit punch well above it's weight.
 

Gaiff

SBI’s Resident Gaslighter
What about rebirth and part 3? I ask because I trust your wisdom on this
Part III I haven't got the faintest idea. Rebirth, I think yes. Maybe? It's already 1152p in Performance Mode so 1440p is only a 56% increase in pixel count or 25% per axis. Once again though, I wouldn't worry too much about raw pixel numbers. The selling point of the Pro is its upscaling technique. If it's as good as DLSS, then you could probably get a better image quality than 1440p for fewer resources.
 

Beechos

Member
for you maybe.
im sick of subpar performance games
This is prob the most stable of all gens so far. Most everything is running at 30fps or more. What games are there that slows thing down to like the teens like in past gens. I guess if you must insist everything be 60fps+.
 

Mr Moose

Member
Compute between a 10.7 tflops 6600xt and 16.1 tflops 6800 is 51% and that translates to 54% in performance.

So if this 62% increase in compute is translating into just 45% more performance than there is something off and we can no longer take 6800 as the PS5 equivalent in raster.

Either the lack of a CPU upgrade is holding them back or Cerny was right all along and increasing CUs without infinity cache is simply not going to give you a 1:1 performance uplift.
My 3060 is over 13TF (it doesn't cap at 1777Mhz) and the PS5 out performs it. TFLOPS don't mean shit.
 

Hunnybun

Member
This is pure raw power of this GPU, of course in RT it will be much faster than 7700XT but only some games have RT. New upscaling will change image quality but nothing else, rendering game in 1440p and upscaling to 4k will still "cost" the same (maybe slighty less compared to FSR2 thanks to dedicated hardware).

Ai stuff is mystery, in PC space tensor cores are used only for upscaling so far.

From what I've seen of FSR and DLSS, you're talking about a significant reduction in the required base resolution for the same IQ. If PSSR is on that kind of level then it's not some minor difference. I'd say in the region of a 50% reduction in base resolution.

My bet is it ends up being the most significant of the 3 major improvements (compute, RT and PSSR).
 

SlimySnake

Flashless at the Golden Globes
My 3060 is over 13TF (it doesn't cap at 1777Mhz) and the PS5 out performs it. TFLOPS don't mean shit.
I am simply going by Sony's own 45% figure. If we were going by tflops, it would be 60% more powerful like we all assumed last year in the other thread. But those extra tflops are not translating into equal performance.

We can also throw out any IPC gains when switching from RDNA2 to RDNA4. Its good they are getting great RT performance but this will not impact standard rasterization which is simply not a big enough upgrade. For example, the 1080p 60 fps non-RT mode in star wars is not even going to get you 1440p 60 fps because that would require 75% more power. At best, you will upgrade from the current 4k FSR Performance mode to 4k DLSS Balanced.
 

Gaiff

SBI’s Resident Gaslighter
My 3060 is over 13TF (it doesn't cap at 1777Mhz) and the PS5 out performs it. TFLOPS don't mean shit.
Yes and no. You're correct that the PS5 outperforms it and to a certain extent, TFLOPs are misleading but, that's because both AMD and NVIDIA do all kinds of fuckery to trick people into thinking the TFLOPs number are a measure of graphics power when they're only part of the equation.

You can't just look at the TFLOPs numbers on the specs sheets since Ampere for NVIDIA and RDNA3 for AMD. The "real" compute performance in gaming workloads is hidden in plain sight.
 
Last edited:
likely the latter. no way in hell do they reduce the clocks. it will cause all kinds of issues.

unless they can increase the clocks in BC mode.


Because we have Nvidia GPUs to compare against. We know exactly how this thing will perform. You are getting a 3070 best case scenario with nvidia quality ray tracing and dlss, not a 3080. 3080 is what you would need to get 2x more performance. I have that card and i dont even get that in non-rt games because in raster its only 60-70% more powerful than the PS5.

If you wanted a 2x upgrade, this aint it. In RT or non-RT games. It's an ok boost with some forward thinking features, but 45% is low and will affect half of the games.

I think an RTX 3070 + 300 AI TOPS (TOPS is greater than RTX3090, between RTX4060 and RTX4060Ti) is going to yield some interesting results, while the 45% won't be enough to get 30 to 60, some games inherently have headroom at their 30fps target (30-50fps depending on scenario) I think once you couple that with PSSR, we'll be looking at some interesting results. Not sure how easy PSSR is going to be to implement, or how often it will be implemented.

I believe this will be delivered for $499/$549 which will create another unique value proposition, like the XSX and PS5 did when they launched.
 
So, the clown, Moore's Bait Is Dead, is the broken clock that's right once in a blue moon. Tom Henderson doesn't miss, he's the most credible of them all when it comes to "insider" knowledge.

The GPU downclock makes me question either active CU count of 60 that Kepler is swearing by on Twitter, or if the clocks remain the same, then the TFLOPS will be way higher than 33.5.
 

Mr Moose

Member
I am simply going by Sony's own 45% figure. If we were going by tflops, it would be 60% more powerful like we all assumed last year in the other thread. But those extra tflops are not translating into equal performance.

We can also throw out any IPC gains when switching from RDNA2 to RDNA4. Its good they are getting great RT performance but this will not impact standard rasterization which is simply not a big enough upgrade. For example, the 1080p 60 fps non-RT mode in star wars is not even going to get you 1440p 60 fps because that would require 75% more power. At best, you will upgrade from the current 4k FSR Performance mode to 4k DLSS Balanced.
They haven't even announced the console, it comes from a YouTuber.
 

SlimySnake

Flashless at the Golden Globes
They haven't even announced the console, it comes from a YouTuber.
well, thats all we have to discuss in this thread. its confirmed to be actual documentation from sony themselves by multiple sources. if it turns out to be fake then fine. I am only discussing the topic of the thread.
 

Bojji

Member
My 3060 is over 13TF (it doesn't cap at 1777Mhz) and the PS5 out performs it. TFLOPS don't mean shit.

It performs like 7.5TF 2070, you can't look at ampere numbers and compare it to RDNA1/2.

From what I've seen of FSR and DLSS, you're talking about a significant reduction in the required base resolution for the same IQ. If PSSR is on that kind of level then it's not some minor difference. I'd say in the region of a 50% reduction in base resolution.

My bet is it ends up being the most significant of the 3 major improvements (compute, RT and PSSR).

Yeah, DLSS2 produce great image from 1080p source on 4k screen. But not with 720p image... Games that are 720p and use FSR2 look like garbage, with this console they will render at higher resolution but 1080p may not be achievable as base resolution to upscale with this (not that great) power increase.
 

Mr Moose

Member
well, thats all we have to discuss in this thread. its confirmed to be actual documentation from sony themselves by multiple sources. if it turns out to be fake then fine. I am only discussing the topic of the thread.
If it's just 45% or whatever has been said with the fancy upscaling, then it's shit. I am hoping it's more than that, I hope the upscaling tech is better than FSR is.
It performs like 7.5TF 2070, you can't look at ampere numbers and compare it to RDNA1/2.
It's a bit better than a 2070 but yeah that's what I mean, TF isn't everything.
 
Last edited:

SlimySnake

Flashless at the Golden Globes
I think an RTX 3070 + 300 AI TOPS (TOPS is greater than RTX3090, between RTX4060 and RTX4060Ti) is going to yield some interesting results, while the 45% won't be enough to get 30 to 60, some games inherently have headroom at their 30fps target (30-50fps depending on scenario) I think once you couple that with PSSR, we'll be looking at some interesting results. Not sure how easy PSSR is going to be to implement, or how often it will be implemented.

I believe this will be delivered for $499/$549 which will create another unique value proposition, like the XSX and PS5 did when they launched.
Are the 40 series AI accelerators used for anything other than DLSS? is framegen using it?

Its definitely going to have interesting results. I am just pushing back against this notion that the people calling out the 45% raster upgrade are out of their minds. There is a lot to like here for ray tracing and dlss fans, but 45% is even lower than the 60% we were hoping for in the rumor threads based on the rumored tflops.

Lets hope Cerny Super Sampling is as easy to implement as DLSS because i really dont have any faith in these devs. Lets not forget the abomination that was RDR2's checkerboard implementation.
 
Last edited:

saintjules

Member
A reminder to everyone that wants one to start saving now. Instead of counting those TFs and CUs count those dollar bills.

Hypothetically if the Pro released on the week of November 11:

$600 USD (guessing price) / 35 weeks from now, you put away $17 or so a week and it won't feel so expensive.
 

Gaiff

SBI’s Resident Gaslighter
Are the 40 series AI accelerators used for anything other than DLSS? is framegen using it?
Yeah, frame gen does use tensor cores and motion flow accelerators (that are also present on Ampere and Turing) but the greedy cunts claimed the results on those archs for frame gen weren’t satisfying.
 

Bojji

Member
If it's just 45% or whatever has been said with the fancy upscaling, then it's shit. I am hoping it's more than that, I hope the upscaling tech is better than FSR is.It's a bit better than a 2070 but yeah that's what I mean, TF isn't everything.

This dual issue stuff really fucks everything up. Now we have 33.5TF console and some people are shitting their pants from happiness haha. When in reality it's on 6800 level, mid RDNA2 GPU from 2020.
 
Last edited:

ChiefDada

Gold Member
Because we have Nvidia GPUs to compare against. We know exactly how this thing will perform. You are getting a 3070 best case scenario with nvidia quality ray tracing and dlss, not a 3080. 3080 is what you would need to get 2x more performance. I have that card and i dont even get that in non-rt games because in raster its only 60-70% more powerful than the PS5.

If you wanted a 2x upgrade, this aint it. In RT or non-RT games. It's an ok boost with some forward thinking features, but 45% is low and will affect half of the games.

3070 at best!??? Really Slimy?

sBiLKu7.jpg


Base PS5 already can compete against 3070 in games. And at 2-4x RT uplift, that is above Ampere RT performance, will further push ahead of these GPUs you claim are comparable. If you really think your 3080 is besting the PS5 Pro in RT then you're in for a rude (or pleasant) awakening.
 

MikeM

Member
I feel like the rasterization numbers are a bit of a red herring. Everything is moving towards ray tracing and AI image reconstruction nowadays. Seeing the ray tracing numbers is pretty exciting and should hopefully allow more games to run with the full suite of RT features instead of just one or two. If PSSR is better than FSR2 and potentially up there with AI solutions like DLSS (we'll see), it would automatically make the unit punch well above it's weight.
I wonder if the bump in raster is to give PSSR a better shot at being effective. We all know upscalers are better with higher base resolutions. Maybe Sony has a hard minimum resolution for it?
 
Last edited:

EDMIX

Member
Look...we didn't need a leak to know this shit was real.

To anyone who kept doubting this shit, why did you really need a "leak" to confirm Sony likes money? What the fuck would they gain NOT releasing a upgrade that is completely optional?
No but the power is irrelevant. Sony just doesn't care.

I agree.

I'm sure it can be done, but shit how many would give a fuck? If you love PS3, you'd still have a working one tbh. I have several.

Its not even saying I'd never use it, merely that maybe once a year I'll go back and play thru an old PS3 game, but that is becoming more rare with lots of remakes and remasters coming from that gen.

So lets say they do some Killzone 2 or Motorstorm remake, they would I play the PS3 version? Realistically, what is even left that has not been remastered or remake that is even worth BC with like less then 2% that even use the function?

So it would be nice, I'm not saying otherwise, but at this point, I don't even know what i'd even play with the BC, 3D Dot Hero....maybe thats it, its like....most of what I re-play is being remade, remastered etc.

MGS4 is rumored to be remastered and so is GTA IV, so even those titles wouldn't really be something I'd go back for, so the deeper we go into the generation, the more reasons I have to not go back to the PS3 gen. It was a great time, but anyone who really cares about that library should be playing native in the first place and maybe should have never sold their PS3 if they gave such a fuck lol
 
Are the 40 series AI accelerators used for anything other than DLSS? is framegen using it?

Its definitely going to have interesting results. I am just pushing back against this notion that the people calling out the 45% raster upgrade are out of their minds. There is a lot to like here for ray tracing and dlss fans, but 45% is even lower than the 60% we were hoping for in the rumor threads based on the rumored tflops.

Lets hope Cerny Super Sampling is as easy to implement as DLSS because i really dont have any faith in these devs. Lets not forget the abomination that was RDR2's checkerboard implementation.

That's a good question, I am pretty sure that DLSS/Framegen operations mostly "live" within tensor cores, I am sure something workload related to DLSS gets spread across the shaders, not 100% sure, but here's an interesting article:




Yeah, if it's checkerboarding 2.0, count me out. I'm just hoping for something comparable to DLSS, but better than FSR at least.
 
Top Bottom