playsaves3
Banned
I was just answering the guys question on what specs Sony would use in this instance
-
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 devkits arrive at third-party studios, Sony expects Pro specs to leak
- Thread starter Bernoulli
- Start date
playsaves3
Banned
we really should want a 699 pro for them to go all in
ChiefDada
Member
Are we 100% sure about this? I don't think a 60% boost in TF via more CUs and a fractional increase in bandwidth would be enough to yield 45% Raw performance uptick. Consider the PS4 vs PS4 Pro which doubled CU AND
33 TF comparable to RDNA2 is plausible imo if the 45% raw raster uplift is to be believed.
Last edited:
playsaves3
Banned
all rt modes from the base model will run completely maxed out now maybe they can go in and a new rt mode that's more like high end pc
playsaves3
Banned
I still think these raster improvement estimates seem super low and conservative Im still expecting a 60-80% raw raster increase everything else looks legit though
playsaves3
Banned
uh isnt that what nearly every leak has been saying for almost a year now. it would be weird if it was much lower than that.
playsaves3
Banned
the guy got a little confused go easy on him and be nicer
Gaiff
SBI’s Resident Gaslighter
This would make it faster than the 7700 XT which is right around 45% faster than the 6700 we compare the regular PS5. That’s without the bandwidth increase too.
Last edited:
thicc_girls_are_teh_best
Member
I just want a sensible box that won't make things impossible for a PS6 that is sensibly powered & offers more than just prettier graphics for the next go-round. Plus I think once you start pushing into the $700+ territory, you start to lose out on part of the perk of consoles vs. PCs.
Dunno; I'm only just saying the 33.5 TF mentioned in the video is probably about dual-issue shader performance, and games would have to specifically target that to get that boost in compute.
FWIW the 45% improvement is just in raster performance; they're claiming for RT the improvements could be as high as 400% over the current PS5 due to the PSSR and other changes.
peish
Member
ChiefDada
Member
The numbers get tricky depending on what resolution you're using to compare. Let's say we put PS5 at 2080ti for comparison sake. Yes it's on the more optimistic end for PS5 but not unheard of in select games. I do this to normalize since techpowerup has 4k baseline for 2080ti cards and better. 45% pure raster would place it slightly above 6900xt level which is 23 TF RDNA2 card.
Gaiff
SBI’s Resident Gaslighter
I think it’s better to lean more conservative with our estimates. Whatever the case, 34 TFLOPs dual-issue compute should put it above a 45% performance increase in rasterzation even though it’s actually only around 60% in single-issue compute. That’s also without considering the lower bus width and bandwidth of the 7700 XT.The numbers get tricky depending on what resolution you're using to compare. Let's say we put PS5 at 2080ti for comparison sake. Yes it's on the more optimistic end for PS5 but not unheard of in select games. I do this to normalize since techpowerup has 4k baseline for 2080ti cards and better. 45% pure raster would place it slightly above 6900xt level which is 23 TF RDNA2 card.
6900 XT’s level of performance isn’t completely impossible and unlike RDNA2 cards, RDNA3 and up don’t crumble at 4K.
Last edited:
Topher
Identifies as young
Not really that weird for most leaks to be completely wrong, no. Not saying it is impossible. Just unlikely.
Last edited:
ChiefDada
Member
It would be insane even for him to be this specific with bs rumor since GDC is right around the corner and inevitably more leaks from more conventional outlets will either prove or disprove his leak very soon. MLiD (and RGT) tends to be very broad and cover all possibilities with most of his rumors. This is all worlds of different from his regular shtick; full of details which is why I decided to pay attention this time around. He injected much needed vitality into this thread at the very least.
Topher
Identifies as young
Pretty sure the video claimed it was FP16 67 TF. FP32 would be half that.
Gaiff
SBI’s Resident Gaslighter
A bit of a different take. It's not FP16 vs FP32. It's FP32 but single-issue vs dual-issue. The numbers AMD reports for FP32 compute on RDNA3 cards are for dual-issue FP32. They probably got tired of all those huge NVIDIA TFLOPs numbers that made them look weak and decided to play the same game.
Just like with double and single precision, just halve the number to get the "real" world performance.
It's explained in more details here if you're interested.
https://chipsandcheese.com/2023/01/07/microbenchmarking-amds-rdna-3-graphics-architecture/
James Sawyer Ford
Member
hmm i doubt that
Sony said cost cutting was going to be very difficult
I’m not expecting any decreases in price this year
Topher
Identifies as young
Ah......ok. I misread
thicc_girls_are_teh_best
then. So marketing bullshit has finally made teraflops completely meaningless, it would seem.
Gaiff
SBI’s Resident Gaslighter
Well, yeah, but there are some interesting implications nonetheless.Ah......ok. I misread
Will we see real gaming applications of that? Probably not. At least not anytime soon because Sony won't start to program games for the PS5 and PS5 Pro differently since RDNA2's SIMD's don't have dual-issue capablities.
foamdino
Member
These look like reasonable specs and they align with some of the rumours over the last few months.
I'm pretty sure these (from MLiD) are the real deal - the timing isn't a coincidence either. As GDC is right around the corner, devs are travelling, anyone wanting to leak stuff can do it much more easily now.
Most interesting parts are the changes made to improve RT and the PsSSR solution. Both of these have been heavily rumoured, so I'm glad to see that these are true.
Look forward to the release (and the fact that CNBC is reporting a pro release this year makes it pretty damn likely)
I'm pretty sure these (from MLiD) are the real deal - the timing isn't a coincidence either. As GDC is right around the corner, devs are travelling, anyone wanting to leak stuff can do it much more easily now.
Most interesting parts are the changes made to improve RT and the PsSSR solution. Both of these have been heavily rumoured, so I'm glad to see that these are true.
Look forward to the release (and the fact that CNBC is reporting a pro release this year makes it pretty damn likely)
James Sawyer Ford
Member
Wasn’t PS5 revealed just before GDC?
I wonder if Sony will announce the pro @ GDC
I wonder if Sony will announce the pro @ GDC
MasterCornholio
Member
I'm also wondering if Microsoft will try to stop it.
Topher
Identifies as young
Yeah, but the article downlplays its impact a bit due to the fact that it relies so much on the compiler which it says are not up to it. Shades of "waiting for tools". Also the article notes that this comes into play for optimization and avoiding bottlenecks. That's not something that can be calculated in a formula like the one used to determine teraflops.
Interesting stuff either way.
Last edited:
Gaiff
SBI’s Resident Gaslighter
This is a bit reminiscent of Intel's scores getting invalidated when enabling Advanced Performance Optimizations in 3D Mark. Right around or before the release of their Arc GPUs, Intel had Advanced Performance Optimization enabled in 3D Mark. This resulted in much higher benchmark scores due to some of the shaders being replaced by their own and boosting performance substantially.
Top score is with Advanced Performance Optimizations enabled and bottom is disabled. 3D Mark invalidates scores with it turned on because it's kind of cheating. Still, I think it's cool to have a real-life application of that, even if in the end it amounted to nothing tangible. Not saying Sony will do that but we know it's at least possible.
vivftp
Member
Nah. They still want to sell PS5 consoles over the next several months til the Pro arrives. Announcing it so early doesn't make any sense as it just tells people to wait til later. They'll likely do what they did with the PS4 Pro and announce it in September for a November launch
bitbydeath
Member
Weren’t most/all PS5 leaks wrong?
Borowski_1
Member
Where are the games tho?
Topher
Identifies as young
Some got pretty close, as I recall, but don't think anyone got it dead on.
Lagspike_exe
Member
The APU was fully leaked, though.
ChiefDada
Member
Some support for MLiD leak from Kepler.
Well it looks like they found a way. 2-4x RT performance uplift is excellent.
Well it looks like they found a way. 2-4x RT performance uplift is excellent.
So its going to be a little slower than a 3080 in raster and will be more bandwidth-limited...I won't lie that is quite disappointing. I thought a 3080 baseline would be the bare minimum but at least there are large gains in reconstruction and rt. Honestly, the only option they would have to showcase a difference is just rt gi patches for all games because that's all its going to be capable of with improved image stability.
Last edited:
Zathalus
Member
I'm not doubting the RT performance, I'm wondering if it might have more RAM then the base PS5.
Panajev2001a
GAF's Pleasant Genius
Maybe a bit more OS dedicated RAM that frees some more memory for developers. It depends on how much they will try to enable 8K or not. Part of the extra 1 GB of RAM freed for developers was to allow for 4K buffers and UI.
A lot more RAM seems nice but it would be mostly unused or would be additional work for devs that may affect the base console negatively. PS5 Pro is not meant to have exclusive software period, it is not meant to be the target console for any dev. So yes, maybe an extra 0.5- 1 GB of RAM dedicated to the OS to reduce or remove OS related accesses to the fast GDDR main pool game code running on CPU and GPU will access (which will improve performance a bit too in and of itself).
ChiefDada
Member
ChiefDada
Member
In the context of real world performance, why?
Because even in real-world performance this is looking like its going to end up inferior to a 3080/4070 which I thought was the minimum bar for a worthy mid gen upgrade. Its bandwidth starved, is quite limited in terms of a compute jump compared even to something like the ps4pro, and it would be optimistic to say its going to equal nvidia in reconstruction and rt In amds first real try with dedicated blocks.
Last edited:
Investor9872
Member
I think you'll be disappointed then.
Gaiff
SBI’s Resident Gaslighter
I was looking around to find what the ray tracing performance of the Pro might be like using those 2-4x RT performance increase figures.
Using the 6700 XT as a baseline, we'd end up with RT performance anywhere between a 6800 XT (2x) to a 3080 12GB (4x). This is a pure ray tracing benchmark (path tracing). Obviously, we probably won't get a fully path-traced game on consoles this generation but we can perhaps extrapolate a bit using this little graph. This would make the pure RT performance of the Pro on par with the 3080 so assuming a similar workload, the Pro should lose about as much performance as the 3080 12GB when turning RT features on in a best-case scenario. I do think the ray tracing on PlayStation is more efficient than on AMD GPUs on PC though. This might seem disappointing but do note that in a best-case scenario in a path-traced benchmark, the Pro would be a bit faster than a 7900 XTX. That's quite good I'd say.
A bit shocked at how bad AMD is at ray tracing I gotta say. Also, the 4090 is a fucking monster.
Using the 6700 XT as a baseline, we'd end up with RT performance anywhere between a 6800 XT (2x) to a 3080 12GB (4x). This is a pure ray tracing benchmark (path tracing). Obviously, we probably won't get a fully path-traced game on consoles this generation but we can perhaps extrapolate a bit using this little graph. This would make the pure RT performance of the Pro on par with the 3080 so assuming a similar workload, the Pro should lose about as much performance as the 3080 12GB when turning RT features on in a best-case scenario. I do think the ray tracing on PlayStation is more efficient than on AMD GPUs on PC though. This might seem disappointing but do note that in a best-case scenario in a path-traced benchmark, the Pro would be a bit faster than a 7900 XTX. That's quite good I'd say.
A bit shocked at how bad AMD is at ray tracing I gotta say. Also, the 4090 is a fucking monster.
Last edited:
Rocco Schiavone
Member
The Pro better runs GTAVI at 60fps or theres no reason for me to buy It. Otherwise I'm building a new gaming PC just for GTAVI.
Panajev2001a
GAF's Pleasant Genius
Maybe a bit more OS dedicated RAM that frees some more memory for developers. It depends on how much they will try to enable 8K or not. Part of the extra 1 GB of RAM freed for developers was to allow for 4K buffers and UI.
A lot more RAM seems nice but it would be mostly unused or would be additional work for devs that may affect the base console negatively. PS5 Pro is not meant to have exclusive software period, it is not meant to be the target console for any dev.
Also, 1080p takes less RAM than full 4K buffers, so the extra memory requirements will not hurt as badly and I do think they will offer a bit of extra RAM to devs to compensate.
In a lot of ways, better framerate, much better RT effects (not saying necessarily more pervasive in many titles), and higher image quality is what to expect. Better than PS5, more refined, smoother and more stable, higher IQ, but not designed to look a generation apart (held back by cross-generation considerations by design and a test-bed for many PS6 ideas).
I do think they are preparing to unleash this puppy a bit more in case Xbox Next releases before PS6 and they want to soften the blow of the news at the time while they prepare the pre-PS6 launch PR.
Panajev2001a
GAF's Pleasant Genius
Even without LFC this is well within VRR window. It would be lovely if consoles actually offered locked 50 Hz framerates most displays would support… such an unused option…I was looking around to find what the ray tracing performance of the Pro might be like using those 2-4x RT performance increase figures.
Using the 6700 XT as a baseline, we'd end up with RT performance anywhere between a 6800 XT (2x) to a 3080 12GB (4x). This is a pure ray tracing benchmark (path tracing). Obviously, we probably won't get a fully path-traced game on consoles this generation but we can perhaps extrapolate a bit using this little graph. This would make the pure RT performance of the Pro on par with the 3080 so assuming a similar workload, the Pro should lose about as much performance as the 3080 12GB when turning RT features on in a best-case scenario. I do think the ray tracing on PlayStation is more efficient than on AMD GPUs on PC though. This might seem disappointing but do note that in a best-case scenario in a path-traced benchmark, the Pro would be a bit faster than a 7900 XTX. That's quite good I'd say.
A bit shocked at how bad AMD is at ray tracing I gotta say. Also, the 4090 is a fucking monster.
Gaiff
SBI’s Resident Gaslighter
Don't think so. 45% would only put it on the level of a 6800 and in some cases, barely above the 2080 Ti. Look at A Plague Tale for instance which seems to favor compute power.
This isn't a very good performer for the PS5. Here the Series X scales 23% above the PS5 which is suspiciously close to the TFLOPs differential. You also have the 2080 Ti beating it by almost 47% which is unusual, but then look at the 3080 which easily outdoes the 6800 XT when they're neck-and-neck at 1440p. The Pro won't be a meager 45% faster than the basic one, especially not if the 60 CUs we've heard about is true. They won't clock them at just 2.2GHz lol. It'll be 60-80% better in rasterization. You can quote me on that.
Zathalus
Member
Having the SSD act as virtual memory is great, but it doesn't fundamentally address my point. A higher resolution, more advanced upscalers, and additional RT effects will all lead to additional memory utilization compared to the game running on a regular PS5. I'm wondering how they are addressing it, be it more RAM or the OS running on slower DDR4 like the PS4 Pro did. It's speculation on my part, nothing more.I really tried, but you left me no choice.
Physiognomonics
Member
The whole thing is odd and numbers do not fit. 300 tops 8bit is a lot and suggest 2450 Mhz GPU clock which doesn't fit with the rest. Why would they downclock the GPU when using 36CUs for PS5 BC?
For instance 2450 Mhz GPU would suggest +37 dual Tflops (+18 single tflops), not 33.
EDIT: Actually this would suggest a GPU clocked at 3.23 Ghz for PS5 BC
Last edited: