Why do you keep making assertions with no evidence to back up the statement? Do you just accept us to blindly accept your assertions as true? If you're going to make claims that are not generally accepted, please provide some reference material so that we can review it and have better understanding. So far you've made like 2 or 3 claims based on your recollection which can be easily disproven. As a result, it's hard to take the claims you make seriously at all.
Sony PlayStation 5 Pro
- Thread starter chris1515
- Start date
davis.anthony
Veteran
I think pretty much every member here would accept that the amount of '60fps' PS5 games that are locked 60fps are far outnumbered by the number of '60fps' games that are not locked.
we're also headed (finally) into a period where last generation support has finally dropped off. And that's where the concern is coming from mainly. To make things look good, object interactivity and density is going to play a larger role, and then comes lighting all of that. All of that is costly, and it's going to hammer both CPU and GPU.
I would agree that 60fps is likely going to be harder and harder maintain for the current generation of consoles going forward. Looking at Avatar for example is really an indication of things to come. It's not atrociously performing, by any means, but it's certainly not locked once there are things happening in the game. We get similar performance result with Alan Wake 2 on PS5. Sort of that hard dip into the 50s. Honestly it's hard to tell just from watching DF videos what is CPU and GPU bound. It's truthfully not always clear. But some titles are certainly worse than others.
Perhaps the 10% boost may be just enough to get things close to lock. But if they want to push things further to 30fps, then that 10% boost isn't going to do much.
DF's example of Baldur's Gate 3 being CPU bound is probably one of the stronger examples for instance. A game which may very well set the tone for all sort of new titles going forward.
Last edited:
I wonder if AMD will have opened up FSR3 to work with PSSR. I see no reason why the hardware can't support frame-gen with the ML hardware upgrades.
Not sure if it'd be performant enough to do both, but rather than 4x res scale it'd be nice to see a target in some games of ~1440-1530p @ 40fps with a ~2x scale and then frame-gen from 40fps to 60fps locked or ~80fps VRR.
Looking forward a little to a theoretical PS6. It'd be pretty great if the system has a built-in, universal, toggleable frame-gen feature for all titles that can hit 40fps internally. In turn giving every game at that level a visual 60fps+. Every game built for the platform from the start that doesn't provide a native 60 target mode could be mandated as having to provide the hooks / expose the relevant information to the system so that it's a fundamental part of the platform rather than something that has to be retrofitted.
My bad, I was going from the 6800 XT "release" date on Techpowerup and not the "availability" date, which was indeed November. PS5 did indeed pip PC RDNA2 to real customer availability by a few days as you say.
My main point was that PS5 isn't really the same generation as PC RDNA 2 though, as it lacks core features of PC / Xbox RDNA 2, using some older elements like the Geometry Engine and ROPs and almost certainly lacking INT4 and INT8 mixed precision support. But it's clearly not PC RDNA 1 equivalent either, as it features RT and clocks really well (as you say).
You probably already know this, but console makers aren't strictly tied to PC GPU generations, and can use things off the roadmap that will be ready on time while not waiting for others that won't. I think calling PS5 "RDNA 2" is an understandable marketing decision, but it kind of obfuscates some interesting differences that hint at the decisions and tradeoffs that must have been made by Sony.
On another note, DF have said that PS5Pro supports VRS, which indicates newer ROP designs. So clearly Sony do want these things when they're available. They may well also be on a newer version of AMD's Geometry Engine that supports amplification shader, and they've probably got Sampler Feedback and that mixed precision int stuff too. It would be funny if Playstation is responsible for games starting to use more features on Xbox too - it's probably already a big part of why developers are actually starting to use Mesh Shader on Xbox.
They are the lead console by a far margin. Developers aren't going to make separate code paths for Xbox. Now that Sony is finally caught up feature wise, perhaps we'll actually see some adoption.
The fact that multiple people are contesting that narrative should tell you that it's not generally accepted. I'm very down to change my mind on the matter but, you must provide evidence. Initially, it was that there is a cpu limit affecting performance. Then it was "actually those 60 fps games are not locked". Now it's there are more unlocked 60fps games that locked. That's all fine and dandy but please provide evidence that 50%+1 games with 60fps modes are not locked. If you do that, then we can all be on the same page. We're not talking about games with 1 off drops here but games that spend a majority of the time below 60fps in that mode.
Doubt as ps5 will be the lead not the pro. Also almost all hardware VRS implementations I've seen with the exception of the Coalition's work on Gears have been poor. A visible degradation in image quality in exchange for an middling performance boost.
That just means we need more good implementations though! Another device to add to the list can only help encourage decent implementations. And while VRS can be a net negative, I think some of the negativity towards it comes from people seeing a zoomed in screenshot of some kind of artifact and then adding to a kind of echochamber.
If it takes DF pausing on a zoomed camera transition of a problematic scene to notice it, it's probably not going to be a problem in reality.
I've been thinking about this for a bit but, if the ps5 pro is only 45% faster in raster rendering than the ps5, how much faster is it than the Series X? The series x has 52 cu's at 1.8ghz but they have that annoying memory problem. IS the PS5 pro only going to be like 25% faster than the x for raster rendering? That would certainly be very disappointing.
agreed.
And VRS is only going to be useful when you are compute bound. But the ideal usecase for VRS is combined with DRS. You run a higher resolution and your variation with VRS ensures you're not having to drop large amounts of resolution. You've just got finer control over things as opposed to turning a large knob on resolution, you now have a finer scale knob as well.
Doom and Gears are currently the front runner for VRS technology. I expect to see more of it in the future.
The only thing we can really safely say at this point, is that games will certainly look sharper and clearer on PS5 Pro.
Yea not much faster unfortunately. I couldn't give an honest answer because we don't know how much cache it has going for it. If it has a lot more than XSX it can probably do a lot more. If it doesn't, I'm not sure.
One of the prevailing theories here is going to be put to rest however: the theory that XSX is a terrible performer due to Memory bandwidth per CU. By release, 5Pro is going to take that crown. And i'm pretty sure it's going to be fine.
So it's good to see that theory die in a fire.
davis.anthony
Veteran
Where?
I'm not going to go through every single PS5 video on Digital Foundry because you don't want to accept it.
So I've done the last 10 games for you instead.
Brothers: A Tale of Two Sons - Not a locked 60fps.
FF7: Rebirth: Not a locked 60fps.
Skull & Bones: Not a perfect lock to 60fps but close enough that I'll allow it a pass.
Helldivers 2: Not a locked 60fps.
Granblue Fantasy Relink: Has 1-2fps drops but close enough to a locked 60fps that I'll give it a pass.
Silent Hill The Short Message: Not a locked 60fps.
Suicide Squad: Not a locked 60fps.
Persona 3 Reloaded: Not a locked 60fps.
Tekken 8: A perfectly locked 60fps.
TLOU2: Not a locked 60fps.
70% of them don't hold a locked 60fps, and in reality only one game (Tekken 8) actually does run a flawless 60fps.
I am done addressing this now, you're an intelligent person, go and watch some Digital Foundry videos.
What? That wasn't me...
Last edited:
There's something maybe a bit interesting about the 33.5 TF performance and talk of being around 45% faster than PS5.
PS5 is 10.23 TF. 45% faster than that is 14.83 TF (all things being equal which they never are).
PRO is 33.5 TF (as far as we know) but even ignoring the dual issue thing and halving that you'd still expect performance around 16.75 TF in old money.
So is there something holding the PRO back a little in real vs theoretical performance compared to the PS5? Is it lacking a large L3 and hitting a BW wall? Does the much wider GPU throttle harder than the one on PS5 (seems plausible)? A bit of both?
Unless there's a decent sized LLC I'm not sure how they could hope to have 2 ~ 4x the RT performance.
A new look at the PS5 (and Pro?) CPU FPUs:
https://chipsandcheese.com/2024/03/20/the-nerfed-fpu-in-ps5s-zen-2-cores/
So definitely cut down, but maybe not too significant under the circumstances?
https://chipsandcheese.com/2024/03/20/the-nerfed-fpu-in-ps5s-zen-2-cores/
So definitely cut down, but maybe not too significant under the circumstances?
Funny)
The least of RT's limitations is the LLC. Consider an Ampere vs. RDNA2 as an example of what I'm talking about. The least effective method for speeding up random pointer chasing in a huge data structure (which doesn't typically fully fit into the cache at all) would be to increase the size of the LLC (which is why ETH mining and ASICs are such a bad fit, as another example). Instead, adding the hardware traversal alone would result in a 2 to 4x speedup. Further improvements could be made by compacting the data structure itself. For instance, if you examine AMD's BVH formats and overall sizes compared to their competitors, you'll quickly realize theirs are significantly more bloated. By making BVH more compact (hello compression and slim formats), you can achieve another 2 to 3x speedup factors in the random walk bandwidth limited scenarios, such as the long range GI. So, I am quite disappointed with what AMD has achieved for the PS5 Pro. I had predicted a 4 to 5x performance growth in RT workloads for it a year or two ago(
Great article.
Yep.
They guy tested it thoroughly and found the nerfed Zen 2 makes virtually no difference in usual gaming applications (like in COD). As I though the FPU silicon is enough while the design helps reducing short power consumption spikes that could create unecessary downclocks. Good design by Cerny and his teams.
Similar threads
- Locked
