Do you think there will be a mid gen refresh console from Sony and Microsoft?
- Thread starter invictis
- Start date
davis.anthony
Veteran
If you look at performance reviews of the PC 3700 you'll see that in a lot of games it's not able to do a locked 60ps as it's IPC is quite low now, and PS5 suffers even more as it's clock speed is also very low by comparison.
Cyberpunk 2077 can't be locked to 60fps on a PC equivalent of PS5's CPU.
Sony might be providing a certain amount of personal customizations or integrated bespoke tech for PS5 Pro's GPU, as they did with the prior Pro system. Maybe, some new AI upscaling techniques beyond FSR, and possibly some type of frame generation technology on achieving higher framerates above 30-40fps. I just don't see Sony [PS5 Pro] simply aiming for higher pixel rasterization without some type of significant boost to performance on selling the system.
Whatever gaming experience Sony is trying to provide in a system that is going to be $100-$150 more than the standard PS5, they will have to prove to consumers and tech-heads why it's worth the additional cost.
That depends on what their target is and if that target needs convincing or not. There are people in this world who'll spend $160 on an in-game cosmetic - there might be a good number who'll spend $600+ on a new 'Pro' console without it doing much more than the old one! Are Sony wanting to make a meaningful device, or just cash in on a subset of their userbase for easy money? The PSPortal to me suggests they aren't against the latter as an option.
I see this type of speculation often, and it always raises the same question for me: What makes people think that Sony would be capable of designing and implementing GPU innovations over and above what AMD can do in their own IP?
The things that would be required here for a big leap, e.g. significant perf/w increases, significant RT performance increases and dedicated AI hardware and AI models for upscaling and frame generation are not applicable only in the console arena. They would be equally beneficial to AMD in the PC space. So if AMD could deliver them - they would.
Sony bringing aspects of RDNA4 which may already incorporate some of these additions ahead of the full RDNA4 release on PC is quite realistic, but expecting Sony to develop these independently of AMD and Frankenstein them into AMD's IP for specific application in the PS5 Pro doesn't seem to be.
I know we have the checkerboard example from PS4P but that had pretty limited applicability that needed to be implemented on a per game basis. I think upscaling techniques have moved beyond that now.
Isn't the HW ID Buffer in PS4Pro precedent?
gamervivek
Regular
The best I'm hoping for is AMD racing to beyond 3Ghz on RDNA4 chips, which was rumored for RDNA3.
So that PS5 Pro can easily clock at 7800XT level with console level TDP.
So that PS5 Pro can easily clock at 7800XT level with console level TDP.
I think it's important to remember that 'physics' is not simply cool ragdoll stuff or whatever. Everything that is moveable and has collision attributes requires physicalization(aka physics). And with GTA, that means a LOT of things. It's one of the big things many people dont appreciate about Bethesda RPG's and the associated demands.
You also have an entire driving model embedded in the game with its own physics, along with other vehicular modes, with various levels of physical interactions possible. NPC's also require AI processes to govern how they act, and that's not just with the simple models randomly wandering around, but also combat AI, driving AI, flying AI, etc. There's also the massive draw distances involved, along with the heavier demands of RT-style lighting. And all this stuff has to be running all at the same time in this hugely dense, high fidelity simulation.
A lot of people simply underestimate how important CPU's still are for games. I think was especially easy to overlook in the XB1/PS4 era because CPU demands were just so relatively light thanks to the lackluster Jaguar CPU's in those consoles that determined how developers built games in that generation.
YET, even in that generation there were actually a lot of 60fps games. Because it really doesn't matter what generation you're talking about, when you're dealing with fixed spec hardware, developers always have the choice whether to target 30fps or 60fps, depending on their ambitions. We've had 60fps games since the 80's.
I think what makes people think that is partly a holdover from days gone when console manufacturers actually did used to work with chip designers to build custom processors for their needs. Chip design is so much more complicated nowadays though, and there's much less room for revolution than in the heyday of innovative console hardware. And with cost restraints and do-it-all APU's, console manufacturers are more reliant than ever on existing tech IP that they can package together into something that will work for themselves.
I'm with you 100% that I dont think either Microsoft nor Sony are offering up actual chip design specifications or targets that are anything beyond what AMD basically already can or will offer. They're moreso just kind of ordering off an a la carte menu of technologies. And given the rumored leak of a late 2024 release for PS5 Pro, I would guess that *if* it has any particular RT boost beyond RDNA3, it wont be some 'Sony special', it will simply be some RDNA4 DNA included, which shouldn't be surprising since RDNA4 is expected to be out by then.
D
Deleted member 11852
Guest
Physics, even when there are tens of thousands of interactions and inter-reactions has not been a challenge for console games since 360/PS3. Oblivion and Skyrim could model the physics of a thousand cheesewheels, it was the rendering side of the engine that struggled to keep up.
The origins of modern compute (GPGPU) hardware was modelling physics. This was largely the basis of why Mark Cerny invested in more CUs in PS4. Video game AI (not ChapGPT) has also been a good fit for compute, once you change your AI to a branch-and-bound structure, i.e. distilling the decision-making process into a bunch of branching determinations, from high level (fight vs flee) down to specific actions (throw a punch, try to kick). Similarly for the artificial stimuli that feeds AI decisions. What can this NPC see, what can the NPC hear.. these are all 3D environmental locality calculations that compute solves really well and these this kind of compute tasks easy to fit between graphics operations.
Game AI is not trying to beat you at chess, games are trying to create a semi-believable world where masses of NPCs are walking around, talking or interacting with the environment which in many cases the game is treating are long, complex animation scripts. There have been a few GDC presentations that have broken this down really well.
Last edited by a moderator:
davis.anthony
Veteran
Didn't they also make changes to async on base PS4
D
Deleted member 11852
Guest
Right. Microsoft and Sony, having specific hardware configurations, provide the development and profiling systems for their platforms, and have access to the environment in which games run, so have unparalleled insight into how game developers are using the hardware and where there are easy wins to modify hardware that could greatly benefit performance. AMD are in a very different position, they have an architecture that needs to fit a white range of architecture capabilities that cannot needs to be more general in its approach.
Microsoft and Sony also have their own in-house developers who can provide frank direction about where bottlenecks are and what improvements might help in very specific hardware configurations.
Not to mention both Microsoft and Sony have massively wide R&D teams looking all sorts of stuff. Sony's DNA is in A/V, they were one of the first companies to put image processing enhancement technology (X-Reality Pro) in a consumer TV, and they make professional movie and TV production equipment. Imagine quality is one of their R&D priorities.
Last edited by a moderator:
I do not believe of any new hardwared dedicated to reconstruction for PS5 Pro. Frame generation is not good from 30fps and AFAIK you'd need a powerful CPU to double PS5 games in many cases.
I think good reconstruction techniques already exist and developers need to use exclusive hardware already there like the ID Buffer that can be used in TAA and anything needing motion vectors like reconstruction. Insomniac and Guerrilla have both excellent reconstruction techniques. But some others can do CBR tech pretty good too like what they used in Monster Hunter World on PS4 Pro or what they used in Dark Souls Remastered. Those clean and efficient solutions just need to be shared and used by more developers. No need of DLSS on consoles yet.
And PS5 cache scrubbers (which is kinda evolved from PS4 volatile bit).
We were talking about major Innovations that can bring a "significant boost to performance", not minor tweaks with limited applicability.
That’s an extremely minor tweak that barely saw any usage and delivered questionable benefits.
How would you know that? Maybe the games where PS5 leads by a 20% performance margin vs XSX (or even has the same performance) is thanks to cache scrubbers.
arandomguy
Veteran
This seems to ignore the realities of how these are businesses ultimately and how IP rights/licensing is going to work. Why is one game company going to just to freely give it's custom bespoke solution away for free to others? Compounding the issue would be that their solution itself may not be openly shareable due to other IP restrictions.
It's likely if Sony (or whatever platform holder) wants to set a universally "good" solution unique to it's platform it will want to provide essentially a black box type solution for developers that then cannot be used elsewhere.
Similar threads
