So...if I'm not mistaken, the one we don't know if the PS5 has is VRS?
What does it do?
What does it do?
That all sounds quite reasonable. We'll know everything soon enough so I agree that with the amount of information available we can do nothing but speculate.
Here, in their Hot Chips presentation, Microsoft talk about the small die area that VRS hardware occupies, alluding to the fact that it's worth it due to the 10-30% performance gain that it provides.
"Variable Rate Shading (VRS)
– Variable Rate Shading increases GPU efficiency by concentrating shader work where it's most needed and reducing shader work in areas where it won't be noticeable. With minimal developer effort, VRS significantly improves GPU performance resulting in more stable and higher resolutions and frame rates with no perceptible loss in visual quality."
Basically it lets them render, for example, the centre viewing area of the frame with more detail and the edges, where you're typically not focusing, with less detail. So you get a performance gain without impacting the viewing experience.
"Variable Rate Shading (VRS)
– Variable Rate Shading increases GPU efficiency by concentrating shader work where it's most needed and reducing shader work in areas where it won't be noticeable. With minimal developer effort, VRS significantly improves GPU performance resulting in more stable and higher resolutions and frame rates with no perceptible loss in visual quality."
Basically it lets them render, for example, the centre viewing area of the frame with more detail and the edges, where you're typically not focusing, with less detail. So you get a performance gain without impacting the viewing experience.
VRS isn't the only feature, which is hardware accelerated and potentially different to PS5 GPU. But here you go a explanation for VRS https://devblogs.microsoft.com/directx/variable-rate-shading-a-scalpel-in-a-world-of-sledgehammers/
"developers have a common set of next-generation tools and performance capabilities that will empower them to deliver transformative gaming experiences across both console and PC."
This is why Microsoft can say they have "full" RDNA2, not that their implementation of RDNA2 is more "full" than Sony's.
I think some people are grasping at straws, for what should be a great marketing claim, from Microsoft's side.
User Banned (3 days): Ignoring staff post
Well, Microsofts strategy this far has been overpromse and underdeliver and Sonys the exact opposite by spoonfeeding information but also releasing impressive next gen looking gamplay of games like ratchet and DS.
Halo Infinite is imo visually disappointing compared to Gears 5 running on a One X.
The best looking gameplay that was shown running on a Series X is actually the updated version of Gears 5 and not a new game.
But I'm sure that great games will eventually released for both consoles and Pre-ordered a PS5 and Series X and looking forward to play Miles Morales and updated Gears 5 und FH4.
Halo Infinite is imo visually disappointing compared to Gears 5 running on a One X.
The best looking gameplay that was shown running on a Series X is actually the updated version of Gears 5 and not a new game.
But I'm sure that great games will eventually released for both consoles and Pre-ordered a PS5 and Series X and looking forward to play Miles Morales and updated Gears 5 und FH4.
Last edited:
Nope, there seems to be differences in hardware, hence MS stated that they've waited for AMD for the most advanced technology. At this point users saying this is just about marketing and can be done via a API are grasping at straws.
Don't get me wrong, I don't think it's going to be a major difference. Both consoles are incredibly powerful and close. But saying "duh marketing" is just silly at this point.
indeed and as with most of these custom features they do not get used in multiplatform games anyway because the hardware/APIs are so different.
For Microsoft, I guess the benefit comes for games that are made with it in mind on PC using DX12.
Call of Duty Modern Warfare uses a software implementation of VRS on PS4 and Xbox.
Games don't use the PS4's checkerboard hardware, they opt for platform-agnostic or a solution that functions better for a specific game.
First, thanks for the link(interesting stuff btw).
And about the features, what is the other one?
(Everything below this point could be wrong and these are my interpretations based on what I read/know, so...I am here to be corrected lol).
SFS is part of Velocity Architeture, and the PS5 has its own I/O stuff, which doesn't mean that SFS couldn't be useful, but not necessary maybe?
DXR, PS5 is capable of RT, we just don't know what are the differences between implementations.
VRS, is the one I know we don't know.
Machine Learning, we know that the PS5 is capable(Based on the second Wired article), but again, we don't know at what extent.
Mesh Shading, I thought its equivalent was the Geometry engine, I was wrong?
As I see it (and uh, yeah, I'm not anyone special, but I do have a pretty good record of projecting this stuff), this doesn't say much beyond the Series X having hardware acceleration for the specific features AMD pointed to as being enabled by RDNA2. I expect the gap between PS5 and Series X to be much smaller than between PS4 Pro and One X (and PS4 and One S), and to actually have some overlap, as Sony have some custom stuff alongside their SOC that MS don't.
Some games (a minority) will show an overall advantage on PS5 as a result, depending on how heavily their design leans on these things.
Sony are going to get developers leveraging their unique stuff, because they're Sony, and because they massively improved their relationship with devs going into last gen. The PS5 sdk, by all accounts, is really developer friendly, so I'm sure those unique features won't be overlooked like say, the PS3's eccentricities.
On the Microsoft side, leaning hard into Direct X 12 Ultimate, and ensuring they have the 'full' RDNA2 feature set, will also encourage developers to use that functionality, since enabling it on Series X, also gets those benefits on PC (or vice versa). Given what the Series X is trying to be, that's a smart move for them. We know that their SDK is a little behind, so I'm sure devs will be closer to leveraging what the PS5 can do early on, than they will be with the Series X, (although naturally neither platform will be doing close to optimal in that regard day one). I do think that advantage will fade.
Both strategies make sense and I do think the balance will change a little as devs get more comfortable Microsoft's SDKs, but I don't think we're going to see the big resolution and performance differences we saw all last gen. I think it's going to be more like running with less settings on high or ultra, that type of thing, with the PS5 having advantages in certain areas, but the Series X having more advantages overall.
Last gen I bought all my multiplats on PS4/PS4 Pro, until the One X came out (well, other than Rock Band 4, cause of all my legacy DLC from the 360 days). Then I bought all my multiplats on One X (unless they had cool PSVR stuff like RE7). This gen I expect to pan out much more like 360 and PS3 era, with Series X *mainly* being where I get multiplats, but PS5 'wins' being common enough that I don't blindly buy one version over another.
And this 'full' support for RDNA2 on Series S/X compared to PS5? I don't expect it to mean much beyond a bit more pixel shader power to throw around.
First party games on both systems are going to be fantastic on a technical level, just as they were this gen. No one is going to regret buying a PS5 over a Series X because of ml, mesh shaders or sampler feedback streaming. But it absolutely makes sense for MS to align the Series X with the PC, and Series X owners will benefit from that in seeing these new features better leveraged than if they were custom like Sony's are. Again, though, Sony's custom features are going to get well leveraged because they're the market leader, and they've been extremely developer friendly with the PS5 and the PS4's hardware and sdks.
Some games (a minority) will show an overall advantage on PS5 as a result, depending on how heavily their design leans on these things.
Sony are going to get developers leveraging their unique stuff, because they're Sony, and because they massively improved their relationship with devs going into last gen. The PS5 sdk, by all accounts, is really developer friendly, so I'm sure those unique features won't be overlooked like say, the PS3's eccentricities.
On the Microsoft side, leaning hard into Direct X 12 Ultimate, and ensuring they have the 'full' RDNA2 feature set, will also encourage developers to use that functionality, since enabling it on Series X, also gets those benefits on PC (or vice versa). Given what the Series X is trying to be, that's a smart move for them. We know that their SDK is a little behind, so I'm sure devs will be closer to leveraging what the PS5 can do early on, than they will be with the Series X, (although naturally neither platform will be doing close to optimal in that regard day one). I do think that advantage will fade.
Both strategies make sense and I do think the balance will change a little as devs get more comfortable Microsoft's SDKs, but I don't think we're going to see the big resolution and performance differences we saw all last gen. I think it's going to be more like running with less settings on high or ultra, that type of thing, with the PS5 having advantages in certain areas, but the Series X having more advantages overall.
Last gen I bought all my multiplats on PS4/PS4 Pro, until the One X came out (well, other than Rock Band 4, cause of all my legacy DLC from the 360 days). Then I bought all my multiplats on One X (unless they had cool PSVR stuff like RE7). This gen I expect to pan out much more like 360 and PS3 era, with Series X *mainly* being where I get multiplats, but PS5 'wins' being common enough that I don't blindly buy one version over another.
And this 'full' support for RDNA2 on Series S/X compared to PS5? I don't expect it to mean much beyond a bit more pixel shader power to throw around.
First party games on both systems are going to be fantastic on a technical level, just as they were this gen. No one is going to regret buying a PS5 over a Series X because of ml, mesh shaders or sampler feedback streaming. But it absolutely makes sense for MS to align the Series X with the PC, and Series X owners will benefit from that in seeing these new features better leveraged than if they were custom like Sony's are. Again, though, Sony's custom features are going to get well leveraged because they're the market leader, and they've been extremely developer friendly with the PS5 and the PS4's hardware and sdks.
You know what would be cool for these no downside, just benefits features? That AMD, Microsoft and Sony sit together and work together on them just like MS and Sony do in the HDR group (founded by both of them).
I don't want console features parity, because this would hinder innovation and different concepts. Both of which are why we are here in gaming or the world in general. But when it comes those small feature so to speak, it would be great if all these companies would work together, so more developers can take advantage of them.
Now that Microsoft have 23 first party studios, it would be reasonable to expect that these hardware accelerated features will be utilised with more frequency as time goes on.
[edit] ... and further to this, down the track as these first party studios do their work, I expect this is when we'll see the Series X start punching above its weight as first party developers make use of the hardware accelerated features. I don't think this should be a controversial statement.
The Road to PS5
PS5 lead system architect Mark Cerny provides a deep dive into PS5’s system architecture and how it will shape the future of games.
youtu.be
"But that feature set is malleable which is to say that we have our own needs for PlayStation and that can factor into what the AMD roadmap becomes. So collaboration is born.
If we bring concepts to AMD that are felt to be widely useful then they can be adopted into RDNA - and used broadly including in PC GPUs.
If the ideas are sufficiently specific to what we're trying to accomplish like the GPU cache scrubbers I was talking about then they end up being just for us.
If you see a similar discrete GPU available as a PC card at roughly the same time as we release our console that means our collaboration with AMD succeeded."
I think Cerny is clear on this.
You do understand that Windows is a complete monopoly when it comes to PC gaming basically? That AMD partner with MS when creating their PC GPU and tailor it around direct x and implementing stuff around that is just as obvious as Nvidia doing the same thing they literally have nowhere else to look on who to collaborate with. That their console GPU closely resemble and implement the same features since it's also running direct x should be quite obvious.
However there is not the same monopoly in console gaming and Xbox is the smallest player on that market out of the big 3 so why would Sony ask to have the same features when they are not running direct x but their own API and hence instead ask for custom hardware that is based around their hardware and API.
It would be really fucking strange if AMD put their custom implementation for the PS5 and it's API on a PC running direct x where it would not be used.
Indeed he is. Similar doesn't mean the same (aka 1:1). Look nobody is saying these are console breaking features, which will give a huge benefit to Xbox. That would be silly. However there are HW differences and Xbox more advanced RDNA2 feature set. Hence full RDNA2.
Rumpestump
Absolutely agreed, hence I said features which have no downside and just benefits. It would be cool if they would work together on those rare and small features. As EvilBoris pointed out many don't seem to use the HW inside the Pro to help with resolution. Wouldn't it be cool, if such small features would be plattform agnostic so every dev takes advantage of them?
Last edited:
It renders parts of the graphics that is not as important (can't see the difference) at lower resolution saving computation.
Matt a developer posting on here developing for both consoles have said both have VRS.
is this your first console gen? Devs never use a systems full capabilities within the first 1-2 years. Look at the 360/ps3. Did launch games look as good as games that came out years later? Did mass effect look as good as mass effect 3?
look at current gen. Did launch games look as good as games released in the last few years?
How do you know that? Plz enlighten us about the PS5 hardware implementation since even DF don't know yet.
Acting as you have insider information is bannable you know that?
agreed. I expect devs like TC, TI, T10, PG games, NT and even Id software to show what the XsX can truly do in a few years.
That's one thing I get too from reading the whole thread. and it's an undeniable fact. but was it really worth it?
If this luanch period end up that PS5 selling millions more due to the fact that they can efficiently produce more, & their devkit and software support were mature long time ago. I think the momentum will continue till the end of 2021, as I expect 3d party games won't differ much between the 2 systems.
by the time XSX shows graphical advatage, sony will be set to release PSVR2, which will provide another excitement & attention to their hardware rather than focus on some graphical advatage from the competition.
So yeah, maybe or maybe not. We all sepculating here. Only time will till really.
the gpu cache scrubbers are likely hardware to allow for a software solution to sampler feedback so it likely won't have the same perf it would otherwise have had if the hardware handled the details for you.
Says who? How about you wait until multiplats are out and we start hearing more about the ins and outs of both consoles before echoing Marketing tweets?
Kind of... Maybe...
Infinity Cache is just AMD's name for Level 3 cache.
Desktop RDNA 2 will have up to 128 MB on-die Level 3 cache
The Xbox Series X has 8 MB of L3 cache.
So... Kind of... Maybe...
more advanced RDNA2 architecture.
I should've said more advanced RDNA2 features. Sorry. As to who? Microsoft and they can't lie about it as a public company. This also fits with the rumors from reliable insiders that Xbox was behind in finishing devkits and so on.
Last edited:
Are you implying that there are untruths in the article written by Andrew Goossen and Jason Ronald? I gather you're using "marketing" as a pejorative, and the article should be disregarded?
A Closer Look at How Xbox Series X|S Integrates Full AMD RDNA 2 Architecture
Could you pick out a few of the key sentences that you think are the most untrue statements?
That it's more advanced in this case once again means tailored and build around direct x.
For RT running on direct x having hardware both on PC and console tailored around it is more "advanced" than something Not tailored for direct x.
Having hardware build around the direct x SFS algorithm is more "advanced" than not having it.
Sony would find no benefit from this "advanced" RDNA2 custom build for direct x when they are not using direct x....
I think the reference is to "more advanced RDNA2 features" which arent in PS5 such as hardware VRS, Mesh shaders, SFS etc.
No one is acting like an insider, its relating to the Xbox blog post above, you know the one the thread is about?
It's possible that PS5 doesn't have the full PC RDNA2 feature set in hardware, but that said feature set isn't more "advanced" than what's in the PS5. It's certainly been hinted by leakers who spoke to Redgamingtech.
Microsoft Xbox Series X's AMD Architecture Deep Dive at Hot Chips 2020
AMD's next-gen console APUs come out of hiding.
I'd recommend watching the video too (not sure how easy it is to find though) they explained a bit on what bottlenecks they tried to tackle with each of the hardware, and the overall design goal for the soc and how to achieve that with the cost/size/power consumption constraints.
I am suggesting that until we know the intricacies of both architectures, it is simply impossible to assert that one is more advanced than the other. It is clear that both vendors have chosen relatively different routes and options/feature sets. But, again, the proof will be in the eating.
That doesn't seem like it's related at all.
Sampler feedback is the texture units giving back information about the textures they sampled.
Cache scrubbers sounds more like something that you'd have when you have other producers/consumers feeding data directly into the ram (the ssd for example) and want to keep coherency without invalidating the whole thing (which means discarding even data that would be valid), but I'm basing this only on the name and the short description Cerny gave.
No Sony have their own custom hardware implementation not build around direct x features and instead their own versions such as the geometry engine GPU scrubbers etc but you won't find those hardware implementation on a PC GPU because it's not meant to run direct x. Running PS5 games on the more "advanced" direct x tailored hardware would not be an improvement at all.
Nah for two reasons.
- This is about hardware and not API
- Nobody would stop Sony from developing their own versions of directX features, while using the HW accelerated parts for them
The RDNA3 features? Not buying it, because when it first started by who knows... Can't remember the name, the person basically said (paraphrased), PS4 Pro had Vega features, perhaps PS5 has RDNA3 features. We know RDNA3 will be a topic of 2022 according to AMD. That's way further out then Vega was when the Pro launched.
Having said that, I've edited my old post, because I wanted to say more advanced RDNA2 features and not more advanced necessarily in the brought term. That's something we probably will never know.
Last edited:
hmm, reads to me as Xbox used the standard RDNA2 that AMD provided while Sony went with a more custom RDNA2 solution, swapping out some features to more custom implementations fitted to their own unique needs. Regardless, I doubt any of the two will have significant gain over the other.
Its absolute silly to think FULL RDNA2 means some great performance edge one has over the other. Both boxes will be crazy close in performance.
Its absolute silly to think FULL RDNA2 means some great performance edge one has over the other. Both boxes will be crazy close in performance.
I agree with you. However I don't think anyone suggested Xbox has "some great performance edge" over the other. Unless said person is already on my ignore list, which means I won't see the post.
yes but you get to choose what data you want. You could for example use it for storing texture metadata and prioritize processing of texture samples within the cache. The whole idea is giving the developer control of the cache eviction policy. You could definitely use it to accelerate texture sampling theoretically by early eviction of all other data. I mean in practice it might be a bad idea, but I could see someone using it for that purpose supposing there's no hardware solution in PS5.
No. Mesh Shading for example is API agnostic but hardware reliant. It works on DX12U, Vulkan, OGL etc as long as the hardware supports it. Turing GPUs and RDNA2 does, RDNA1 doesnt.
PS5 does not have the hardware for Mesh Shading, XSX does. Hence a more advanced RDNA2 featureset.
1.The hardware is build around the software you don't build hardware and hope someone will figure out how to do software around it.
2. They already have done this for several generations including this one. Rember all the benefits direct x12 would enable for Xbox only speak from MS last gen and how lots of people on the banned site at the time had this same conversation and that it would mean Sony could not use it or it's features. Sony already had the same solution with another name in their GNX api and in the end direct x provided nothing GNX could not do.
PS5 have custom hardware for running their implementation of next gen software solutions once again in the PS5 and just because it's not named the same thing as the direct x counterpart does not mean it will perform better or worse than it.
Microsoft Xbox Series X's AMD Architecture Deep Dive at Hot Chips 2020
AMD's next-gen console APUs come out of hiding.www.tomshardware.com
I'd recommend watching the video too (not sure how easy it is to find though) they explained a bit on what bottlenecks they tried to tackle with each of the hardware, and the overall design goal for the soc and how to achieve that with the cost/size/power consumption constraints.
Where in this presentation did you get confirmation of hardware accelerated VRS, mesh shaders etc? I can't find any reference to it in the article. It speaks of hardware accelerated ray tracing, but when it comes to other features it merely has the following to say.
"We still don't have concrete details on whether the Xbox Series X will have any additional hardware related to machine learning, for example like Nvidia's Tensor cores that are used for DLSS 2.0. The slides do mention Machine Learning Acceleration, and note an "ML inference acceleration" of 3-10x for a small die area cost, but that could be via FP16 or INT8 calculations run on the CU clusters."
There was a very long deep dive thread on here based on the difference between mesh shaders and primitive shaders Sony talked about and their geometry engine. The conclusion of that deep dive was that if the API allows good hardware detail you should be able to do pretty much the same thing and Sony's API have always been very "close to the metal".
If you're saying hardware is built with specific software tasks in mind, then I agree, but it sounds like you may not be familiar with how that plays out.
Some features rely on physical hardware to be practical at all, and the software implementation "grows" around that hardware after it becomes available. In many generations, new hardware features (some as basic as multi-core/multi-thread CPUs) weren't supported in game software until well after its deployment (I say this with experience as an old-school G4/G5 owner). Between development experience and individual engine support (which isn't the same as just a renderer), it's fairly common for new hardware to exist "in the wild" before actually being taken advantage of.
Events like GDC/Hot Chips and company-specific events like the AMD presentation yesterday a part of the efforts to make those more features known to developers who will be working with them, in addition to marketing those potential perks to users.
That quote was based on the leaked slides before the presentation, not the full presentation itself.
The full slides have been linked above along with the details on die space vs performance. The first slide lists the hardware features, subesequent slides dive deeper and give details on die space.
Edit: Wet Jimmy kindly linked the relevant slides above.
Yeah, that sounds about right. Would have been silly if the DirectXbox didn't fully support DX12U. It'll be interesting to see how Sony's implementation compare.
They probably don't really need Infinity Cache either since there's only one pool of RAM, but it is a key feature of the desktop GPUs.