• 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.

Next-Gen PS5 & XSX |OT| Console tEch threaD

Status
Not open for further replies.

SonGoku

Member
I suppose this console generation has me thinking about (for fun and genuine curiosity) which console will reach its peak performance most efficiently, within their own engineered parameters if that makes sense 🤔?
Its just a simple equation, the more compute units the more more threads needed to reach same level of utilization as the smaller gpu.
Neither will reach 100% VALU utilization consistently but the PS5 will likely get closer to the peak at any given game. XSX will still perform better

Having more CUs is not inefficient, clock speeds can only take you so far before hitting hard diminishing returns
 

PaintTinJr

Member
Its just a simple equation, the more compute units the more more threads needed to reach same level of utilization as the smaller gpu.
Neither will reach 100% VALU utilization consistently but the PS5 will likely get closer to the peak at any given game. XSX will still perform better

Having more CUs is not inefficient, clock speeds can only take you so far before hitting hard diminishing returns
It is inefficient use of caches if they have to dupplicate cache data to split a workload that could have ran in 1 frame on 1 CU - not to mention the additional coordination of split workloads being another performance cost that on PC is likely incurred by CPU and RAM bandwidth in the memory mapped video driver. For every CU a workload needs shared with, that is reducing cache memory per frame by that factor (AFAIK).

Also, where is this so-called hard limit? If you mean overclocking parts on a PC, then that isn't really comparative because the bottlenecks that mark cerny described would still be there and at orders of magnitudes bigger than the clockboost benefit. But that doesn't mean in a balanced processing setup that clocking higher wouldn't continue to yield greater performance,
 
Last edited:

bitbydeath

Gold Member
A lot of people still don’t understand the evolutionary differences of XSX and PS5.

Comparing kraken and BCPack is irrelevant because with Sony having removed all surrounding bottlenecks means that task performs 100 times faster and not just that task either but all of them.

XSX is still using the traditional method of loading a game, decompressing textures, everything before the game starts.

PS5 has flipped everything upside down and does everything in less than a second on the fly. You turn your character, it decompresses, it maps, it does everything required to generate the scene in under a second.

The fast travel is also quoted as being so fast due to this new practice that it feels unnatural and they’ve had to implement pauses.

These consoles may have similar hardware but they are operating very differently.
 
Last edited:

Bo_Hazem

Banned
I'm all for sony in this console war, but thats not how any of this works.

Both consoles will use ssds. Both consoles won't need duplicate data. So as long as the game developer cares they will be able to push the game without duplicate data to the consoles.
Both consoles won't have seek times for files and so on because they don't use hdds. ( not taking into account harddrives connected via usb )

In case you're wondering what is that coming from, it's if that 100GB is meant to be a partition which is discussed by some here. Let's see this low budget NVMe SSD in action in the future anyway.
 

Bo_Hazem

Banned
I think he might be referring to the engine data that will go into the 10GBs(effectively VRAM) of XsX. Mark explained a very simple setup of streaming just FOV data a full quality in the time it took the player to rotate. The split access pattern on XsX - along with slower SSD and no IO complex - may result in some of the old design style loading of non-visible data being needed to multi-plat PS5 design lead games, which in turn would require a drop in the FOV asset quality because the effective size of 10GB would be reduced by the out of view assets.

Could we see in the next gen assets being loaded in a 3-4 phase style (potato-low-mid-high quality) like in Assassin's Creed Odyssey when you turn and see assets taking time to load?
 

Bo_Hazem

Banned
It is inefficient use of caches if they have to dupplicate cache data to split a workload that could have ran in 1 frame on 1 CU - not to mention the additional coordination of split workloads being another performance cost that on PC is likely incurred by CPU and RAM bandwidth in the memory mapped video driver. For every CU a workload needs shared with, that is reducing cache memory per frame by that factor (AFAIK).

Also, where is this so-called hard limit? If you mean overclocking parts on a PC, then that isn't really comparative because the bottlenecks that mark cerny described would still be there and at orders of magnitudes bigger than the clockboost benefit. But that doesn't mean in a balanced processing setup that clocking higher wouldn't continue to yield greater performance,

3waydx.jpg
 
Remember this come from the guy who said to exhaustion the PS5 doesn't has hardware RT because reasons

I will ignore that guy before I he start with his conspirations

What in the world are you talking about? Not only do I have no idea what you are talking about, I've fully backed up everything I've ever said. Meanwhile you are trying to dismiss my point without even an attempt at making a credible argument. You are the one talking nonsense, and it sounds like you know it, which is why you are tying to declare victory and slink away.
 

EliteSmurf

Member
What in the world are you talking about? Not only do I have no idea what you are talking about, I've fully backed up everything I've ever said. Meanwhile you are trying to dismiss my point without even an attempt at making a credible argument. You are the one talking nonsense, and it sounds like you know it, which is why you are tying to declare victory and slink away.
He might be referring to SenjutsuSage SenjutsuSage who believed Cerny was lying about hardware RT after he confirmed the PS5 had it because of ... reasons ?
 

SonGoku

Member
It is inefficient use of caches if they have to dupplicate cache data to split a workload that could have ran in 1 frame on 1 CU - not to mention the additional coordination of split workloads being another performance cost that on PC is likely incurred by CPU and RAM bandwidth in the memory mapped video driver. For every CU a workload needs shared with, that is reducing cache memory per frame by that factor (AFAIK).

Also, where is this so-called hard limit? If you mean overclocking parts on a PC, then that isn't really comparative because the bottlenecks that mark cerny described would still be there and at orders of magnitudes bigger than the clockboost benefit. But that doesn't mean in a balanced processing setup that clocking higher wouldn't continue to yield greater performance,
It depends of the target each company sets
Sony was able to meet their 10TF target by prioritizing frequency but MS's 12TF was likely unreachable without going wider.

I think both consoles frequencies are the sweet spot for their respective chip design
Cerny mentioned they could go higher but they capped it to guarantee the chip logic operates properly, on the DF foundry interview he also mentioned the CPU&GPU clock frequency hit a thermal density sweet spot.

I agree with you, these aren't comparable to overclocks in the slightest, the respective frequencies were carefully selected and tested to hit a sweetspot using a model SoC
 
Last edited:
What in the world are you talking about? Not only do I have no idea what you are talking about, I've fully backed up everything I've ever said. Meanwhile you are trying to dismiss my point without even an attempt at making a credible argument. You are the one talking nonsense, and it sounds like you know it, which is why you are tying to declare victory and slink away.
Hahaha not you EventHorizon, calm down, response was in order you remember what kind of person you were replying
 

rnlval

Member
Correct

Wrong. The same hardware block handles both formats

You are the only person i came across with the (incorrect) interpretation that there's two decompressing blocks. DF is clear, only one
Two decompression blocks for fix function design instead of yet another processor.
 

joe_zazen

Member
OK, so if xsx memory speed takes a nose dive when more than 10GB are used and OS takes 2.5GB, does that mean if more than 7.5 is used for a game the memory speed is worse than ps5?
 
It's the 100GB partition that works like ancient gaming duplicates going forward. If that's not the case, games that are 100GB on XSX might end up being 20GB on PS5, so to avoid that they made this afterthought trick to compensate for its inferior SSD and I/O.

It doesn't work like you're thinking. Don't forget Xbox has a new compression system specifically designed for gpu textures that compresses GPU textures down to 50%+ their original size. GPU textures are responsible for most of a game's size. In fact, the lion's share.

Xbox Series X also has a fully custom streaming system designed for the series x that makes irrelevant all this copy stuff you're referring to. Instead of loading entire textures into physical memory like the PS5 intends to do at high speeds, Xbox will simply do something potentially a lot better and may actually turn out to be even faster with additional benefits on top of that. It will only load the section of the texture that's actually visible in the scene. As Microsoft put it, usually the largest mip of a 4K texture is often 8MB and often more. But if they only need to 25% of the largest mip, that data that would be brought into RAM goes from a potential full 8MB or more down to just 2MB. Imagine all the physical RAM savings this can create across a game. Microsoft is flat out calling Sampler Feedback Streaming a game changer because it will add a multiplying effect to how physical memory is used, upwards of 2x-3x (and in some cases higher) they've stated. But the benefit doesn't stop there. If you are copying way less, much smaller pieces of texture data into physical RAM, it also means you are getting significantly more bang for your buck from your SSD performance, too. If you're copying much less data in general, that's a win, win. Devs usually want as much ram as possible so they don't need to do any copying at all. Series X appears to have a custom built feature that actually comes the closest to achieving this without a much larger jump in available system RAM.

So to reiterate an example you just gave about game sizes being smaller on PS5, that may not be the case at all. Then again, if devs want, due to all the extra RAM they may end up saving due to Sampler Feedback Streaming, they may opt to use even higher quality textures on series x, possibly producing higher quality textures than what may be possible on the PS5 side of things, if only cause they know they can get away with only loading in 25% of the texture.

And just to confirm, though Sampler Feedback may be a supported RDNA2 and even Nvidia Turing feature, Sampler Feedback Streaming isn't actually a standard RDNA2 feature that will automatically be found in all RDNA2 GPUs. That along with all the parts of Xbox Velocity Architecture are custom to Xbox Series X. Sony has a beast of an SSD, but it looks to me like Sony's solution is more based on the raw speed of that SSD whereas Microsoft has taken numerous steps to try and change the way the work is actually carried out. Sampler Feedback Streaming looks more revolutionary than just a plain faster SSD. Basically both Sony and Microsoft have approached the goal in two different ways.



Stanard is a Graphics Optimization R&D & Engine Architect at Microsoft.



 

Imtjnotu

Member
The guy who made the tweets is this guy:

Rich Geldreich is a founder of Binomial, the company behind the Basis Universal Supercompressed GPU Texture Codec. And apparently Microsoft is using this technology in XSX.

https://github.com/BinomialLLC/basis_universal/blob/master/README.md

And apparently he is still working with MS to improve the compression format to match the 6GB/s (minus overhead ) compression of the XSX HW block.

Also I have read that the XSX's implementation of Direct Storage seems to use a special game format dedicated to SSD so both PC and XSX and can take advantage of the storage and compression benefits exposed in the DS API.

This may be one of the reasons why they require XSX enhanced titles to operate strictly from their specified SSD offerings and not USB externally connected devices.
It's all apparently. Where is the actual telling. All he has said mentioning the texture compression is that it potentially could be than kraken but no where has anything been factual thus far
 

PaintTinJr

Member
It depends of the target each company sets
Sony was able to meet their 10TF target by prioritizing frequency but MS's 12TF was likely unreachable without going wider.

I think both consoles frequencies are the sweet spot for their respective chip design
Cerny mentioned they could go higher but they capped it to guarantee the chip logic operates properly, on the DF foundry interview he also mentioned the CPU&GPU clock frequency hit a thermal density sweet spot.

I agree with you, these aren't comparable to overclocks in the slightest, the respective frequencies were carefully selected and tested to hit a sweetspot using a model SoC

From all the stuff I’ve extrapolated from the Sony GDC talk and everything I’ve heard from DF, etc about XsX, I’ve now concluded that the real difference in the consoles is in their clocking strategies, and how they handle clock by clock thermal latency.

PCs, current consoles and the XsX seem to be using the old style thermal latency, where you react 5 seconds or more behind the rise in temp on the chip thermo sensor(so potentially 150 frames or more) and then react by ramping up the fan speed.

The paradigm shift by Sony seems like they are eliminating thermal latency by starting their deterministic workload analysis at frame -1 (for thermal latency and workload), and are instead analysing frame 0’s workload at frame -1, and then boosting without supplying any more power. The net effect being more power that would have idly turned to heat in frame 0 and turned up at frame 150 has been repurposed as more work done in frame 0 – because of an instantaneous undershoot in power needed for the boost clock. The system then turns less of its power to heat at frame 0, and because this info was known at frame -1, the fan has already adjust by a deterministic amount (in frame 0) to counter act the smaller rise, keeping the system cool, quiet, performant, power efficient, and able to find a much higher sweet spot before thermal latency starts to re-emerge.
 
Last edited:

DaGwaphics

Member
PS5
CPU@3.5Ghz@16 threads
1 or 2 for the OS?
14 threads for games?

XSX
CPU@3.6Ghz@16 threads
1 core, 2 threads = reserved for OS, 14 threads for games
OR (no SMT aka Hyperthreading)
2 cores at 3.8Ghz + 6 cores@3.6Ghz
1 core, 1 thread reserved for OS, 6 or 7 for games

What´s the point of the latter option?

I believe all cores are at 3.8 with SMT disabled. This option will likely work well for existing engines that are designed for only 6 threads on the current consoles.
 
Last edited:

SonGoku

Member
OK, so if xsx memory speed takes a nose dive when more than 10GB are used and OS takes 2.5GB, does that mean if more than 7.5 is used for a game the memory speed is worse than ps5?
It drops total average bandwidth but not to the point is lower than PS5. They both have roughly the same GB/s per TF
Whoa, where did that come from? Nice ad!
Playstation magazine
 
Last edited:

SonGoku

Member
Sampler Feedback may be a supported RDNA2 and even Nvidia Turing feature, Sampler Feedback Streaming isn't actually a standard RDNA2 feature that will automatically be found in all RDNA2 GPUs
Sampler Feedback its a software implementation (API) that makes texture space shading (HW) more effective and simpler to use.
Sony wouldn't use DX API anyways, they'll develop their own software implementation to take advantage of TSS.
 

ethomaz

Banned
It doesn't work like you're thinking. Don't forget Xbox has a new compression system specifically designed for gpu textures that compresses GPU textures down to 50%+ their original size. GPU textures are responsible for most of a game's size. In fact, the lion's share.

Xbox Series X also has a fully custom streaming system designed for the series x that makes irrelevant all this copy stuff you're referring to. Instead of loading entire textures into physical memory like the PS5 intends to do at high speeds, Xbox will simply do something potentially a lot better and may actually turn out to be even faster with additional benefits on top of that. It will only load the section of the texture that's actually visible in the scene. As Microsoft put it, usually the largest mip of a 4K texture is often 8MB and often more. But if they only need to 25% of the largest mip, that data that would be brought into RAM goes from a potential full 8MB or more down to just 2MB. Imagine all the physical RAM savings this can create across a game. Microsoft is flat out calling Sampler Feedback Streaming a game changer because it will add a multiplying effect to how physical memory is used, upwards of 2x-3x (and in some cases higher) they've stated. But the benefit doesn't stop there. If you are copying way less, much smaller pieces of texture data into physical RAM, it also means you are getting significantly more bang for your buck from your SSD performance, too. If you're copying much less data in general, that's a win, win. Devs usually want as much ram as possible so they don't need to do any copying at all. Series X appears to have a custom built feature that actually comes the closest to achieving this without a much larger jump in available system RAM.

So to reiterate an example you just gave about game sizes being smaller on PS5, that may not be the case at all. Then again, if devs want, due to all the extra RAM they may end up saving due to Sampler Feedback Streaming, they may opt to use even higher quality textures on series x, possibly producing higher quality textures than what may be possible on the PS5 side of things, if only cause they know they can get away with only loading in 25% of the texture.

And just to confirm, though Sampler Feedback may be a supported RDNA2 and even Nvidia Turing feature, Sampler Feedback Streaming isn't actually a standard RDNA2 feature that will automatically be found in all RDNA2 GPUs. That along with all the parts of Xbox Velocity Architecture are custom to Xbox Series X. Sony has a beast of an SSD, but it looks to me like Sony's solution is more based on the raw speed of that SSD whereas Microsoft has taken numerous steps to try and change the way the work is actually carried out. Sampler Feedback Streaming looks more revolutionary than just a plain faster SSD. Basically both Sony and Microsoft have approached the goal in two different ways.



Stanard is a Graphics Optimization R&D & Engine Architect at Microsoft.




Sampler Feedback is just software... not hardware.

I mean it is a software logic that helps streaming textures and/or TSS.

There is no dedicated hardware involved... just the standard hardware units found in any nVidia and AMD GPUs (probably Intel too but I’m not sure) that supports DX12.
 
Last edited:
Sampler Feedback its a software implementation (API) that makes texture space shading (HW) more effective and simpler to use.
Sony wouldn't use DX API anyways, they'll develop their own software implementation to take advantage of TSS.

Sampler Feedback isn't just a software implementation API as you put it, it requires explicit hardware support to be used. That's like calling Mesh Shaders just a software implementation API not understanding the hardware needs to be there to support it. Same is true for VRS, there needs to be hardware support, up to a specific tier to support the quality of VRS that even Turing GPUs support. Up to this point there's zero evidence PS5 supports Sampler Feedback. Being RDNA2 doesn't automatically mean it is assured to be one thing or another, as confirmed by cerny in his talk.

To date the only features confirmed for the PS5 GPU are ray tracing, Primitive Shaders (a feature from Vega that was finally made active in RDNA1) and he discussed the cache scrubbers they had put in place. Cerny didn't touch on anything else, not even VRS in his talk. And this was billed as a ps5 architecture deep dive. He touched on every piece of custom hardware crucial to the SSD, dug down into primitive shaders on the GPU, cache scrubbers, and went into amazing detail on the audio chip. I would think if other high profile features were there he would have talked about them. Cerny doesn't have a reputation for leaving things out when he decides to go public with specific technical details like what he has done with DF. I once doubted hardware ray tracing prior because the communication made me suspicious as to whether it was truly there, but that's been put to bed with the road to ps5 presentation. Unlike ray tracing where I thought communication seemed off, there has been zero communication on VRS, Sampler Feedback, or Texture Space shading (a feature of sampler feedback)

Sampler Feedback is far, far more than just texture space shading. In order for texture space shading to even work you require sampler feedback to help provide the necessary information.

Sampler Feedback enables better visual quality, shorter load time, and less stuttering by providing detailed information to enable developers to only load in textures when needed.

Suppose you are a game developer shading a complicated 3D scene. The camera moves swiftly throughout the scene, causing some objects to be moved into different levels of detail. Since you need to aggressively optimize for memory, you bind resources to cope with the demand for different LODs. Perhaps you use a texture streaming system; perhaps it uses tiled resources to keep those gigantic 4K mip 0s non-resident if you don’t need them. Anyway, you have a shader which samples a mipped texture using A Very Complicated sampling pattern. Pick your favorite one, say anisotropic.

The sampling in this shader has you asking some questions.

What mip level did it ultimately sample? Seems like a very basic question. In a world before Sampler Feedback there’s no easy way to know. You could cobble together a heuristic. You can get to thinking about the sampling pattern, and make some educated guesses. But 1) You don’t have time for that, and 2) there’s no way it’d be 100% reliable.

Where exactly in the resource did it sample? More specifically, what you really need to know is— which tiles? Could be in the top left corner, or right in the middle of the texture. Your streaming system would really benefit from this so that you’d know which mips to load up next.c

Sampler feedback solves this by allowing a shader to efficiently query what part of a texture would have been needed to satisfy a sampling request, without actually carrying out the sample operation. This information can then be fed back into the game’s asset streaming system, allowing it to make more intelligent, precise decisions about what data to stream in next. In conjunction with the D3D12 tiled resources feature, this allows games to render larger, more detailed textures while using less video memory
.

And where texture space shading comes in, and why it's necessary to have support for sampler feedback to even be able to use texture space shading.



Sampler feedback also enables Texture-space shading (TSS), a rendering technique which de-couples the shading of an object in world space from the rasterization of the shape of that object to the final target.

TSS is a technique that allows game developers to do expensive lighting computations in object space, and write them to a texture— for example, something that looks like a UVW unwrapping of the object. Since nothing is being rasterized the shading can be done using compute, without the graphics pipeline at all. Then, in a separate step, bind the texture and rasterize to screen space, performing a dead simple sample. This approach reduces aliasing and allows computing lighting less often than rasterization. Decoupling these two rates allows the use of more sophisticated lighting techniques at higher framerates.


Setup of a scene using texture-space shading
One obstacle in getting TSS to work well is figuring out what in object space to shade for each object. Everything? That would be hardly efficient. What if only the left-hand side of an object is visible? With the power of sampler feedback, the rasterization step can simply record what texels are being requested and only perform the application’s expensive lighting computation on those.

So as you see, texture space shading is dependent on sampler feedback and it must be supported in hardware. It isn't just some software, because if it was Xbox One X and PS4 Pro could do it. Are there software methods for trying to emulate what sampler feedback does? Absolutely, but the problem is they are simply not as effective and the performance benefits are not as good without hardware support.

Based on how I've seen AMD handle Primitive Shaders and Mesh Shaders, I'm honestly not even sure if Mesh Shaders and Primitive Shaders are even exactly the same thing. As AMD went from calling Primitive Shaders one thing in Vega and RDNA1 to now marketing Mesh Shaders as a specific RDNA 2 feature. It's entirely possible that Mesh Shaders goes a bit further than Primitive Shaders, and so that's why it's marketed as an RDNA 2 feature. Or mesh shaders is just the DX 12 Ultimate name for Primitive Shaders as at a basic level they do really sound very similar. but what's always left out is the level of detail and extra capability we can read about in Mesh Shaders. Nvidia, for example, has always called it Mesh Shaders since turing. Anandtech for example confirmed that primitive shaders gave zero performance benefit in Vega and it wasn't until RDNA that AMD managed to fix some things and enable it, but it may still not be the same as Mesh Shaders. For the sake of argument though because it's hard to tell I'll assume it's the same.
 

ethomaz

Banned
Sampler Feedback isn't just a software implementation API as you put it, it requires explicit hardware support to be used. That's like calling Mesh Shaders just a software implementation API not understanding the hardware needs to be there to support it. Same is true for VRS, there needs to be hardware support, up to a specific tier to support the quality of VRS that even Turing GPUs support. Up to this point there's zero evidence PS5 supports Sampler Feedback. Being RDNA2 doesn't automatically mean it is assured to be one thing or another, as confirmed by cerny in his talk.

To date the only features confirmed for the PS5 GPU are ray tracing, Primitive Shaders (a feature from Vega that was finally made active in RDNA1) and he discussed the cache scrubbers they had put in place. Cerny didn't touch on anything else, not even VRS in his talk. And this was billed as a ps5 architecture deep dive. He touched on every piece of custom hardware crucial to the SSD, dug down into primitive shaders on the GPU, cache scrubbers, and went into amazing detail on the audio chip. I would think if other high profile features were there he would have talked about them. Cerny doesn't have a reputation for leaving things out when he decides to go public with specific technical details like what he has done with DF. I once doubted hardware ray tracing prior because the communication made me suspicious as to whether it was truly there, but that's been put to bed with the road to ps5 presentation. Unlike ray tracing where I thought communication seemed off, there has been zero communication on VRS, Sampler Feedback, or Texture Space shading (a feature of sampler feedback)

Sampler Feedback is far, far more than just texture space shading. In order for texture space shading to even work you require sampler feedback to help provide the necessary information.



And where texture space shading comes in, and why it's necessary to have support for sampler feedback to even be able to use texture space shading.





So as you see, texture space shading is dependent on sampler feedback and it must be supported in hardware. It isn't just some software, because if it was Xbox One X and PS4 Pro could do it. Are there software methods for trying to emulate what sampler feedback does? Absolutely, but the problem is they are simply not as effective and the performance benefits are not as good without hardware support.

Based on how I've seen AMD handle Primitive Shaders and Mesh Shaders, I'm honestly not even sure if Mesh Shaders and Primitive Shaders are even exactly the same thing. As AMD went from calling Primitive Shaders one thing in Vega and RDNA1 to now marketing Mesh Shaders as a specific RDNA 2 feature. It's entirely possible that Mesh Shaders goes a bit further than Primitive Shaders, and so that's why it's marketed as an RDNA 2 feature. Or mesh shaders is just the DX 12 Ultimate name for Primitive Shaders as at a basic level they do really sound very similar. but what's always left out is the level of detail and extra capability we can read about in Mesh Shaders. Nvidia, for example, has always called it Mesh Shaders since turing. Anandtech for example confirmed that primitive shaders gave zero performance benefit in Vega and it wasn't until RDNA that AMD managed to fix some things and enable it, but it may still not be the same as Mesh Shaders. For the sake of argument though because it's hard to tell I'll assume it's the same.
Mesh Shades / Primitive Shaders indeed require hardware feature.

Sampler Feedback not.

MS documentation confirms that.
 
Last edited:

SonGoku

Member
Sampler Feedback isn't just a software implementation API as you put it, it requires explicit hardware support to be used.
Yes and that hardware support its called TSS found in Turing and RDNA2
Nothing stops Sony from developing their own software implementation to properly exploit it.
Up to this point there's zero evidence PS5 supports Sampler Feedback. Being RDNA2 doesn't automatically mean it is assured to be one thing or another, as confirmed by cerny in his talk.
Cerny all but confirmed PS5 will match (and exceed in some areas) the feature set of RDNA2 cards available during that time frame
There's zero evidence that implies PS5 wont support basic RDNA2 features.

PS4 & PS4Pro matched (and exceeded) the feature set of their respective GPU architecture, where does this narrative comes from that Sony will go out of their way to remove basic RDNA2 features from PS5's GPU?
 
Last edited:
Sampler Feedback is just software... not hardware.

I mean it is a software logic that helps streaming textures and/or TSS.

There is no dedicated hardware involved... just the standard hardware units found in any nVidia and AMD GPUs (probably Intel too but I’m not sure).

It's definitely not just software, this is why it was considered a new architectural feature added to Nvidia Turing GPU. It's hardware coded into the shaders. It's a shader capability. It's not just anything that any old AMD or Nvidia GPU can do.

Page 47 touches on it. It's something hardware built into the shaders. It isn't just simple software.


 

ethomaz

Banned
It's definitely not just software, this is why it was considered a new architectural feature added to Nvidia Turing GPU. It's hardware coded into the shaders. It's a shader capability. It's not just anything that any old AMD or Nvidia GPU can do.

Page 47 touches on it. It's something hardware built into the shaders. It isn't just simple software.


?

But old nVidia and AMD GPUs support it.
There is even a limited support in the DirectX 12 Ultimate API for very old hardware that doesn’t support all feature of Sampler Ferdback.

You call a function and it will say if there is no support, limited or full support.

Sampler Feedback is not something that requires hardware function added in for example RDNA 2 or Turing.

Anyway RDNA 2 fully support Sampler Ferdback (Tier 1)... there is no MS custom hardware for that like you supposed.

But Sampler Feedback is logic patented by MS so Sony can’t use that logic but they can created their own logic for the same function.
 
Last edited:

SonGoku

Member
Page 47 touches on it. It's something hardware built into the shaders. It isn't just simple software.
That's what been said to you... TTS is the hardware feature, sampler feedback is a software implementation designed to effectively exploit it
Sampler Feedback is a Direct3D feature for capturing and recording texture sampling information and locations. Without sampler feedback, these details would be opaque to the developer.

rMCOiDc.png
 
?

But old nVidia and AMD GPUs support it.
There is even a limited support in the DirectX 12 Ultimate API for very old hardware that doesn’t support all feature of Sampler Ferdback.

You call a function and it will say if there is no support, limited or full support.

Sampler Feedback is not something that requires hardware function added in for example RDNA 2 or Turing.

Anyway RDNA 2 fully support Sampler Ferdback... there is no MS custom hardware for that like you supposed.

I'll take your word for it then, but apparently my understanding is that older GPUs can't use it in the same way. For example, Microsoft has specific hardware built into the Series X GPU to support Sampler Feedback, as confirmed through digital foundry.
 

ethomaz

Banned
I'll take your word for it then, but apparently my understanding is that older GPUs can't use it in the same way. For example, Microsoft has specific hardware built into the Series X GPU to support Sampler Feedback, as confirmed through digital foundry.
Let’s say what you say is true.
So how any RDNA 2 and Turing GPU supports Sampler Feedback?

They lack that MS “custom” hardware unit only found in Xbox Series X, no?
 
Last edited:

SonGoku

Member
Sampler Feedback is not something that requires hardware function added in for example RDNA 2 or Turing.
I'll take your word for it then, but apparently my understanding is that older GPUs can't use it in the same way. For example, Microsoft has specific hardware built into the Series X GPU to support Sampler Feedback, as confirmed through digital foundry.
While older GPUs can use Sampler Feedback, the GPUs that support it at the hardware level will do it more effectively
Same as VRS, some current gen games used VRS on PS4/XB1 without having hardware support for it, but the new GPUs will be better at it.
 
Last edited:

ethomaz

Banned
While older GPUs can use Sampler Feedback, the GPUs that support it at the hardware level will do it more effectively
Same as VRS, some current gen games used VRS on PS4/XB1 without having hardware support for it, but the new GPUs will be better at it.
Yeap there is 3 tiers support.

- NO TIER = No support
- TIER_0.9 = Limited support for some specific cases
- TIER_1.0 = Full support

There is a function that you can know the TIER support of the hardware you are running the code.

But I’m not entering in which hardware support it or not discussion... RDNA 2 and Turing fully support it so how it can be a customized feature only found in Xbox Series X? It just makes no sense.
 
Last edited:

longdi

Banned
PS5 cooling system is supposed to have a hole in PCB? IIRC from rumors off some patent.
Sorta like this, with its base positioned vertical and dual sided.

thermaltake-spinq-vt-cpu-cooler-2.jpg


This is it guys! The top chip is the apu and bottom is the other parts.
zFDqdoG.png



Well, MS also using a split PCB design too 🤷‍♀️
 
Last edited:
Status
Not open for further replies.
Top Bottom