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

Status
Not open for further replies.
Sony really needs to show off a really good tech demo of the I/O capabilities. Because that would clear up a ton of confusion over it.

I think this pretty clearly states devs just need to use the provided API? Hardly any effort required. https://www.eurogamer.net/articles/digitalfoundry-2020-playstation-5-the-mark-cerny-tech-deep-dive

Behind the scenes, the SSD's dedicated Kraken compression block, DMA controller, coherency engines and I/O co-processors ensure that developers can easily tap into the speed of the SSD without requiring bespoke code to get the best out of the solid-state solution. A significant silicon investment in the flash controller ensures top performance: the developer simply needs to use the new API. It's a great example of a piece of technology that should deliver instant benefits, and won't require extensive developer buy-in to utilise it.
 
They may well do. The one thing that Sony sells better than Microsoft, and it's quite true too; Is how they build the console tailor made and laser focused on devs. They nearly lost their empire with the PS3 difficulties and learnt the hard way.

I'm convinced the PS5 is the dev dream machine where Xbox runs second in that aspect but still clearly solid.

Sony also knows, that their dev love = dev support.


Ever since Mark Cerny designed the PS4 his focus was on making the hardware as accessible to developers as possible. I'm pretty sure the Series X is still a great console to make games for but the PS5 will be easier make games for.
 
You also quoted a custom console GPU that was even further ahead than that.

So PS4 was built with 8 ACEs which is what AMD did when they launched Hawaii. Hawaii and PS4 came out in the same time frame.

Ps5 and xbox are both using RDNA2 which is also likely to launch in and around the consoles towards the end of the year.

RDNA3 will be a late 2021 or early 2022 GPU. To think a console launching in 2020 will be using it is nonsense.
 
This is really interesting. She's not an engineer who'd be involved in details. If she mentions "special sauce" using those exact words, those features might be significant and I mean for AMD, not Sony.
Chill Spencer not of fan of special sauce. He knows something.
 
Not sure if you read what I quoted but he specifically mentioned the GPU Cache Scrubber modification as a potential future RDNA feature.

No he didn't; he implied it could be a future implementation within the PC GPU space. The specific term 'RDNA' never came out of his mouth in that segment.

Not really, I think that after saying something clearly and seeing plenty of people take it the worst possible confusing way he was a bit annoyed and tried to set expectations more directly. He talked about customisations you may see on an AMD discredete GPU around PS5's launch or after and asked people to keep in mind that, like for PS4, that would not mean those features were not developed by Sony or in collaboration with Sony. I think that was pretty clear. MS may have the same depending on how much they deviated: like with Xbox One X vs PS4 Pro I see Sony being the one focusing on trying to add longer term lower level small bits and bobs to allow devs to extract extra performance they are not providing with additional CU's (trying to do a bit more with less).

I can understand Cerny's frustration there but a lot of that also came down to Sony just being in a state of flux, with so many key positions being reshuffled, phased out, new hires etc. Arguably they are probably still trying to settle on some of that. And just my opinion, but Jim Ryan has the personality of a plain donut, so that obviously has a role to play given he's running the division nowadays.

That said I think you bring up a historical precedent that, for what seems like to some anyway, isn't probably the case now. When MS designed the XBO, they were not designing a gaming-first console. They made sure the system was capable enough (by their standards) to play then-next gen games, but the chief focus was on multimedia and Kinect. That informed the XBO's design to its very core.

But that isn't the case this time around. It's pretty clear they've gone for a device designed with gaming as its chief concern, so when I hear some people try insinuating their design philosophy is the same as with XBO's, or that they're "just building a gaming PC", that comes off as disingenuous. It's ignoring the fact that of course they'd want to ensure cross-compatibility with the PC market in terms of feature sets and hardware standards, since they have an actual stake in that market. It doesn't mean they simply took PC features and tried shoving them into a console-style box and went with some brute force approach. And it also ignores the fact that even in terms of there being cross-compatibility between the two platforms, there are still aspects of those customized only for XSX, like mesh shader level support, and aspects of BCPack.

Even regarding the expansion card storage, that isn't commonplace on PCs, not with that form factor. Maybe it will be some time in the future, or maybe it won't, who's to say. My point is, there's a prevailing thought that MS may or may not have done certain customizations due to what they did with the XBO and X (which, TBF, had to still play by XBO's rules in a lot of ways in terms of just how much things could be changed), when it's pretty clear MS is not being informed by that same design philosophy for the Xbox division anymore, and they're better off for it.

Wasn't CELL made by IBM though?

It was a menage-a-trois between IBM, Sony and Toshiba.
 
Last edited:
So PS4 was built with 8 ACEs which is what AMD did when they launched Hawaii. Hawaii and PS4 came out in the same time frame.

Ps5 and xbox are both using RDNA2 which is also likely to launch in and around the consoles towards the end of the year.

RDNA3 will be a late 2021 or early 2022 GPU. To think a console launching in 2020 will be using it is nonsense.

The PS4 Pro had vega features too and that didn't arrive until nearly a year later
 
Well, IF PS5 is using any RDNA3 tech it would be advantageous to use any feature that gets the heat or wattage lower.
Has anyone even heard what potential features they could be siphoning off the RDNA3 tech?
 
Your right about that. Somebody mentioned earlier that she worked on RSX which is why I was confused.

"Also through the division, she represented IBM in a collaboration to create next-generation chips with Sony and Toshiba. Ken Kutaragi charged the collaboration with "improving the performance of game machine processors by a factor of 1,000", and Su's team eventually came up with the idea for a nine-processor chip, which later became the Cell microprocessor used to power devices such as the Sony PlayStation 3. As of 2006, she continued to serve as vice president of the semiconductor research and development center at IBM, holding the role until May 2007."
 
That said I think you bring up a historical precedent that, for what seems like to some anyway, isn't probably the case now. When MS designed the XBO, they were not designing a gaming-first console.
I specifically mentioned Xbox One X and PS4 Pro specifically.

Even back then you could see a machine with a focus on much higher raw performance vs one with more custom features trying to punch above its weight a bit more (Sony generally encourages lower level HW access and was hoping more developers took advantage of those extra features). This time it changed as you have a smaller performance delta and the consoles launching at around the same time.

customized only for XSX, like mesh shader level support, and aspects of BCPack
BCPack is more of a customisation on the I/O pipeline than the GPU itself and mesh shaders is a feature supported by both consoles (it is possible Sony has a more advanced geometry processing stage with the Geometry Engine while MS has a more advanced VRS implementation though).

there's a prevailing thought that MS may or may not have done certain customizations due to what they did with the XBO and X (which, TBF, had to still play by XBO's rules in a lot of ways in terms of just how much things could be changed
Xbox One X change everything compared to the original Xbox One: more custom CPU, totally different GPU generation, totally different memory setup, etc... not sure if the argument they were restricted is that strong in this case. There is historic precedent and current data known that seems to show a different pattern in what each console architect thinks developers want in an easy to develop for and powerful architecture. These are the bets that will make third party software great on both, but first party ones really really generational leaps even in the first year on the market I believe.
 
Last edited:
I don't get why people are so irritated about that topic. Both MS and Sony worked together with amd to build these consoles. Both console manufacturers have their own ideas and resources to co-design certain cpu/gpu/apu/overall hardware features.

AMD had the combined resources of all 3 companys at their disposal and might therefore had new information, ideas and concepts for future features that might be relevant for their line of products.

Both MS and Sony probably paid some money and signed some hefty ndas to get informations about AMDs R&D anf future specs and features, and might have again paid to use some of those in their next gen consoles which will release before AMD finishes their RDNA3 designs.
AMD profits hugely from this cooperation in terms of money and knowledge. They are already investing even more money into R&D.

Basically MS and Sony are accelerating AMDs progress.

Noone is saying MS doesn't have RDNA3 features. However we don't know which possible RDNA3 features which console has at this point.
However we will learn about those eventually.
Just imagine this situation the console maker (MIcrosoft or Xbox) were working with AMD in a specific feature, but this will be not ready for RDNA for
some requirement or still need more tweaks then the console maker decide to continue with that implementation finish themselves but that also
could means:

-That specific feature is not necessary a feature for RDNA 3 maybe yes the feature is working good enough in the console can help the next gen in AMD maybe no because they follow another path
-If some console maker has this is probably they cannot tell us for two reason 1)the other console maker doesn't has it 2)AMD request to give any detail too soon to NVDIA about a product
which is probably ready until the next year.
-The feature is not the not necessarily the second coming of Jesus
-That feature could be a sister from another, this can have a similar purpose but implemented slightly differently or be something different
-That happened before

Is not impossible to have that a new feature but at we don't have any info just don't theorize about this we need more info is just not worth it.
 
Last edited:
Without specifying names. It is evidence that PS5 has features that are not (nor will be) in XSX (some have already been said, others not yet), for the simple fact that they are features that Sony and AMD have developed jointly. Technology that is already known to be in the next AMD processors and PS5 (exclusively). Anything else that has not been developed in conjunction with Sony can be found on XSX. The rest of the things are not.
 
Without specifying names. It is evidence that PS5 has features that are not (nor will be) in XSX (some have already been said, others not yet), for the simple fact that they are features that Sony and AMD have developed jointly. Technology that is already known to be in the next AMD processors and PS5 (exclusively). Anything else that has not been developed in conjunction with Sony can be found on XSX. The rest of the things are not.
How are you feeling chief?
 
Without specifying names. It is evidence that PS5 has features that are not (nor will be) in XSX (some have already been said, others not yet), for the simple fact that they are features that Sony and AMD have developed jointly. Technology that is already known to be in the next AMD processors and PS5 (exclusively). Anything else that has not been developed in conjunction with Sony can be found on XSX. The rest of the things are not.
You're not going to make friends in the X-Box forums, that's for sure! :messenger_beaming:
 
Without specifying names. It is evidence that PS5 has features that are not (nor will be) in XSX (some have already been said, others not yet), for the simple fact that they are features that Sony and AMD have developed jointly. Technology that is already known to be in the next AMD processors and PS5 (exclusively). Anything else that has not been developed in conjunction with Sony can be found on XSX. The rest of the things are not.

You can say names, we ready know about "tempest engine"

And why do you speak like you are assuming these details?

I thought you were a dev that ready has seen/hands on with the ps5 dev kits and a couple months ago were assuring everyone that the ps5 was more powerful tf-wise?
 
Without specifying names. It is evidence that PS5 has features that are not (nor will be) in XSX (some have already been said, others not yet), for the simple fact that they are features that Sony and AMD have developed jointly. Technology that is already known to be in the next AMD processors and PS5 (exclusively). Anything else that has not been developed in conjunction with Sony can be found on XSX. The rest of the things are not.
This was expected from the Cerny speak. Happened as well with the hardware accelerated checkerboard and ID per object in the PSPro that were features that didn't make it to the later Scorpio.
 
Anyone expecting Rockstar to show up at the Inside Xbox event next week? GTAV series X gameplay? Surprise GTAVI trailer?
This is one of the few companies which doesn't need Xbox, Sony,Nintendo to make a big announcement so I don't think so is more
probably will be in the show of Geoff Keighley but I think is too soon.
 
You can say names, we ready know about "tempest engine"

And why do you speak like you are assuming these details?

I thought you were a dev that ready has seen/hands on with the ps5 dev kits and a couple months ago were assuring everyone that the ps5 was more powerful tf-wise?

He said Ps5 performed well, he did not say how many TF, there is a difference which people cannot grasp, they just want 1 simple number.

Now the debate is how useful is the logic or function.

Reading up on VRS, amd the DF video saying you can get 10 % extra performance but there are still artefacts - so no difference to checkerboard rendering, which lets face it VRS is a checkerboard type solution in smaller areas - but I expect both to have it and its usefulness will be with large blank areas, like sky in racing games etc, or VR for outside of your focus.

VRS was started by Nvidia, and its hardly a secret sauce.

My guess is something to do with geometry engine...guess.
 
Last edited:
You can say names, we ready know about "tempest engine"

And why do you speak like you are assuming these details?

I thought you were a dev that ready has seen/hands on with the ps5 dev kits and a couple months ago were assuring everyone that the ps5 was more powerful tf-wise?
that's how fud is created. did you consulted your discord xbox chat before writing this? or you are coached in advance?
 
The one culling example you gave is called z-cull, in traditional pipeline (without primitive shader) it doesn't happen until the very last stage of the pipeline. Culling performed before rasterization are clip space culling, back-face culling and small triangles culling, none are particularly computationally expensive.

And load/store instructions on GPUs (and any modern processor for that matter) are not asynchronous, so you cannot go ahead while waiting for memory.
I was talking about triangle culling in mesh shaders. If a mesh is only partially clipped by a plane, whether it's worth it to find out exactly which triangles in the mesh are clipped depends on how large your triagles are and how much compute time you can afford. With the PS5's higher compute headroom it should be able to do the calculations for smaller triangles and less clipped meshes. I also assume it should be able to test for more obstructions in the geometry phase before it even gets to the z-buffer.

Here is a Microsoft video discussing the concept. The part I'm refering to starts at 22:22, but earlier in the video it mentions that if a mesh is only clipped by 5-10%, it's not worth it to do per triangle testing. However if your GPU is clocked faster like the PS5's, then those should be free tests hidden by memory access delays.
 
Without specifying names. It is evidence that PS5 has features that are not (nor will be) in XSX (some have already been said, others not yet), for the simple fact that they are features that Sony and AMD have developed jointly. Technology that is already known to be in the next AMD processors and PS5 (exclusively). Anything else that has not been developed in conjunction with Sony can be found on XSX. The rest of the things are not.
Doesn't the opposite apply too, MS/AMD co-developed tech only present in SX?
 
then i guess cerny is wrong and you are right. can you turn RT from hw solution to software? or maybe whole GPU to just software code?
No. The RDNA 2 GPU runs the geometry processing code from the developer. How that code gets to the hardware is through the API. On the PS5 that API/architecture is called the Geometry Engine. On the XSX it's called DirectX Ultimate. Since both platforms are using the same base RDNA 2 GPU to do the actual geometry processing, they both have roughly the same capability.

Like I previously pointed out, either platform could have tweaked the hardware for their own needs, like the PS5 did with it's better cache management, but the base functionality is the same. The cache is still doing cache things. As Cerny said, if one of the platforms did come up with a new feature that was generally applicable, it would have become a standard part of RDNA 2 and everyone would have access to it.
 
Can someone elighten me on the difference between VRS and DLSS? Is one basically rendering what needs to be rendered in the vision of the player to reduce draw calls while the other is AI filling in the image?
 
Last edited:
And load/store instructions on GPUs (and any modern processor for that matter) are not asynchronous, so you cannot go ahead while waiting for memory

Firing off loads, or prefetching memory (or in some architectures running ahead just to warm up the predictors and loca caches, see Niagara's speculative threads approach), and scheduling non data dependent instructions is quite common place in any out of order front CPU. GPU wise is also pretty common to execute non data dependent math heavy workload waiting for data to be loaded in the chip (it is common practice to balance texture/memory loads instructions with compute work and or run shaders that have data ready to go... GPU's are latency tolerant throughout machines).
 
No. The RDNA 2 GPU runs the geometry processing code from the developer. How that code gets to the hardware is through the API. On the PS5 that API/architecture is called the Geometry Engine. On the XSX it's called DirectX Ultimate. Since both platforms are using the same base RDNA 2 GPU to do the actual geometry processing, they both have roughly the same capability.

Like I previously pointed out, either platform could have tweaked the hardware for their own needs, like the PS5 did with it's better cache management, but the base functionality is the same. The cache is still doing cache things. As Cerny said, if one of the platforms did come up with a new feature that was generally applicable, it would have become a standard part of RDNA 2 and everyone would have access to it.

Cerny said that the desktop part would have access to it, not necessarily XSX. Same thing for the customisations MS did on top of the base RDNA2 specs.
 
Doesn't the opposite apply too, MS/AMD co-developed tech only present in SX?

Of course, but MS being MS they give it nice sounding names like velocity architechure and mention it 1000 times, so I am confortable that we would know by now.

Even Nvidia ones that made it to RDNA2 like VRS MS talk about it and DX12 implementation......so it is what it is...

Cerny said there is 2 and nobody has mentioned them, maybe there big or small...good or worthless....who knows..
 
Last edited:
Of course, but MS being MS they give it nice sounding names like velocity architechure and mention it 1000 times, so I am confortable that we would know by now.

Even Nvidia ones that made it to RDNA2 MS talk about...so...

Cerny said there is 2 and nobody has mentioned them, myeb there nig or small...who knows..

Is the hype around the velocity architecture deserved?

I'm seeing plenty of speculation over it and it appears that some people are over exaggerating it's capabilities.
 
Doesn't the opposite apply too, MS/AMD co-developed tech only present in SX?
Microsoft is a company that develops "software". I don't see Microsoft collaborating with AMD to develop Hardware, much less the software for that new Hardware. If you have not even had the decency to bother to properly customize DX12U for XSX and have opted to use the generic version designed for PC. That does not mean that it is bad, but it certainly does not look good for a software company to give the impression that it has not made much effort to create decent tools that really measure up to the powerful machine they have bought. That will make that power not squeeze all its performance easily, and less from the beginning. That is the only weak point that I see Microsoft. Power without sufficient short-term control.

You can say names, we ready know about "tempest engine"

And why do you speak like you are assuming these details?

I thought you were a dev that ready has seen/hands on with the ps5 dev kits and a couple months ago were assuring everyone that the ps5 was more powerful tf-wise?
Look for the message in which I said that the PS5 GPU was more powerful in theoretical numbers than the XSX. It is not me who is stubborn in trying to compare those numbers. I have always said that PS5 is more efficient. I thought about it in January, and at this point I still think that it will continue to be so when consoles are on the market.

You're not going to make friends in the X-Box forums, that's for sure! :messenger_beaming:
Fortunately, I don't pay any, I say what I see, if you don't like what I say, I'll go to sleep just as calmly. I think self-deception is for the weak. Although it is true that if Microsoft had continued betting on VR perhaps it could say more positive things about it.

How are you feeling chief?
Perfectly for ... 2 weeks already? (I do not remember). Back and working hard. Right now I am writing this message while I render some renders. And right now I'm going to finish watching "The Empire of the Sun".

Cheers.
 
Microsoft is a company that develops "software". I don't see Microsoft collaborating with AMD to develop Hardware, much less the software for that new Hardware.

Do you have any proof of that? From all that I've read it seems like they worked with AMD to design hardware but maybe I'm missing something.
 
Can someone elighten me on the difference between VRS and DLSS? Is one basically rendering what needs to be rendered in the vision of the player to reduce draw calls while the other is AI filling in the image?
VRS = Rendering parts of the image that are less important (such as the sky) for example because said parts are quite uniform at a lower resolution to save GPU resources.
DLSS = Literally AI upscaling that shits all over primitive upscaling. Seems like magic to some, because it kinda is lol.
 
VRS = Rendering parts of the image that are less important (such as the sky) for example because said parts are quite uniform at a lower resolution to save GPU resources.
DLSS = Literally AI upscaling that shits all over primitive upscaling. Seems like magic to some, because it kinda is lol.

DLSS blew my mind in Control. I was able to get 60 fps with full RTX on my 2060 with a 2600x.

Couldn't believe it.
 
Microsoft is a company that develops "software". I don't see Microsoft collaborating with AMD to develop Hardware, much less the software for that new Hardware. If you have not even had the decency to bother to properly customize DX12U for XSX and have opted to use the generic version designed for PC. That does not mean that it is bad, but it certainly does not look good for a software company to give the impression that it has not made much effort to create decent tools that really measure up to the powerful machine they have bought. That will make that power not squeeze all its performance easily, and less from the beginning. That is the only weak point that I see Microsoft. Power without sufficient short-term control.


Look for the message in which I said that the PS5 GPU was more powerful in theoretical numbers than the XSX. It is not me who is stubborn in trying to compare those numbers. I have always said that PS5 is more efficient. I thought about it in January, and at this point I still think that it will continue to be so when consoles are on the market.


Fortunately, I don't pay any, I say what I see, if you don't like what I say, I'll go to sleep just as calmly. I think self-deception is for the weak. Although it is true that if Microsoft had continued betting on VR perhaps it could say more positive things about it.


Perfectly for ... 2 weeks already? (I do not remember). Back and working hard. Right now I am writing this message while I render some renders. And right now I'm going to finish watching "The Empire of the Sun".

Cheers.
Damn ur comments sounds like the crytek dev where he said xsx api is not user friendly or as easy as ps5 and dared that u wont find a single dev who can say xsx api is better than ps5 api. I remeber he also mentioned xsx api is rather "old" or something like that . Like they are reusing the dx12 which is rather old and not tailored to the hardware
 
Last edited:
No he didn't; he implied it could be a future implementation within the PC GPU space. The specific term 'RDNA' never came out of his mouth in that segment.

Future PC GPU = RDNA3+.

Furthermore with the final part I quoted he alludes that the changes made on PS5 has created even further efficiencies along both the GPU and CU than what is standard for RDNA2.

Continuous improvement in AMD technology means it's dangerous to rely on teraflops as an absolute indicator of performance and CU count should be avoided as well.

Continuous improvement in AMD = jumps from GCN to RDNA1 or RDNA1 -> RDNA 2 etc.
 
Last edited:

2 advanced GPU "features"

Mark Cerny said:

How it works is that we sit down with AMD, who are terribly collaborative. It's a real pleasure to work with them. So basically, we go ahead and say how many CUs we want to have and we look at the roadmap features and we look at area and we make some decisions and we even - in this case - have the opportunity, from time to time, to have a feature in our chip before it's in a discrete GPU. We have two of these this time, which is very nice.
 
Last edited:
Status
Not open for further replies.
Top Bottom