Next Generation Hardware Speculation with a Technical Spin [post E3 2019, pre GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
If PS5 GPU was RDNA 2 based, why would a Sony PS5 software engineer be asking "rhetorical questions" about possible RT related issues right after Xbox announcement of having RDNA 2 based GPU? Doesn't make any sense to me. Why would he ask very specific questions about RT related methods relating to RDNA 2, if PS5 has such features? Either Sony engineers have no clue how to resolve these specific issues, or PS5 GPU isn't RDNA 2, but something else.

Also, in some of his following replies he hinted towards why he was asking such questions.

Alternatively if they know what RT in RDNA 2 entails, they could be asking those questions from a POV of how DXR might be more limited than the actual hardware implementation of RT in RDNA 2.

The tweet is about as vague as MS saying RT in XBSX is based on DXR. The current version of DXR or a future version of DXR? Just like DX changes over time, DXR is also going to change over time. If we go back to when DX and later D3D were introduced, changes to DX and D3D came fast and furious. I see no logical reason that DXR won't also see similarly rapid revisions as hardware accelerated RT is in its infancy just as consumer level hardware accelerated 3D was still in its infancy when MS attempted to provide cross hardware compatibility through D3D (versus all the vendor specific APIs suitable for games that were available at the time).

At which point D3D and hardware accelerated 3D hardware evolved mostly in step with each other with MS (and software developers through MS) influencing hardware design while hardware design influenced D3D. RT hardware acceleration and DXR will likely follow a similar trajectory.

However until that happened there were sometimes gaps in what X hardware could do and what the API allowed access to. Alternatively there were features that the API allowed access to that X hardware couldn't do. It's just as likely that Sony engineer could have been referencing something along these lines as it is that it was a purely rhetorical question or that it could be a non AMD RT solution.

Unless or until mobile embraces RT, PC hardware GPU makers will likely dictate the future of RT hardware acceleration. While being influenced by...say one console...could net you around a 100 million user install base, that also locks the hardware features to once every 5-8 years. Versus PC where you'll have user (hardware) penetration in the hundreds of millions (gamers, businesses, research institutions, etc.) with your hardware being capable of evolving at a much faster pace.

Regards,
SB
 
Last edited:
Follow up tweet....

He is specifically addressing Microsoft with that first tweet. Not some random / drive-by questions directed towards anyone else.

Again still vague. Could be talking about how DXR is more limited than the RT in RDNA 2, could be purely rhetorical, or could be hinting at non AMD RT hardware.

[edit] One more thing it could be hinting at. Perhaps the problems he listed are problems Sony also has and he's poking to see if Microsoft may have found some way to solved them if they are using the same hardware.

Regards,
SB
 
Again still vague. Could be talking about how DXR is more limited than the RT in RDNA 2, could be purely rhetorical, or could be hinting at non AMD RT hardware.

Not necessarily disagreeing either. But these questions could have been asked days, weeks or months prior to yesterday's Xbox RDNA 2 announcement. Better yet, they have partners (AMD) and other capable internal teams (ICE, etc.) for such needs.

Perhaps the problems he listed are problems Sony also has and he's poking to see if Microsoft may have found some way to solved them if they are using the same hardware.

Regards,
SB

This could be a possibility. But I can't picture Sony's software engineering teams having such issues when they have hired such individuals (years ago) whom specifically specialize in RT software/hardware.
 
This could be a possibility. But I can't picture Sony's software engineering teams having such issues when they have hired such individuals (years ago) whom specifically specialize in RT software/hardware.

Which also applies to MS. It's not like DXR and either separately or relatedly RT for XBSX were something that MS just accidentally stumbled on or failed to do research on or didn't have anyone in their vast R&D teams with RT experience.

In general MS doesn't talk much about the myriad things they do research on in their R&D teams. Things have changed a little with Nadella OKing potential research projects for potential commercial deployment and thus public disclosure slightly more frequently, but the vast majority of research that goes on at MS isn't heard of, outside of MS.

Hell, they were already doing RT research as a solution to problems back in the 80's. :p

https://www.microsoft.com/en-us/res...plex-models-containing-surface-tessellations/

RT and the related path tracing applied to various fields (Kinect, for example) have featured in many of their research papers over the years, through the 80's, 90's, 2000's, and 2010's. And that's only what their researchers have been allowed to publicly reveal.

Regards,
SB
 
Last edited:
Which also applies to MS. It's not like DXR and either separately or relatedly RT for XBSX were something that MS just accidentally stumbled on or failed to do research on or didn't have anyone in their vast R&D teams with RT experience.

Regards,
SB

Hold your horses, this isn't a MS vs. Sony fanboy thing. I'm just posing other possibilities (like others) about his tweet. Nothing more, nothing less.

And yes, I'm fully aware of Microsoft's capabilities when it comes to R&D. :yep2:
 
Hold your horses, this isn't a MS vs. Sony fanboy thing. I'm just posing other possibilities (like others) about his tweet. Nothing more, nothing less.

And yes, I'm fully aware of Microsoft's capabilities when it comes to R&D. :yep2:

Yes, many companies have done research into RT. Heck, just look at Sony and that one demo on the PS2. I just took umbrage at...

But I can't picture Sony's software engineering teams having such issues when they have hired such individuals (years ago) whom specifically specialize in RT software/hardware.

The implication being that Microsoft's (or AMDs, or NVs, or any other team doing RT research) engineers would have such issues but Sony's wouldn't? Basically, it comes across as, "Sony can't have these problems because they hired people in the RT field, but other companies who employ people in the RT field can have those issues."

I'm not saying that the Sony engineers haven't solved a problem that other engineers haven't solved. But it also doesn't mean that Sony's (or anyone elses) engineers have solved the problem. And it certainly shouldn't be implied that only Sony's engineers are capable of solving the problem.

Regards,
SB
 
Last edited:
The implication being that Microsoft's (or AMDs, or NVs, or any other team doing RT research) engineers would have such issues but Sony's wouldn't? Basically, it comes across as, "Sony can't have these problems because they hired people, but other companies can have those issues."

Regards,
SB

That wasn't what I was implying. I was implying that I can't picture Sony software engineers asking such questions "openly" when they have such resources as AMD or ICE on resolving such problems (inclusive of specialist). This wasn't implying anything negative against Microsoft. So...
 
That wasn't what I was implying. I was implying that I can't picture Sony software engineers asking such questions "openly" when they have such resources such as AMD and ICE on resolving such problems. This wasn't implying anything negative against Microsoft. So...

Gotcha. And I didn't take as something negative against MS specifically, but implying things about anyone that isn't Sony. There are many capable RT researchers out there not employed by MS or Sony as well as ones that are employed by MS and Sony.

Regards,
SB
 
The implication being that Microsoft's (or AMDs, or NVs, or any other team doing RT research) engineers would have such issues but Sony's wouldn't?
As I understand it, regardless what R&D Ms has done in the past (they have their own patents on things like photon mapping), MS is using AMD's hardware and so will have to solve problems with their software stack on AMD hardware. The theoretical possibility here is Sony have their own hardware, so can use hardware solutions to these problem. Hence it has very little to with research and ability, instead being about choices how to implement HW and whether to go with their own R&D or AMD's.
 
As I understand it, regardless what R&D Ms has done in the past (they have their own patents on things like photon mapping), MS is using AMD's hardware and so will have to solve problems with their software stack on AMD hardware. The theoretical possibility here is Sony have their own hardware, so can use hardware solutions to these problem. Hence it has very little to with research and ability, instead being about choices how to implement HW and whether to go with their own R&D or AMD's.

Yes, but all this is related to that one tweet by the Sony engineer. So...
  • Is the limitation in RDNA 2?
    • Possible implication that Sony aren't using the RT block from RDNA 2.
    • Possible that the tweet is just rhetorical.
    • Possible that the engineer is just poking to see if MS have solved the problem.
  • Is the limitation in the current implementation of DXR?
    • Possible implication that Sony is using the RT block of RDNA 2 but without the limitations of the current implementation of DXR.
    • Could still possibly imply that Sony isn't using the RT block of RDNA 2.
So, yes, it's one of a myriad of possibilities. That was never in question.

But I can't picture Sony's software engineering teams having such issues when they have hired such individuals (years ago) whom specifically specialize in RT software/hardware.

That was what rubbed me the wrong way. Microsoft or AMD (or anyone else) could have those problems but Sony can't possibly have those problems? But the tweet doesn't say that MS or AMD have those problems. It's questioning whether it does or not.

However, it's all been cleared up. :) He didn't mean it the way it came across when I read it.

Regards,
SB
 
As I understand it, regardless what R&D Ms has done in the past (they have their own patents on things like photon mapping), MS is using AMD's hardware and so will have to solve problems with their software stack on AMD hardware. The theoretical possibility here is Sony have their own hardware, so can use hardware solutions to these problem. Hence it has very little to with research and ability, instead being about choices how to implement HW and whether to go with their own R&D or AMD's.

Phil Spencer has used the language "co-engineered" on more than one occasion when referring to the silicon in the XSX. I believe that makes it likely that DXR (or at least MS' learnings on RT) have informed the hardware design of RT in XSX and more broadly, RDNA 2. The way I interpret the tweet in this context, is that Sony veered off on their own between RDNA 1 and RDNA 2 with regards to a solution for RT. When RDNA 2 was announced yesterday, the Sony engineer was curious how the combination of hardware and software arrived upon by AMD\MS in RDNA 2 would address what they (Sony) viewed as the primary obstacles to RT performance and quality.
 
Again still vague. Could be talking about how DXR is more limited than the RT in RDNA 2, could be purely rhetorical, or could be hinting at non AMD RT hardware.

[edit] One more thing it could be hinting at. Perhaps the problems he listed are problems Sony also has and he's poking to see if Microsoft may have found some way to solved them if they are using the same hardware.

Regards,
SB
Could also be that RDNA2 RT is limited and so will have issues with meeting the requirements of both current DXR and future/evolving requirements. There I think we've run the gamut.

The questions to what his underlying meaning is depends on whether he/Sony knows the answer to the questions he asks or not. Yes (and it was easy or and it was fk'n hard) or No (I wonder if they solve what we couldn't).

My personal opinion is that he is well aware of the strengths and weaknesses or Sony's RT solution (obviously) and of RDNA's (which I'm sure has been evaluated more than once and in more than one state of development) and that the answers to the questions he's asking led to the decisions that were made by Sony.

I think it also means we are reaching the end to how long the knowledgeable people can be kept quiet. Its why GDC is going to be a sieve of leaks if Sony tries to maintain its silence.
 
Honestly, the mind reading that's going on about this guys tweets deserves to be in the baseless thread. Does anyone know what he actually does now, or how long ago he left PS5? There are a half dozen ways you could interpret the tweets and any of them could be true.

He left not long ago because his linkedin account show he worked until february 2020 on PS5.

After I doubt his tweet has nothing to do with the PS5 because he is not rendering software programmer but it is probably something he heard by people working at Sony HQ like him.
 
Last edited:
  • Is the limitation in the current implementation of DXR?
    • Possible implication that Sony is using the RT block of RDNA 2 but without the limitations of the current implementation of DXR.
    • Could still possibly imply that Sony isn't using the RT block of RDNA 2.
Regardless of anything - I do not think there is some magical solving of transparency or multiple (real) bounces in for RT without better hardware. I do not think the bottleneck is the currently available API, rather hardware and softare development ontop of that API. It is still "stunde null" for RTRT.

The questions about transparency or multibounce are conondrums for RT performance with or without realtime HWRT being considered. Stuff is just generally... expensive.
 
Regardless of anything - I do not think there is some magical solving of transparency or multiple (real) bounces in for RT without better hardware. I do not think the bottleneck is the currently available API, rather hardware and softare development ontop of that API. It is still "stunde null" for RTRT.

He worked on the OS side of the PS5 with C#, React, javascript for the UI and tools of the PS5. I don't believe, he will asked this questions if it was some random thought about rendering. This is very precise for a non rendering dev. For example the viewport and frustrum question, he speaks about hardware having at least tiers 2 VRS.

What is sure it is not something he knows because of is his job but probably talking with people on the rendering teams in studios or at Sony headquarter.

His linkedin account:
https://www.linkedin.com/in/matthargett/
 
Last edited:
https://community.amd.com/community/gaming/blog/2019/06/09/amd-powers-microsoft-project-scarlett

Last year AMD already announced "next-generation Radeon™ RDNA gaming architecture ". Some people argue that it doesn't mean "next-generation of RDNA". But since AMD''s words are straightforward we don't need to be so surprised to see RDNA2 announcement.

Besides AMD already admitted xbox using their RT solution early but they never mentioned PS5. This is why so many speculation about different solution of RT in PS5.
 
https://community.amd.com/community/gaming/blog/2019/06/09/amd-powers-microsoft-project-scarlett

Last year AMD already announced "next-generation Radeon™ RDNA gaming architecture ". Some people argue that it doesn't mean "next-generation of RDNA". But since AMD''s words are straightforward we don't need to be so surprised to see RDNA2 announcement.

Besides AMD already admitted xbox using their RT solution early but they never mentioned PS5. This is why so many speculation about different solution of RT in PS5.
Yeah, they don't disclose technology that the partner (or themselves) didn't want disclosed yet. "based on RDNA" or "next gen RDNA" was the most they could say until RDNA 2 was officially unveiled.

Anything custom cannot be disclosed by AMD first, so it's a reasonable angle to argue it might not be amd's implementation, but it could also be just a slightly modified version of amd's implementation.
 
Status
Not open for further replies.
Back
Top