AMD Execution Thread [2023]

Status
Not open for further replies.
Why is they talking about ML instructions after mentioning AL? The two has nothing in common.

I'd expect AL to work similarly to Reflex in that there's an app side and the driver side, and for optimal performance both must be present. This means that FSR3's latency will likely be worse on non-AMD h/w where there won't be the driver side of the AL part.
 
The 7800xt offers basically the same experience as the 6800xt which has been available at this price for over a year. There is no market for this card that wouldn’t have already bought a 6800xt. It’s hardly more power efficient at a meager 10% reduction.
There are people that will be more attracted to the 'new' thing rather than the discounted old thing.

And I'll bet that there will be scope for optimizing power efficiency to a level that the 6800XT could not do with the same performance.

In general, I dont like the whole 'but the new thing costs the same as discounted old part' argument to cry about pricing, cuz usually those older parts are only discounted cuz they're desperately trying to get rid of stock and soon will become unavailable. It's not typically a fair comparison point. But I'd agree here that these RDNA2 parts being available and discounted for so long makes things feel quite different.

Though I'd say that even if they hadn't been available for a long time, getting the performance of a $650 GPU for $500 nearly three years later is just lousy however you want to cut it. Just generally, these minimal improvements in performance per dollar are just hurting the excitement for basically all new GPU's.
 
And I'll bet that there will be scope for optimizing power efficiency to a level that the 6800XT could not do with the same performance.
Yes, beside double issue, RDNA3 has doubled register file and LDS size, iirc. That's raising a standard which was set with PS4 and has not changed since that (idk about NV).
That's a big optimization potential - bad occupancy due to register pressure and being tight on LDS should be gone, so we can crank things up.
I can increase the quality of my GI for example, something that just more cores does not really allow. I need more LDS.

That's why i'm pretty interested in RDNA3. It's much more of a game changer than many notice.
Though I'd say that even if they hadn't been available for a long time, getting the performance of a $650 GPU for $500 nearly three years later is just lousy however you want to cut it. Just generally, these minimal improvements in performance per dollar are just hurting the excitement for basically all new GPU's.
And that's why i'm not, and why i think people will loose excitement about the former status symbol of gaming. GPUs is an increasing source of frustration, so rejecting the race on higher fps and res in favor of efficient / low cost iGPU should make sense for an increasing number of people.
 
getting the performance of a $650 GPU for $500 nearly three years later is just lousy however you want to cut it.
It's no secret who we should thank for the surge in GPU prices. As for the performance, it's all about clocks and the overall number of CUs - so when AMD stays on the same or very similar process node, there is no additional performance to extract from higher transistor density and lower power draw.

The same situation was for GCN1/2/3 parts which stayed on TSMC '28 nm' for 3 years - and some substantial performance jumps only materialized when AMD moved to GloFo/Samsung 14LP/12LP with GCN4/5, then to TSMC N7/N6/N5 with RDNA1/2/3.

So whatever parts they would release on TSMC N3/N2, they will be more attractive even if they used the same RDNA3 architecture as today.

There is no market for this card that wouldn’t have already bought a 6800xt.
I've found you a market that's potentially 20 times the size of current RX 6800 XT owner share, 5.5% vs 0.3%, just by looking at the latest Steam Hardware Survey results for July 2023:

ALL VIDEO CARDS - Steam Hardware Survey July 2023
AMD Radeon RX 580​
1,07%​
AMD Radeon RX 570​
0,80%​
AMD Radeon RX 5700 XT​
0,68%​
AMD Radeon RX 6600​
0,56%​
AMD Radeon RX 6600 XT​
0,41%​
AMD Radeon RX 550​
0,38%​
AMD Radeon RX 580 2048SP​
0,36%​
AMD Radeon RX 5600 XT​
0,25%​
AMD Radeon RX 5500 XT​
0,22%​
AMD Radeon RX 6500 XT​
0,21%​
AMD Radeon RX 6650 XT​
0,21%​
AMD Radeon RX 560​
0,19%​
AMD Radeon RX 480​
0,16%​
Potential incentive to upgrade to Radeon RX 7800 XT
5,5%
AMD Radeon RX 6700 XT​
0,63%​
AMD Radeon RX 6750 XT​
0,21%​
AMD Radeon RX 6800 XT​
0,31%​
AMD Radeon RX 6900 XT​
0,25%​
AMD Radeon RX 7900 XTX​
0,17%​
No incentive to upgrade to Radeon RX 7800 XT
1,57%

Nvidia GTX 7/9/10/16 cards make 29,65%, RTX cards make 31,63%, the rest are notebook RTX cards at 8,36% and MX/GTX-M cards at 1,79%, integrated Intel/AMD processor graphics at 11,46%, and 'other' (unrecognized) at 10,01%. AMD total is 7,07% and Nvidia total is 71,43%.
 

Attachments

  • SteamSurvey201307.zip
    22.5 KB · Views: 0
Last edited:
I've found you a market that's potentially 10 times the size of current RX 6800 XT owner share, 6% vs 0.3%, just by looking at the latest Steam Hardware Survey results for July 2023:


ALL VIDEO CARDS - Steam Hardware Survey July 2023
AMD Radeon RX 580​
1,07%​
AMD Radeon RX 570​
0,80%​
AMD Radeon RX 5700 XT​
0,68%​
AMD Radeon RX 6700 XT​
0,63%​
AMD Radeon RX 6600​
0,56%​
AMD Radeon RX 6600 XT​
0,41%​
AMD Radeon RX 550​
0,38%​
AMD Radeon RX 580 2048SP​
0,36%​
AMD Radeon RX 5600 XT​
0,25%​
AMD Radeon RX 5500 XT​
0,22%​
AMD Radeon RX 6500 XT​
0,21%​
AMD Radeon RX 6650 XT​
0,21%​
AMD Radeon RX 560​
0,19%​
AMD Radeon RX 480​
0,16%​
Potential incentive to upgrade to Radeon RX 7800 XT
6,13%
AMD Radeon RX 6750 XT​
0,21%​
AMD Radeon RX 6800 XT​
0,31%​
AMD Radeon RX 6900 XT​
0,25%​
AMD Radeon RX 7900 XTX​
0,17%​
No incentive to upgrade to Radeon RX 7800 XT
0,94%

Nvidia GTX 7/9/10/16 cards make 29,65% and Nvidia RTX cards make 31,63%, the rest are notebook Nvidia RTX cards at 8,36% and MX/GTX-M at 1,79%, integrated processor graphics at 11,46%, and 'other' (unrecognized) at 10,01%.
If that 6% was interested in this performance level for 500$ why wouldn't they have already bought a 6800xt over the last year+?
 
If that 6% was interested in this performance level for 500$ why wouldn't they have already bought a 6800xt over the last year+?
Because they didn't want to buy an older GPU and were specifically waiting for new GPU's? Not like people knew for sure what $500 was going to get them a year ahead of time.

I also dont know why we're assuming only AMD owners would buy these new GPU's.
 
why wouldn't they have already bought a 6800xt

Good news, I've just received a call from my friend Group Captain The Lord Obvious - he just polled them all by phone, and it looks like they were frantically waiting for AMD to release the Radeon RX 7800 XT :ROFLMAO:
 
Because they didn't want to buy an older GPU and were specifically waiting for new GPU's? Not like people knew for sure what $500 was going to get them a year ahead of time.

I also dont know why we're assuming only AMD owners would buy these new GPU's.
They would have known when the 7900 series launched 10 months ago at the very latest.
 
If that 6% was interested in this performance level for 500$ why wouldn't they have already bought a 6800xt over the last year+?
Your question is irrelevant, we have no way to predict the buying behaviour of the 6% with this little information
 
AMDs Frank Azorhas gone on quote saying there's nothing stopping Starfield from adding DLSS, it's up to Bethesda if they want to put it in, despite few people claiming for a fact AMD is banning them from doing it because they hadn't specificly denied they're not. Azor also added that when ever any dev they support ask if they can integrate DLSS, AMD says yes. They technically don't even require prioritizing AMD techs - it's assumed of course, they request it, but don't demand it.

 
AMDs Frank Azorhas gone on quote saying there's nothing stopping Starfield from adding DLSS, it's up to Bethesda if they want to put it in, despite few people claiming for a fact AMD is banning them from doing it because they hadn't specificly denied they're not. Azor also added that when ever any dev they support ask if they can integrate DLSS, AMD says yes. They technically don't even require prioritizing AMD techs - it's assumed of course, they request it, but don't demand it.

So why no comment when asked about this a couple of months back?
 
So why no comment when asked about this a couple of months back?
Maybe as part of the co-marketing agreement they were only allowed to talk about AMD products and FSR.

Have Bethesda ever actually integrated DLSS or FSR into a product before? I don't think even the current Fallout 76 updates use any kind of 3rd party upscaler (but I've never played it so can't say for sure).

It's possible that without AMD's intervention, Starfield wouldn't support any kind of advanced temporal upscaler. AMD's engineers might have been updating the engine to support FSR.

At this point, I don't think we can take it for granted that Starfield would have supported DLSS in a timeline where AMD weren't involved.
 
Yes, many times. Deathloop, Redfall, Doom Eternal, etc.

Sorry, I should have been more specific that I meant Bethesda Game Studios - of Todd Howard fame - that uses the Gamebryo / Creation / Creation 2 lineage of engine and that are the ones making Starfield and using that engine line.

Deathloop, Redfall, Doom Eternal are all by different developers, and they don't use the Creation Engine.

Just because a sister studio of Bethesda Game Studios have implemented DLSS or FSR into a completely different engine, it doesn't automatically mean that BGS's Creation Engine will have imbibed those none trivial engine changes.

Technologies like DLSS or FSR are integrated at the engine or studio level - it's not really a publisher level matter (unless the publisher tries to force a common set of technologies, but that isn't the case with Zenimax).

Have BGS ever integrated DLSS or FSR into their engine before?
 
Have BGS ever integrated DLSS or FSR into their engine before?
It is possible the BGS team lacks the technical competence to add such a feature, almost as surprising as the game lacking ray tracing.
I believe there is a DLSS/XeSS/FSR mod available for Skyrim by modder PureDark.
 
Yeah, but it breaks much more important ENB so it's useless at the moment.
Not really. People are using DLAA + ENB until the DLSS + ENB version is complete.
  1. DLAA+ENB, REALLY WORK! FINALLY! I don't need to use DSR, don't need to see soap by TAA. Thank you, take my vote for this mod!
  2. I hope for possibility use DLSS + ENB, but DLAA work pretty good
 
Last edited by a moderator:
Sorry, I should have been more specific that I meant Bethesda Game Studios - of Todd Howard fame - that uses the Gamebryo / Creation / Creation 2 lineage of engine and that are the ones making Starfield and using that engine line.

Deathloop, Redfall, Doom Eternal are all by different developers, and they don't use the Creation Engine.
If the engine has TAA then integrating any upscale solution is a trivial task, and if there are no graphics programmers in BGS (which is the only way this would be a problem) then they can ask id to help them - which they did for some gameplay elements already anyway apparently.

In other words, this suggestion makes no sense.
 
If the engine has TAA then integrating any upscale solution is a trivial task, and if there are no graphics programmers in BGS (which is the only way this would be a problem) then they can ask id to help them - which they did for some gameplay elements already anyway apparently.

In other words, this suggestion makes no sense.
And yet we've seen plenty of sub par implementations including with fsr, dlss xess.

Whether that was the reason or not I'm sure getting multiple versions in the game was probably lowest on their priorities, regardless of how easy it is to do.
With all the time spent debugging fixing improving what they had.
Especially once they had something that was suitable across everything.

Hell, even got people complaining why haven't they said will have FSR3 or why's it's not on the list of games that will get support.
Hopefully most won't sttetch to say it's because of contract.
 
If the engine has TAA then integrating any upscale solution is a trivial task, and if there are no graphics programmers in BGS (which is the only way this would be a problem) then they can ask id to help them - which they did for some gameplay elements already anyway apparently.


Not all TAA systems would meet FSR's requirements for three colour, depth and velocity buffers. And have Bethesda (the game studio, not the publisher) used TAA in their Creation engine before? I don't know.

I asked you explicitly about DLSS and FSR in the post you just responded to (you edited out the question, perhaps accidentally). But have they?

From consoles I'm only aware of regular current frame upscaling, but have they (Bethesda the studio, not the publisher) used TAA with Creation in any capacity before? They may have, I don't know. But I'm asking.

And why would a graphics programmer's priority on Starfield naturally be to implement an arbitrary upscaling system if project priorities had them do more crucial work? The game is delayed, and people have been crunching hard on the game.

And why should a hypothetical of your own making - iD coming into to do coding on an engine that they don't know and don't use - be the litmus test of whether Bethesda (the studio, not the publisher) could benefit from AMD engineers assisting for free?

In other words, this suggestion makes no sense.

It only makes no sense if you create loaded, artificial parameters within which the suggestion can make no sense. Which I suspect is what you just tried to do.

And riddle me this, if it's: super easy; super trivial; any graphics programmer could do it (and it would be their highest priority); the game can already basically do it; iD can do it easily for free; ..... why have AMD engineers spent time hands on with the game's code making it FSR ready?
 
Status
Not open for further replies.
Back
Top