AMD Radeon RDNA2 Navi (RX 6500, 6600, 6700, 6800, 6900 XT)

[pjbliverpool]

I don't think I'm alone in preferring the flashy lights, fancy mirrors, etc..

Definitely not. I can't wait to see RT in CB 2077, it seems like the perfect title to show it off.

News from hothardware:

The answer of the rasterizer let me get more ??? This is teh answer from HotHardware from AMD Scott Herkelman:

Someone should ask them why they chose to go from 1 primitive unit per SA to 1 per SE while in the consoles they kept them at 1 per SE.

In today's world I think it's better to have better rasterization than RT since the number of games with RT are very limited and the number of games with useful RT are like unicorns. And all of the "but future proof!" is no sense to me. Nvidia is barely making it in tern of performance and AMD is barely barely making so I don't think today's GPU will be able to use full RT in next gem games.

I'm not sure I agree with this. Most games releasing on the new gen consoles use some level of RT which means that the majority of PC games from now on are likely to include RT. It may soon become the case that benchmarking without RT is as relevant as benchmarking without AA 10 years ago.

I really have no idea why we have RT GPUs...*at least* we are 1 gem away from truly usable perf. and the space on Die could be use to have faster and better GPUs capable of more "traditional" effect that would be more useful than slightly better shadows that eat 20-30-50% of the frame rate just so you can "admire" the difference in a static pic in a side by side comparative. Ridiculous...

But when you look at the difference it makes in Miles Morales on relatively anemic hardware there's clearly an argument for it, especially on high end Ampere GPU's which likely have 3-4x that level of performance at least. Personally I do agree that shadows are a questionable use of the technology given the level of performance it seems to suck up, although I'd need to see a more detailed analysis to finalise that conclusion. But the reflections in Watch Dogs for example are transformative IMO.

 

[manux]

For fun I tried bmw blender scene on my 4670k+3070fe. CPU rendering I aborted at 2 minutes. Estimate was render would have taken 12 minutes. Optix with 3070 took 16.9s. It's insane what old machine can do with modern gpu. 3070fe I have seems to like 1950Mhz gpu clock during blender usage. Quick overclock with afterburner and gpu clock was solid 2100Mhz and mem clock +926(about 15.8Gbit). Rendering time overclocked was 15.3s.

 

[DavidGraham]

Even the 3070 is 20% faster than 6800XT in Call Of Duty Black Ops with RT @4K.

 

[PSman1700]

Difficult to believe this statement. For example blender in benchmark below is 3080 11.9s versus 2080ti 19.9s or 2080 super 26.1s. Pure ray tracing perf in ampere is much improved. Also minecraft or quake2 rtx would show how pure raytracing perf has advanced. I'm not 100% sure where 6800xt falls, but I believe it's somewhere around 38s for same bmw blender benchmark.

https://www.phoronix.com/scan.php?page=article&item=blender-290-rtx3080&num=3

Ye, thought of sharing the quake 2 RTX improvement going from turing to ampere, which saw a leap in RT performance more then normal rasterization did. The performance improvement going from turing to ampere in pure RT is on a whole different level.

But when you look at the difference it makes in Miles Morales on relatively anemic hardware there's clearly an argument for it, especially on high end Ampere GPU's which likely have 3-4x that level of performance at least. Personally I do agree that shadows are a questionable use of the technology given the level of performance it seems to suck up, although I'd need to see a more detailed analysis to finalise that conclusion. But the reflections in Watch Dogs for example are transformative IMO.

Couldnt agree more.

Even the 3070 is 20% faster than 6800XT in Call Of Duty Black Ops with RT @4K.

And thats that, in CoD BO, where the game isnt all that taxing or even close to what say CP2077 is doing. Its the combination of normal rendering and use of RT without alot of compromises that actually have to do with RT.

 

[xEx]

Difficult to believe this statement. For example blender in benchmark below is 3080 11.9s versus 2080ti 19.9s or 2080 super 26.1s. Pure ray tracing perf in ampere is much improved. Also minecraft or quake2 rtx would show how pure raytracing perf has advanced. I'm not 100% sure where 6800xt falls, but I believe it's somewhere around 38s for same bmw blender benchmark.

https://www.phoronix.com/scan.php?page=article&item=blender-290-rtx3080&num=3

But my friend, you are forgetting that the 3080 has a lot more horse power? If you look at the 3070 vs 2080TI which has roughly the same performance in raster, they also have the same perf in RT. Ampere is better because it has more RT cores not better RT cores. Not the same thing.

No matter when they chose to bring hardware accelerated RT to market the fact is that there would be zero game support on day one and it would be many years before there is wide market adoption. So given those facts what percentage of die area would you suggest be dedicated to the first RT implementation? 25%, 50%?

I think right now no1 knows the answer to that question.

Now, as a consumer yeah you can say "it's not fast enough yet, I won't weight RT in the balance", and that's perfectly fine. But nVidia, AMD, Intel, Imgtech whoever ... have to start somewhere...

IMHO I think it would be better for next gem when GPUs could play 4K like they do for 1440p today. So u don't get to play at 60FPS to get slightly better shadows. or 30 for a meaningful experience. That is my point.

Definitely not. I can't wait to see RT in CB 2077, it seems like the perfect title to show it off.



Someone should ask them why they chose to go from 1 primitive unit per SA to 1 per SE while in the consoles they kept them at 1 per SE.



I'm not sure I agree with this. Most games releasing on the new gen consoles use some level of RT which means that the majority of PC games from now on are likely to include RT. It may soon become the case that benchmarking without RT is as relevant as benchmarking without AA 10 years ago.



But when you look at the difference it makes in Miles Morales on relatively anemic hardware there's clearly an argument for it, especially on high end Ampere GPU's which likely have 3-4x that level of performance at least. Personally I do agree that shadows are a questionable use of the technology given the level of performance it seems to suck up, although I'd need to see a more detailed analysis to finalise that conclusion. But the reflections in Watch Dogs for example are transformative IMO.

Yes but I think we gonna start seeing it for the next gem games half way 2021 and 2022.Yeah we will have WD and CP and maybe it would make sense to get a 700 dollar VGA to play those games at all of it glory but the VGA below them just doesn't have enough power to do so. 6800 is barely making sense for RT. below that what can you play? Dirt 5?

 

[Kaotik]

Even the 3070 is 20% faster than 6800XT in Call Of Duty Black Ops with RT @4K.

WCCFTechs numbers in that video are quite a way off from FPS Reviews numbers


vs



https://www.thefpsreview.com/2020/11/18/amd-radeon-rx-6800-xt-and-radeon-rx-6800-review/6/

(they hadn't had the time to test all the cards in between so it's just the 3080 vs Radeons)

 

[troyan]

It depends on the scene. Cold War supports RT sun light shadows, point light shadows and AO. Scenes without the sun will always less taxable on the GPU than scenes with daylight.

 

[CarstenS]

E.g. AMD already took out Alternate.de (the store that started selling their RX 6800XT at 1049€) from their official list of recommended retailers.

The rage did not last long it seems. It's been put back there.

 

[trinibwoy]

I think right now no1 knows the answer to that question.

Which is even more reason to get the ball rolling sooner rather than later. Why wait when we don’t know what we’re waiting for.

IMHO I think it would be better for next gem when GPUs could play 4K like they do for 1440p today. So u don't get to play at 60FPS to get slightly better shadows. or 30 for a meaningful experience. That is my point.

I hope you can appreciate the irony in this position. How can you advocate for 4K while downplaying the IQ impact of RT. In what way is 4K worth the performance hit over 1440p?

 

[j^aws]

Someone should ask them why they chose to go from 1 primitive unit per SA to 1 per SE while in the consoles they kept them at 1 per SE.

This would be a good question. I would also ask if the culling rate is still 2 triangles per clock per Prim Unit, or is it higher now?

BTW, we know from Hotchips the Prim Unit arrangement for XSX is 1 per Shader Array (and driver leak also confirms RDNA1 arrangement), but we still don't know arrangement for PS5.

 

[Digidi]

This would be a good question. I would also ask if the culling rate is still 2 triangles per clock per Prim Unit, or is it higher now?

BTW, we know from Hotchips the Prim Unit arrangement for XSX is 1 per Shader Array (and driver leak also confirms RDNA1 arrangement), but we still don't know arrangement for PS5.

Culling rate is the same like Navi 10 8 Prim/clock culling there are also 4 prim units on navi21. check slides

 

[j^aws]

Culling rate is the same like Navi 10 8 Prim/clock culling there are also 4 prim units on navi21. check slides

I've seen the slide where they show 4 Prim Units for Navi21, 1 Prim Unit per Shader Engine rather than Shader Array. Which slide talks about culling rate, as there are so many, I might have missed it?

 

[xEx]

Which is even more reason to get the ball rolling sooner rather than later. Why wait when we don’t know what we’re waiting for.



I hope you can appreciate the irony in this position. How can you advocate for 4K while downplaying the IQ impact of RT. In what way is 4K worth the performance hit over 1440p?

I give u that half way. Playing in a living with a TV does need a 4K image. but my point is, get to RT when there is no better use for that performance. Yes is going to be a try and error circle but at least do in when the perf is there to play with meaningful effects at 30FPS with 700 dollars VGAs. I think RT would be the perfect addition to next gem cards when the performance will be there to back it up.

 

[pjbliverpool]

I've seen the slide where they show 4 Prim Units for Navi21, 1 Prim Unit per Shader Engine rather than Shader Array. Which slide talks about culling rate, as there are so many, I might have missed it?

I can't find it at the moment but there have definitely been a couple of slides that confirm it is 2 pre-culled in, 1 culled out for the Primitive units.

 

[Digidi]

Yes ant this is strange. You get only 4 primitves after culling but you have 8 scan converter which convert the primitives to pixels. This will be a total unbalanc? This makes no sense.

No, of course not. Big Navi is some 60% faster in pixel fill and more than twice as fast in some triangle throughputs, not so much in others.

Also this makes no sense. If both have the same geometry processor Navi21 could not be faster than Navi10.

Maybe @CarstenS can provide some detail information what he have measured with values? Thank you in advanced.

 

[trinibwoy]

I give u that half way. Playing in a living with a TV does need a 4K image. but my point is, get to RT when there is no better use for that performance. Yes is going to be a try and error circle but at least do in when the perf is there to play with meaningful effects at 30FPS with 700 dollars VGAs. I think RT would be the perfect addition to next gem cards when the performance will be there to back it up.

That’s a good way to put it. I think we’re already at the point where there’s no better use for those transistors.

Mesh shaders address the inflexibility and lack of scalability in the classic geometry rasterization pipeline for primary visibility. RT provides an elegant and scalable solution for indirect lighting. Both are much better uses of performance than scaling up resolutions on the same old stuff.

 

[Putas]

Also this makes no sense. If both have the same geometry processor Navi21 could not be faster than Navi10.

Wasn't the NGG supposed to optionally finish rasterization via shaders? The rest can be clocks.

 

[DavidGraham]

AMD GPUs perform worse with DXR1.1 3D Mark test than DXR1.0.





https://www.legitreviews.com/amd-radeon-rx-6800-xt-and-radeon-rx-6800-review_223774/9

 

[dorf]

WCCFTechs numbers in that video are quite a way off from FPS Reviews numbers

Tom's numbers are in line with Wccftech. Maybe FPS Review is testing some simpler scene.

 

[Jawed]

Finally, for TU102 that means ~29.4912mm2 BVH transversal in total, which is 3.9% of total die size. This is only a comparison of BVH area there're probably other regions that needs to be beefed up for RT acceleration.

It makes me think that NVidia developed ray tracing acceleration for the professional rendering industry and then persuaded Microsoft that it was time to add it to D3D. Same as how tensor cores were designed for the machine learning industry.

The key problem is that brute force ray tracing as seen in Control, produces terrible performance. Brute force appears to have hit a ceiling with Ampere (if we talk about rays per unit bandwidth).

Difficult to believe this statement. For example blender in benchmark below is 3080 11.9s versus 2080ti 19.9s or 2080 super 26.1s. Pure ray tracing perf in ampere is much improved. Also minecraft or quake2 rtx would show how pure raytracing perf has advanced. I'm not 100% sure where 6800xt falls, but I believe it's somewhere around 38s for same bmw blender benchmark.

https://www.phoronix.com/scan.php?page=article&item=blender-290-rtx3080&num=3

Brute force hardware is good for "professional" ray tracing.

I hope you can appreciate the irony in this position. How can you advocate for 4K while downplaying the IQ impact of RT. In what way is 4K worth the performance hit over 1440p?

There's little doubt that 720p video looks more realistic than 4K gameplay

So apparently what we need is upscaled 320p real time path tracing for gaming