Friday, March 6th 2020

AMD RDNA2 Graphics Architecture Detailed, Offers +50% Perf-per-Watt over RDNA

With its 7 nm RDNA architecture that debuted in July 2019, AMD achieved a nearly 50% gain in performance/Watt over the previous "Vega" architecture. At its 2020 Financial Analyst Day event, AMD made a big disclosure: that its upcoming RDNA2 architecture will offer a similar 50% performance/Watt jump over RDNA. The new RDNA2 graphics architecture is expected to leverage 7 nm+ (7 nm EUV), which offers up to 18% transistor-density increase over 7 nm DUV, among other process-level improvements. AMD could tap into this to increase price-performance by serving up more compute units at existing price-points, running at higher clock speeds.

AMD has two key design goals with RDNA2 that helps it close the feature-set gap with NVIDIA: real-time ray-tracing, and variable-rate shading, both of which have been standardized by Microsoft under DirectX 12 DXR and VRS APIs. AMD announced that RDNA2 will feature dedicated ray-tracing hardware on die. On the software side, the hardware will leverage industry-standard DXR 1.1 API. The company is supplying RDNA2 to next-generation game console manufacturers such as Sony and Microsoft, so it's highly likely that AMD's approach to standardized ray-tracing will have more takers than NVIDIA's RTX ecosystem that tops up DXR feature-sets with its own RTX feature-set.
AMD GPU Architecture Roadmap RDNA2 RDNA3 AMD RDNA2 Efficiency Roadmap AMD RDNA2 Performance per Watt AMD RDNA2 Raytracing
Variable-rate shading is another key feature that has been missing on AMD GPUs. The feature allows a graphics application to apply different rates of shading detail to different areas of the 3D scene being rendered, to conserve system resources. NVIDIA and Intel already implement VRS tier-1 standardized by Microsoft, and NVIDIA "Turing" goes a step further in supporting even VRS tier-2. AMD didn't detail its VRS tier support.

AMD hopes to deploy RDNA2 on everything from desktop discrete client graphics, to professional graphics for creators, to mobile (notebook/tablet) graphics, and lastly cloud graphics (for cloud-based gaming platforms such as Stadia). Its biggest takers, however, will be the next-generation Xbox and PlayStation game consoles, who will also shepherd game developers toward standardized ray-tracing and VRS implementations.

AMD also briefly touched upon the next-generation RDNA3 graphics architecture without revealing any features. All we know about RDNA3 for now, is that it will leverage a process node more advanced than 7 nm (likely 6 nm or 5 nm, AMD won't say); and that it will come out some time between 2021 and 2022. RDNA2 will extensively power AMD client graphics products over the next 5-6 calendar quarters, at least.
Add your own comment

242 Comments on AMD RDNA2 Graphics Architecture Detailed, Offers +50% Perf-per-Watt over RDNA

#26
GeorgeMan
W1zzard
Considering that currently shipping Navi is "Navi 10" and "Navi 14", which can be summarized as "Navi 1x", I would assume that the next GPUs are "Navi 20" and "Navi 30", so the x stands for "any number", not "x-times improvement"
Exactly that. I don't understand why it confuses people. The slides are crystal clear.
Posted on Reply
#27
R0H1T
No they're not, the naming scheme or (internal) chip jargons shouldn't be referenced in such a way!
Posted on Reply
#28
Valantar
Sorry, but how on earth does anyone see "Navi 2X" in friggin' quotes without any further data and think "Oh, that must mean 2x the performance"? Sorry, but that is a rather extreme leap of the imagination. Also, x as a multiplier is generally lower case, this is upper case, which is generally X as an unknown variable. 2X = 20, 21, 22, etc. is much more reasonable of an assumption than 2X = 2x performance.

2X is the generational code name for all consumer-oriented non-semi custom RDNA 2 silicon, with each piece of silicon then having a distinct second digit. End discussion.
Posted on Reply
#29
ShurikN
Hopefully the cards launch in Q3 rather than Q4. Having new gaming products in time for CP2077 would be huge.
Posted on Reply
#30
Vya Domus
R0H1T
Oh in that case AMD should fire the wise guy who made that slide, I mean it makes little sense even now!
They shouldn't fire anyone, everyone's comprehension is appalling.
Posted on Reply
#31
Bruno Vieira
btarunr
Fixed title, sorry for the confusion.

NVIDIA went a step further than Intel in supporting not just tier-1, but also tier-2. The complete sentence was:

"NVIDIA and Intel already implement VRS tier-1 standardized by Microsoft, and NVIDIA "Turing" goes a step further in supporting even VRS tier-2 ."
AMD also stayed that the VRS and Raytracing implementation were made in conjunction with microsoft, so it has to have the highest tier available
Posted on Reply
#32
droopyRO
I hope that the "plague" dose not slow the production of this chips too much.
Posted on Reply
#33
HD64G
droopyRO
I hope that the "plague" dose not slow the production of this chips too much.
Maybe the delay in next-gen Navi GPUs is due to that. Will know for sure once they launch them.
Posted on Reply
#34
jmcslob
I feel as though the confusion here was mostly caused because English isn't everyone's first language.

...and by "everyone's" I mean everyone in this thread but me.
Posted on Reply
#35
kapone32
50% more performance per watt. So if a 200 Watt GPU gives you 80 FPS the next gen would give you 120 FPS? Or a more real thought would be 100 FPS.
Posted on Reply
#36
kings
Talk is cheap, I believe when I see it.

Vega's architecture was also supposed to bring 4X the performance per watt than previous generation.

But I hope this time it’s what AMD says.
Posted on Reply
#37
Valantar
kings
Talk is cheap, I believe when I see it.

Vega's architecture was also supposed to bring 4X the performance per watt than previous generation.

But I hope this time it’s what AMD says.
Talk isn't particularly cheap when you're talking to investors and financial analysts. Fail to meet your promises and at best your stock tanks, at worst you get sued by shareholders for lying to them. Still, 50% sounds like a lot. Fingers crossed that it turns out that way - then we'll have a real fight on our hands in the next GPU generation, and prices ought to reflect that.
Posted on Reply
#38
EarthDog
Considering that currently shipping Navi is "Navi 10" and "Navi 14", which can be summarized as "Navi 1x", I would assume that the next GPUs are "Navi 20" and "Navi 30", so the x stands for "any number", not "x-times improvement"
There we go....logic and intelligence prevail!! :)

Anyway, I can't wait to see these on the market and AMD catch up to performance per /W over the 12nm Turing parts. The put some special sauce in the 5600 XT which put it on par with Nvidia, so this should be interesting, as well as seeing an apples to apples comparison with Ampre and its increase in efficiency p /w along with the shrink to 7nm...I bet NV still holds that lead........
Posted on Reply
#39
Slizzo
EarthDog
There we go....logic and intelligence prevail!! :)

Anyway, I can't wait to see these on the market and AMD catch up to performance per /W over the 12nm Turing parts. The put some special sauce in the 5600 XT which put it on par with Nvidia, so this should be interesting, as well as seeing an apples to apples comparison with Ampre and its increase in efficiency p /w along with the shrink to 7nm...I bet NV still holds that lead........
"special sauce"? If by special sauce you mean they freaked out and pushed out a firmware right at launch that blew initial thermal and power targets, then, yeah, "Special".
Posted on Reply
#40
oxidized
If only performance was your main problem AMD...
Posted on Reply
#41
EarthDog
Slizzo
"special sauce"? If by special sauce you mean they freaked out and pushed out a firmware right at launch that blew initial thermal and power targets, then, yeah, "Special".
Ehh, it was still comparable to the RTX 2060 which it competes with... that is different than we saw with 5500 XT and 5700/5700XT.
Posted on Reply
#42
efikkan
I feel it's disappointing to see that there are no major new architecture in sight; just more iterations of Navi.

Chomiq
If they'll actually deliver the only thing that will prevent me from buying this would be the driver issues that we saw with almost every launch from team red's GPU camp.
It has been a recurring subject with every release, since the underlying driver problems remains unfixed.

kings
Talk is cheap, I believe when I see it.

Vega's architecture was also supposed to bring 4X the performance per watt than previous generation.
And Polaris promised 2.5x performance per watt, while it turned out that they meant if it ran at 850 MHz vs. an older GCN at a higher clock…
AMD's GPU department have a long standing tradition of over-promising and under-delivering, unfortunately.
Posted on Reply
#43
kapone32
efikkan
I feel it's disappointing to see that there are no major new architecture in sight; just more iterations of Navi.


It has been a recurring subject with every release, since the underlying driver problems remains unfixed.


And Polaris promised 2.5x performance per watt, while it turned out that they meant if it ran at 850 MHz vs. an older GCN at a higher clock…
AMD's GPU department have a long standing tradition of over-promising and under-delivering, unfortunately.
I am not sure about that Polaris is faster than Vega and the clocks for Vega are 1630 MHz.

kapone32
I am not sure about that Polaris is faster than Vega and the clocks for Vega are 1630 MHz.
[
Forgive I should have said Tahiti at 1100 Mhz but Tahiti was no joke.
Posted on Reply
#44
Vya Domus
kings
Vega's architecture was also supposed to bring 4X the performance per watt than previous generation.
Are you on hallucinogenics ? Vega was never supposed to bring 4X the performance per watt, I swear you're all gonna scour the depths of the internet just to find that one fake leak to make your point.



This was from a fake Aprils fools leak. Come on, just how low will you fanboys go.

https://wccftech.com/vega-teaser-slides-leak-nda/
Posted on Reply
#45
efikkan
kapone32
[quote=efikkan]
And Polaris promised 2.5x performance per watt, while it turned out that they meant if it ran at 850 MHz vs. an older GCN at a higher clock…
I am not sure about that Polaris is faster than Vega and the clocks for Vega are 1630 MHz.[/quote]That's not what I said. Try again ;)
Posted on Reply
#46
Cheeseball
Not a Potato
EarthDog
There we go....logic and intelligence prevail!! :)

Anyway, I can't wait to see these on the market and AMD catch up to performance per /W over the 12nm Turing parts. The put some special sauce in the 5600 XT which put it on par with Nvidia, so this should be interesting, as well as seeing an apples to apples comparison with Ampre and its increase in efficiency p /w along with the shrink to 7nm...I bet NV still holds that lead........
It was more like that they wanted to easily defeat the 1660 Super and 1660 Ti, but because it was priced so close to the RTX 2060 non-Super, they decided to increase the clocks to make it competitive with it. They just changed targets at that price range, and it was a good idea.

kapone32
I am not sure about that Polaris is faster than Vega and the clocks for Vega are 1630 MHz.
You're getting mixed up there dude. Polaris and Vega are GCN (4th and 5th gen) architectures.
Posted on Reply
#47
kapone32
Cheeseball
It was more like that they wanted to easily defeat the 1660 Super and 1660 Ti, but because it was priced so close to the RTX 2060 non-Super, they decided to increase the clocks to make it competitive with it. They just changed targets at that price range, and it was a good idea.



You're getting mixed up there dude. Polaris and Vega are GCN (5th gen) architectures.
Yeah I was too quick with the trigger I should have said Vega vs Navi which really impressed me with the fact that the 5700XT is faster than the Vega 64 with 1/2 the ROPs.
Posted on Reply
#48
IceShroom
Vya Domus
Are you on hallucinogenics ? Vega was never supposed to bring 4X the performance per watt, I swear you're all gonna scour the depths of the internet just to find that one fake leak to make your point.



This was from a fake Aprils fools leak. Come on, just how low will you fanboys go.

https://wccftech.com/vega-teaser-slides-leak-nda/
Looks like Nvidia guys cant tell apart which are fake and which are official AMD slide. And the fake slide is spread by WCCFTECH.o_O

Dont worry someone will take this video as official AMD video also.(Dont click)
Posted on Reply
#49
Cheeseball
Not a Potato
kapone32
Yeah I was too quick with the trigger I should have said Vega vs Navi which really impressed me with the fact that the 5700XT is faster than the Vega 64 with 1/2 the ROPs.
While it is technically impressive, please note that RDNA is quite different to GCN at SIMD-level, where RDNA works with SIMD32 (native Wave32!!) and single-cycle instructions.

GCN (5th gen) used SIMD16, which means it issues instructions every 4(??) cycles, where as RDNA issues it every cycle. This inherently makes a 40 CU (RX 5700 XT) cluster faster than the previous 64 CU cards (Vega 64/Radeon VII).

Depending on what you're trying to achieve (raw core performance vs. optimized IPC), GCN5 can still compete well against its younger sibling. However RDNA can do everything GCN5 can do, except beating it in raw compute loads.
Posted on Reply
#50
kings
Vya Domus
Are you on hallucinogenics ? Vega was never supposed to bring 4X the performance per watt, I swear you're all gonna scour the depths of the internet just to find that one fake leak to make your point.

This was from a fake Aprils fools leak. Come on, just how low will you fanboys go.
My mistake then, I apologize. I didn't know about these Aprils fools slides.

As for the fanboy part, you're wrong, but you're entitled to your opinion.
Posted on Reply
Add your own comment