AMD: Speculation, Rumors, and Discussion (Archive)

Status
Not open for further replies.
That's why WCCF did a second video which I posted earlier. At least we can sort of extrapolate the impact of Async on/off now. Hopefully it won't take too long for id to get it to work on Nvidia GPUs.
Yeah fingers crossed soon for Pascal, unfortunately it would still need to be benchmarked with the various shadow settings just like PCGameshardware did originally, along with different zones.
Shame Guru3d did not test a 980ti as it seems quite a few people on various forums are seeing the Vulkan boost for it (separate to the async compute additional boost); would help to show whether something unusual with the Pascal cards-driver or maybe setting/zone used by Guru3d.
I also notice Guru3d performance figures also seem a bit low for the 480 between OpenGL and Vulkan - again could be various reasons.
Cheers
 
Last edited:
Nvidida GPUs seeing increases with Vulkan where held back buy the CPU. If the system was GPU bound and not CPU bound then the improvements are marginal or simply non-existent (Vulkan is actualy slower than OGL).

All following benchs were done with Async OFF (FXXA was selected unfortunitaly so you can add a few more FPS to the RX480 scores:
http://www.gamersnexus.net/game-bench/2510-doom-vulkan-vs-opengl-benchmark-rx-480-gtx-1080

vulkan-doom-1080p.png
vulkan-doom-1080p-percent.png


vulkan-doom-1440p.png
vulkan-doom-1440p-percent.png


vulkan-doom-4k.png

vulkan-doom-4k-percent.png

More at the link..
 
http://videocardz.com/62117/amd-confirms-radeon-rx-470-and-rx-460-specifications

PCWS8np.jpg


fiROc1Q.jpg



I confess I'm quite disappointed at the number of active CUs in the RX460. Why only 14 if the GPU has 16?
Unless they're clocking the whole chip at close to 1500MHz, I find it very hard to believe this will get Pitcairn levels of performance.

Sure, they can introduce a RX465 further down the line, but for the following months they're stuck with this.

However, I'm really glad they went with the true mITX form factor as the default PCB size and cooling solution. With that price, hardware video codec and output capabilities this is definitely the card to get for HTPC users. At least until Kaby Lake and motherboards with HDMI 2.0 that is..
 
http://videocardz.com/62117/amd-confirms-radeon-rx-470-and-rx-460-specifications

I confess I'm quite disappointed at the number of active CUs in the RX460. Why only 14 if the GPU has 16?
Unless they're clocking the whole chip at close to 1500MHz, I find it very hard to believe this will get Pitcairn levels of performance.

Sure, they can introduce a RX465 further down the line, but for the following months they're stuck with this.

However, I'm really glad they went with the true mITX form factor as the default PCB size and cooling solution. With that price, hardware video codec and output capabilities this is definitely the card to get for HTPC users. At least until Kaby Lake and motherboards with HDMI 2.0 that is..
What's more curious is the fact that no matter how many leaked slides they get, they apparently don't get the official slide decks - those specs were confirmed already on RX 480 launch
 
  • Like
Reactions: xEx
Nvidida GPUs seeing increases with Vulkan where held back buy the CPU. If the system was GPU bound and not CPU bound then the improvements are marginal or simply non-existent (Vulkan is actualy slower than OGL).

All following benchs were done with Async OFF (FXXA was selected unfortunitaly so you can add a few more FPS to the RX480 scores:
http://www.gamersnexus.net/game-bench/2510-doom-vulkan-vs-opengl-benchmark-rx-480-gtx-1080

vulkan-doom-1080p.png
vulkan-doom-1080p-percent.png


vulkan-doom-1440p.png
vulkan-doom-1440p-percent.png


vulkan-doom-4k.png

vulkan-doom-4k-percent.png

More at the link..


I wanna see more benchmarks with various GPU/CPU/settings combos. It seems some older GPUs get unreal boost.

https://www.reddit.com/r/Amd/comments/4sd75l/doom_123_vulcan_performance_increase_on_r9_280x/
 
Last edited:
I confess I'm quite disappointed at the number of active CUs in the RX460. Why only 14 if the GPU has 16?

There is no external power. Maybe two CUs are disabled to get under the 75W spec?
 
http://videocardz.com/62117/amd-confirms-radeon-rx-470-and-rx-460-specifications

I confess I'm quite disappointed at the number of active CUs in the RX460. Why only 14 if the GPU has 16?
Unless they're clocking the whole chip at close to 1500MHz, I find it very hard to believe this will get Pitcairn levels of performance.

The best Polaris 11 dies will be going towards laptop design wins, where the margins are a fair bit higher than US$99 desktop cards. AMD has messaged Polaris 11 as a laptop-focused chip.

The RX480M is mentioned in AMD marketing materials as 16CU Polaris 11 for laptops at 35W TDP (clock speed not given).
 
That shouldn't be hard. The 480 has 36CUs and draws about 170W: 14/36×170W = 66W. There's more to GPUs than CUs, but then again most other components should be scaled down as well, and it should also be clocked a bit lower.
 
There is no external power. Maybe two CUs are disabled to get under the 75W spec?
That laptop design win has Polaris 11 replacing a GPU that goes up to 45W (according to wikipedia). Moreover, at 16CUs it's practically "half" of a RX 470 which has a TDP of 110W in its desktop form.
If Polaris 11 needs to disable CUs to reach 75W then the chip is DOA, in my opinion.


The best Polaris 11 dies will be going towards laptop design wins, where the margins are a fair bit higher than US$99 desktop cards. AMD has messaged Polaris 11 as a laptop-focused chip.
That makes more sense, though I wonder if a tiny chip like P11 does get higher margins in a laptop than a desktop graphics card at $99. The larger chips like GM204 in MXM cards do carry a hefty premium but you can see the smaller GPUs inside every other laptop starting at ~650€.


I hope AMD has made the clocks go up in order to compensate for the CU deficit, otherwise it's just a replacement to Bonaire and not Pitcairn.
 
Last edited by a moderator:
I wanna see more benchmarks with various GPU/CPU/settings combos. It seems some older GPUs get unreal boost.

https://www.reddit.com/r/Amd/comments/4sd75l/doom_123_vulcan_performance_increase_on_r9_280x/

Well I tried the Vulkan patch briefly last night and I was very suprised. My 7970 was hovering around high 40s to high 60FPS most of the time and very playable with OpenGL. As I have a 3770K overclocked to 4.6GHz I wasn't expecting much, however when I loaded the Titan level last night I was pretty shocked to see framerates in excess of 100FPS when wondering around and High 70s/ 80FPS in the few fire fights I played.
 
Just installed the non-whql-64bit-radeon-software-crimson-16.7.2-win10-win8.1-win7-july9.exe hotfix driver for some DOOM-testing action, and with this version the Radeon Settings app crashes repeatably every time the "Games" tab is clicked. This means I can't access the 3D quality settings or Overdrive... :(

Edit:
Unable to enable (blah!) Vulkan support in DOOM. After changing API the game states "preparing to restart" and then nothing more happens. Waited like 5 minutes, you just see sparks flying by and music playing. Doesn't seem normal...
 
Last edited:
Just installed the non-whql-64bit-radeon-software-crimson-16.7.2-win10-win8.1-win7-july9.exe hotfix driver for some DOOM-testing action, and with this version the Radeon Settings app crashes repeatably every time the "Games" tab is clicked. This means I can't access the 3D quality settings or Overdrive... :(

Edit:
Unable to enable (blah!) Vulkan support in DOOM. After changing API the game states "preparing to restart" and then nothing more happens. Waited like 5 minutes, you just see sparks flying by and music playing. Doesn't seem normal...

Have you tried the usual, DDU in safe mode? Make sure you have the latest Vulkan run time installed too.
 
This corroborates my experience with regards to the huge performance boots on the Fury X coupled with a 6700K. It totally nukes a 1070 at 1080p and 1440P (and is probably as fast or faster than a 1080).

https://www.computerbase.de/2016-07/doom-vulkan-benchmarks-amd-nvidia/
Although it is another review like Guru3D that saw no improvements for any Nvidia.
Something is not right with all the testing because some are seeing still pretty good improvements for Nvidia cards (albeit not comparable to GCN but still into double figure % increase) and some reviews or tests seeing zero.
Bearing in mind this improvement is not just Async Compute but also low level API benefits.
Fingers crossed a publication actually does a very in-depth analysis, looking at the diverse variables that may influence performance.
Cheers
 
Status
Not open for further replies.
Back
Top