[AMD_Robert] Concerning the AOTS image quality controversy

[Azix]

Because this might get buried and its potentially important. Should do some comparisons to see if only the 1080 runs it improperly.

Hi. Now that I'm off of my 10-hour airplane ride to Oz, and I have reliable internet, I can share some insight.

System specs:

• CPU: i7 5930K
• RAM: 32GB DDR4-2400Mhz
• Motherboard: Asrock X99M Killer
• GPU config 1: 2x Radeon RX 480 @ PCIE 3.0 x16 for each GPU
• GPU config 2: Founders Edition GTX 1080
• OS: Win 10 64bit
• AMD Driver: 16.30-160525n-230356E
• NV Driver: 368.19

In Game Settings for both configs: Crazy Settings | 1080P | 8x MSAA | VSYNC OFF

Ashes Game Version: v1.12.19928

Benchmark results:

2x Radeon RX 480 - 62.5 fps | Single Batch GPU Util: 51% | Med Batch GPU Util: 71.9 | Heavy Batch GPU Util: 92.3% GTX 1080 – 58.7 fps | Single Batch GPU Util: 98.7%| Med Batch GPU Util: 97.9% | Heavy Batch GPU Util: 98.7%

The elephant in the room:

Ashes uses procedural generation based on a randomized seed at launch. The benchmark does look slightly different every time it is run. But that, many have noted, does not fully explain the quality difference people noticed.
At present the GTX 1080 is incorrectly executing the terrain shaders responsible for populating the environment with the appropriate amount of snow. The GTX 1080 is doing less work to render AOTS than it otherwise would if the shader were being run properly. Snow is somewhat flat and boring in color compared to shiny rocks, which gives the illusion that less is being rendered, but this is an incorrect interpretation of how the terrain shaders are functioning in this title.

The content being rendered by the RX 480--the one with greater snow coverage in the side-by-side (the left in these images)--is the correct execution of the terrain shaders.

So, even with fudgy image quality on the GTX 1080 that could improve their performance a few percent, dual RX 480 still came out ahead.

As a parting note, I will mention we ran this test 10x prior to going on-stage to confirm the performance delta was accurate. Moving up to 1440p at the same settings maintains the same performance delta within +/-1%.

https://www.reddit.com/r/Amd/comments/4m692q/concerning_the_aots_image_quality_controversy/

https://www.youtube.com/watch?v=KT2dQZpns-c

980ti. looks like the 480 video but may not be same drivers/game version.

 

[BlitzWulf]

I got a lot more out of him in the comments section

 

[Azix]

I got a lot more out of him in the comments section

Link? Maybe could compile them here

 

[Kenmitch]

Bolded part is interesting....Hmm.

Before thread turns ugly.

AMD_RobertTechnical Marketing 12 points 3 hours ago
I never said it was intentional. I simply stated the fact that the image quality is a bit fudgy, and it could improve perf by a few percent. There was no implication of malice.

 

[BlitzWulf]

https://www.reddit.com/r/Amd/commen...s_image_quality_controversy/d3swm7s?context=3

 

[RaulF]

You guys remember a while back a big controversy of a uk player quality issues in BF4 comparing Nvidia and AMD that the videos look different and it appear that Nvidia was showing less detail and some stuff was missing!

Oh uh.


Sent from my iPhone using Tapatalk

 

[Rvenger]

I don't know. Most people don't hold much weight to AoTS. I wouldn't come to any type of conclusion for either camp on this title.

 

[Bacon1]

You guys remember a while back a big controversy of a uk player quality issues in BF4 comparing Nvidia and AMD that the videos look different and it appear that Nvidia was showing less detail and some stuff was missing!

Oh uh.


Sent from my iPhone using Tapatalk

Yeah he had AF turned off in Nvidia control panel or something IIRC.

Still, it is very important for reviewers to watch the games and look for these differences.

One reason I do like Digital Foundry reviews, though they often only show playback from one card not side by side.

 

[Flapdrol1337]

You guys remember a while back a big controversy of a uk player quality issues in BF4 comparing Nvidia and AMD that the videos look different and it appear that Nvidia was showing less detail and some stuff was missing!

Oh uh.

Yeah I remember that, turned out to be nothing. Had to force off anisotropic filtering to create that result.

Edit:beaten.

 

[swilli89]

Because this might get buried and its potentially important. Should do some comparisons to see if only the 1080 runs it improperly.

So its looking like this is actually more like worst case performance for the 480's and not best case as some have claimed. The rendering difference could actually have a quite large impact on performance so this comparison is basically back to useless now.

You guys remember a while back a big controversy of a uk player quality issues in BF4 comparing Nvidia and AMD that the videos look different and it appear that Nvidia was showing less detail and some stuff was missing!

Oh uh.


Sent from my iPhone using Tapatalk

Yep.. reminds me of 10th grade in high school. I saved up all summer to buy the Geforce 6800GT on release day. A few weeks later I started to notice that the textures looked different somehow in Everquest 2 and Homeworld 2 (when I would pan the camera across ships textures would like downgrade in quality and their lines would become blurry) compared to my friend who was running a Radeon X800. Well lo and behold guess what? A few weeks later I read an article how nVidia had purposefully miffed rendering textures like the examples in my two games. They released new drivers that properly executed the game code and performance dropped 10-15%.

I've never been more disgusted with a purchase in my life than I was with my 6800GT. Not surprising then that I have never supported nvidia again and have always stuck with ATI/AMD since. People like to generalize, they like to say AMD is a for-profit company just like nVidia. They say they are the same. The problem with that statement is it whitewashes the huge difference in the companies. nVidia has an almost palpable corporate culture that they will do ANYTHING to remain on top, and anything can and has involved deceiving their loyal buyers. AMD hasn't always had the very best product, but at the end of the day the user experience has always been pretty much identical and I can be happy that my money is going to a company that at least seems to operate with integrity.

Sorry for all that /rant.

 

[IllogicalGlory]

If this result is representative of the RX 480's performance, then it's slower than a stock 290.

62.5/1.83 = 34.15

On 1080p crazy, my 290 gets over 35 FPS, and it's clocked at the standard 947/1250 clocks. If this is true, then the 480 is an absolute failure unless it uses like 80W and has unbelievable OCing headroom.

 

[Riek]

If this result is representative of the RX 480's performance, then it's slower than a stock 290.

62.5/1.83 = 34.15

On 1080p crazy, my 290 gets over 35 FPS, and it's clocked at the standard 947/1250 clocks. If this is true, then the 480 is an absolute failure unless it uses like 80W and has unbelievable OCing headroom.

consider that it was also more efficient than 1080 --> you are probably looking at each card using less than 90W. So you are not that far off.

 

[parvadomus]

If this result is representative of the RX 480's performance, then it's slower than a stock 290.

62.5/1.83 = 34.15

On 1080p crazy, my 290 gets over 35 FPS, and it's clocked at the standard 947/1250 clocks. If this is true, then the 480 is an absolute failure unless it uses like 80W and has unbelievable OCing headroom.

Dont think so, based on the majority of benchmarks, I have come to the conclusion that today games rely heavily on pixel rate (RX 480 has 64 rops running at 1200mhz, so it should be faster than 290s).
I expect RX 480 to be on Fury X / 980Ti levels or a bit more once overclocked.

 

[Bacon1]

If this result is representative of the RX 480's performance, then it's slower than a stock 290.

62.5/1.83 = 34.15

On 1080p crazy, my 290 gets over 35 FPS, and it's clocked at the standard 947/1250 clocks. If this is true, then the 480 is an absolute failure unless it uses like 80W and has unbelievable OCing headroom.

They are running @ 1440p not 1080p

 

[crisium]

They are running @ 1440p not 1080p

That's a negatory. See the OP.

 

[LTC8K6]

I think we should hear both sides before making any judgements.

Has anyone representing NV commented?

 

[poofyhairguy]

I don't know. Most people don't hold much weight to AoTS.

I think AoTS did a great job of showing us what was to come in full Directx 12 titles.

 

[LTC8K6]

Does 368.19 support Pascal?
368.22 doesn't seem to.

 

[Freddy1765]

nvm

 

[crisium]

I don't know that much about this game and it's settings. If when AMD Robert says "In Game Settings for both configs: Crazy Settings | 1080P | 8x MSAA | VSYNC OFF" he means 1440p, then okay I guess I'll take your word for it.

 

[Bacon1]

That's a negatory. See the OP.

Ah you are right, I was thinking of the 1440p benchmarks people were linking from ashes website.

 

[swilli89]

I think AoTS did a great job of showing us what was to come in full Directx 12 titles.

Exactly. See Warhammer Total War, pretty much same results. Evidence starting to strongly point to AMD built a better DX12 architecture.

 

[topmounter]

nm

 

[Bacon1]

Exactly. See Warhammer Total War, pretty much same results. Evidence starting to strongly point to AMD built a better DX12 architecture.

Yep, Frostbite engine is going to show it off very well too. DICE has some amazing developers and were the ones that spearheaded Mantle with AMD.

BF1, maybe even Mirrors Edge too.

 

[Adul]

Link? Maybe could compile them here

For the questions he asked and got answers for.

Thanks for the post, may I ask one question though? When it's claimed that there is only 51% GPU utilization does that mean 51% of each GPU is being utilized or that the performance scaling is equivalent to 151% of a single card?
permalinkembed

[–]AMD_RobertTechnical Marketing 162 points 9 hours ago*
Scaling is 151% of a single card.
//EDIT: To clarify this, the scaling from 1->2 GPUs in the dual RX 480 test we assembled is 1.83x. The OP was looking only at the lowest draw call rates when asking about the 51%. The single batch GPU utilization is 51% (CPU-bound), medium is 71.9% utilization (less CPU-bound) and heavy batch utilization is 92.3% (not CPU-bound). All together for the entire test, there is 1.83X the performance of a single GPU in what users saw on YouTube. The mGPU subsystem of AOTS is very robust.

There has been footage/photo
s of DOOM running on an RX 480, some people have claimed that this demo was at 1080p resolution I am under the impression that all DOOM demos run on the RX 480 at Computex were using 1440p VSR on 1080p monitors,am I mistaken?
permalinkembedparent

[–]AMD_RobertTechnical Marketing 104 points 9 hours ago
1080p monitor running at 1440p with VSR.