[AMD_Robert] Concerning the AOTS image quality controversy

[itsmydamnation]

i think your missing his point which is:
http://www.ashesofthesingularity.co...-details/8b748568-fc96-4e48-9fed-22666a7149f5

i would have to test on my 290 see how much perf loss there is between crazy and extreme. Also what is your cpu clocked at? ifs its greater then stock that could be an FPS or two advantage vs 3.5 6 core ( i tested CPU scaling and it isn't linear).

 

[Bacon1]

I wish. That was the Doom demo.

Obviously, wait for reviews, but this demo should be showing the 480 at its peak. Maybe it's a compute performance bottleneck or something, but...

Still, maybe this thing's crazy efficient, but I have worries about performance compared to die size and potential unused throughput in Vega chips. Or maybe the clocks are really low and increasing them would kill the efficiency.

In your post you quoted him saying that moving upto 1440p was a change of +/- 1%

Moving up to 1440p at the same settings maintains the same performance delta within +/-1%.

Which means that we can basically say it was running @ 1440p since it gains or loses by 1% (CPU bound).

 

[IllogicalGlory]

i think your missing his point which is:
http://www.ashesofthesingularity.co...-details/8b748568-fc96-4e48-9fed-22666a7149f5

i would have to test on my 290 see how much perf loss there is between crazy and extreme. Also what is your cpu clocked at? ifs its greater then stock that could be an FPS or two advantage vs 3.5 6 core ( i tested CPU scaling and it isn't linear).

How am I missing the point? I tested my GPU at the same resolution and preset as both cited 480 tests (Crazy 1080p and Extreme 1440p).

The performance loss between presets is quite large, as you can see performance at 1440p Extreme is higher than performance at 1080p Crazy. 2560x1080 Extreme was the preset I liked to use while overclocked when playing the game.

I know that result you linked, and those are the settings I tested in my last post, and the performance difference is still almost nothing, against a stock 290 no less.

My CPU is at 4.7GHz, but it looks like there's plenty of headroom for my chip and AMD's chip anyway.

Which means that we can basically say it was running @ 1440p since it gains or loses by 1% (CPU bound).

But it wasn't, and while the comparison to the 1080 doesn't change, the comparison to my 290 would. However, it's running at 1080p and I tested it at the same settings, so it's a valid comparison - as is my 1440p test to the one in the benchmark database. My extrapolation of its single GPU performance is based on official information from AMD too.

 

[Armsdealer]

How am I missing the point? I tested my GPU at the same resolution and preset as both cited 480 tests (Crazy 1080p and Extreme 1440p).

I know that result he's referring to and my last test used the same settings.

My CPU is at 4.7GHz, but it looks like there's plenty of headroom for my chip and AMD's chip anyway.

I see your point and it's certainly a valid one, but your cpu at least explains part it:

Your r9 290 on a 4.7 ghz i7 is similar in performance to a c7 on a 3.5 ghz i7 in AotS 1440 extreme in a game that has high cpu requirements.

I hardly think it's a bad thing. Keep in mind even in 3dmark dx11 the 1080 is only
75% faster than a 390. In dx12 it's probably going to be around 60%. I think people are backward calculating using what is almost certainly a bad cf scaling factor: the c7 probably will have scored around 41 fps on the AMD test system 1080 crazy.

Edit: based on your results I'm probably wrong on that 41 figure. Your argument holds more water than I initially gave it credit for...

 

[3DVagabond]

Nvidia needs to be pushed to fix this rendering error on the 1080 then. Every review I saw of the 1080 included AOTS and made a big to do about the performance increase in that title over the 980ti/Titan X. No doubt it was part of the reviewer's guide for the card.

There may be a small to negligible difference with this fixed, or there may be a difference worth noting, making reviews of the card using results of this game incorrect.

what really annoys me Groove is that nobody mentioned it. Nobody!

Remember the "texture shimmering" episode on one older game (don't recall the game ATM) for AMD? Remember 290 throttling? Remember 290 reference cards performing worse than the review models (because of poor heatsink assembly, IIRC). And of course the dropped frames in Crossfire? Everyone was reporting, testing, buying retail cards, and on and on.

Are we to believe they are so thorough with AMD reviews but simply miss all this stuff with nVidia?

 

[Bacon1]

But it wasn't, and while the comparison to the 1080 doesn't change, the comparison to my 290 would. However, it's running at 1080p and I tested it at the same settings, so it's a valid comparison - as is my 1440p test to the one in the benchmark database. My extrapolation of its single GPU performance is based on official information from AMD too.

But we know that performance was +/- 1% which means the #s would be the same for 1440p

 

[Hitman928]

But we know that performance was +/- 1% which means the #s would be the same for 1440p

Robert said the performance difference between the 1080 and 480 cf was +/- 1% when moving to 1440p, not that the absolute performance of the cards didn't change.

 

[ul256]

This is actually not quite true. My initial run was using 4X MSAA, not 8X. Still the difference is only 1 FPS.

XFX 290 Double Dissipation at 947/1250, 1080p Crazy + 8X MSAA (matching the settings on the Polaris demo).

Single GPU RX 480 ~ 34.1 FPS.

Just wanted to point out one thing. I noticed that AMD never said what their settings for point light quality, glare quality, and shadow quality were in the benchmark. I see that you are using the high setting. Could you try using the ultra setting on these options to see if the results change? thanks

 

[kagui]

i dont understand what you guys are discussing, under the same settings the 1080 is 98% and dual 480 is 51% , but it depends on the game , will you get the same result on fallout no, will you get the result on newer games yes, as long as the dev implement it, and if you dont like the result you can buy AMD or NVIDIA and running it on crossli , as long as both card's are of similar performance

 

[ShintaiDK]

i dont understand what you guys are discussing, under the same settings the 1080 is 98% and dual 480 is 51% , but it depends on the game , will you get the same result on fallout no, will you get the result on newer games yes, as long as the dev implement it, and if you dont like the result you can buy AMD or NVIDIA and running it on crossli , as long as both card's are of similar performance

Did you read where the 51% number comes from?

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%

Dont let yourself fool from PR. They picked their cherry sponsored AOTS for the sole reason that its the only place its happening and that they can select this single number out of 3

 

[Layer8]

They picked their cherry sponsored AOTS for the sole reason that its the only place its happening and that they can select this single number out of 3

Do you have any sources for it being sponsored?
I do only find that they treat all GPU providers the same:
http://www.oxidegames.com/2015/08/16/the-birth-of-a-new-api/

The answer is that we have an open access policy. Our goal is to make our game run as fast as possible on everyone’s machine, regardless of what hardware our players have.

To this end, we have made our source code available to Microsoft, Nvidia, AMD and Intel for over a year. We have received a huge amount of feedback. For example, when Nvidia noticed that a specific shader was taking a particularly long time on their hardware, they offered an optimized shader that made things faster which we integrated into our code.

 

[Flapdrol1337]

Do you have any sources for it being sponsored?
I do only find that they treat all GPU providers the same:
http://www.oxidegames.com/2015/08/16/the-birth-of-a-new-api/

Go to the ashes of the singularity website, scroll down, observe amd logo.

 

[ShintaiDK]

Do you have any sources for it being sponsored?

Its an AMD Evolved Game.

 

[sirmo]

Its an AMD Evolved Game.

Saying it's an AMD evolved game [which it isn't] is implying that there are some shenanigans at play here, when in reality there are none. Nvidia was given access to the source code of the game. Stop spreading FUD.

Virtually all DX12 games have shown a strong propensity to run better on GCN.

 

[flopper]

Its an AMD Evolved Game.

Nvidia cant do true async compute, they had years to prepare and Pascal cant do it.
cant blame someone else for Nvidias incomptence there.

 

[beginner99]

I have to agree. Looks like one of those RTS games where you play it once or twice or never? Its like one big DX12 tech demo that has a bleak future because successful franchises like Totalwar, SC have their own engines that already do pretty damn well on top of the entire genre becoming a rarity these days. The former has a pretty good DX12 implementation already and SC2 for instance caters toward even the lowest end rigs (the gameplay speaks for itself without the need of all that fancy DX12 effects etc -

You are absolutely wrong. I haven't understood that dx12 is about CPU limits and not about effects.

SC2 would greatly benefit from dx12 and in general much better multi threading. it's basically same fail as SC1 was. Try playing 4v4 on money maps. it's CPU bound all the way. Original Starcraft had a min requirement of pentium 1 90 mhz but needed a PIII to actually play 4v4 without lag.

The unit limit of 200 is mainly also there due to CPU limits. Because it's single-threaded. dx12 in AoTS is all about CPU limits and not fancy dx12 effects.

 

[Galatian]

Dont let yourself fool from PR. They picked their cherry sponsored AOTS for the sole reason that its the only place its happening and that they can select this single number out of 3

What single number? Do you even know what they mean? "Single Batch GPU Util" is a CPU bound test and the lower number on AMD system actually means that the Polaris chips can't stretch their feet because of the CPU. Polaris CF will before even better on a OCed CPU.

That being said as a owner of two R9 290X, which only cost me 300€ each, when they were sold off in 2014, I'm in an awkward position. There is still no clear upgrade path and Polaris seems to bring similar performance for a similar or slightly lower price. Also power consumption is heavily reduced. Which is of course nice as a replacement for the R9 x90 lineup, but it's not like you were not able to have had the horse power for the same price years ago.

All this is showing is that the R9 290X aged very well and was a good card, especially when you purchased it during the sale.

 

[96Firebird]

This is obviously a best case scenario game for AMD to showcase, no doubt about that. The interpolations made from these numbers are comical and will not reflect in real world reviews. But then posters here will just blame the review sites for benchmarking wrong, as usual.

 

[monstercameron]

This is obviously a best case scenario game for AMD to showcase, no doubt about that. The interpolations made from these numbers are comical and will not reflect in real world reviews. But then posters here will just blame the review sites for benchmarking wrong, as usual.

So many words, and nothing said. As always riveting meta commentary. :thumbsdown:

 

[Krteq]

Wow

We (Stardock/Oxide) are looking into whether someone has reduced the precision on the new FP16 pipe.
Both AMD and NV have access to the Ashes source base. Once we obtain the new cards, we can evaluate whether someone is trying to modify the game behavior (i.e. reduce visual quality) to get a higher score.
In the meantime, taking side by side screenshots and videos will help ensure GPU developers are dissuaded from trying to boost numbers at the cost of visuals.

So, this happened right after nVidia launched their first cards with mixed precision (FP16/FP32/FP64) support, coincidence?

AMD supports mixed precision since GCN Gen. 3 (Tonga/Fiji), Pascal can manage FP16 in 2:1 ration of FP32, GCN Gen 3 in 1:1 only. So who will benefit more by this?

reddit - Concerning the AOTS image quality controversy

 

[renderstate]

So many words, and nothing said. As always riveting meta commentary. :thumbsdown:

Actually he/she said something that makes completely sense. Let me add that if you are presenting a new product and to demonstrate its performance you decide to showcase it in a contrived way in an application everyone knows heavily favors it showing numbers that make no sense..well, we immediately know you are in trouble. We don't know what the problem is yet but it's a matter of time.

The (very well welcomed) super low price only confirms that.

 

[swilli89]

SC2 would greatly benefit from dx12 and in general much better multi threading. it's basically same fail as SC1 was. Try playing 4v4 on money maps. it's CPU bound all the way. Original Starcraft had a min requirement of pentium 1 90 mhz but needed a PIII to actually play 4v4 without lag.

The unit limit of 200 is mainly also there due to CPU limits. Because it's single-threaded. dx12 in AoTS is all about CPU limits and not fancy dx12 effects.

You're completely remembering it wrong. SC and SC: Broodwar ran like a champ and was hindered by network performance on 4v4 not CPU. I used to LAN Broodwar and play 8 player battles and it ran beautifully.

 

[Sweepr]

This is obviously a best case scenario game for AMD to showcase, no doubt about that. The interpolations made from these numbers are comical and will not reflect in real world reviews. But then posters here will just blame the review sites for benchmarking wrong, as usual.

Precisely. Would be the same if NVIDIA did a demo of Pascal vs AMD running Rise of the Tomb Raider or another title that favours their cards. Even under this best case scenario the best they came up with was a dual vs single GPU 'efficiency' comparison, meh.

 

[boozzer]

Precisely. Would be the same if NVIDIA did a demo of Pascal vs AMD running Rise of the Tomb Raider or another title that favours their cards. Even under this best case scenario the best they came up with was a dual vs single GPU 'efficiency' comparison, meh.

and it is still 40% cheaper meh.

 

[Glo.]

Wow
So, this happened right after nVidia launched their first cards with mixed precision (FP16/FP32/FP64) support, coincidence?

AMD supports mixed precision since GCN Gen. 3 (Tonga/Fiji), Pascal can manage FP16 in 2:1 ration of FP32, GCN Gen 3 in 1:1 only. So who will benefit more by this?

reddit - Concerning the AOTS image quality controversy

http://semiaccurate.com/forums/showthread.php?p=263871#post263871

I think you should read this post from Fottemberg, about FP16 on Pascal GPUs.

The post is from 29th may. Few days before the AMD presentation, and the benchmarks. We have to wait for effects of investigation by Oxide to conclude anything.