[AMD_Robert] Concerning the AOTS image quality controversy

[sirmo]

yes and they were offline: https://www.reddit.com/r/Amd/comments/4m692q/concerning_the_aots_image_quality_controversy/d3tvaax

 

[sirmo]

It's a bit embarrassing because you don't seem to know what the word conjecture means.

Looks like I do, yay me

 

[gamervivek]

So the 1080's better showing over the 980Ti/Titan X in DX12 is due to a dodgy driver that skips rendering shaders??

Wow. Cheating busted, where's the tech press up in arms about IHVs cheating with drivers?

*crickets chirping*

That's right. USA tech sites, all bought out. Leave it to EU tech sites to cover this..

Not necessarily. Watch 2:15 onwards.

https://www.youtube.com/watch?v=i9ZG7pA5INc

 

[renderstate]

So they went to buy a gtx 1080, came back, got the GPU out of the box and put it in a new machine and then what? A magical driver they are not supposed to have appeared on their hard drive?

Perhaps NVIDIA erroneously released 368.19 to some partners?

 

[Bacon1]

I love how this has turned from IQ was incorrect due to rendering issue because of Nvidia's drivers into a conspiracy around how AMD got the bad rendering drivers in the first place. If there is a few % difference in performance no one would have noticed and all the reviews would have been done with the bad driver.

But because people were so die hard on making AMD look bad because they thought that AMD was rendering incorrectly they are spinning the issue to once again try to make AMD look bad.

 

[Vaporizer]

Ok. No problem. Calm down. It was a simple lack of communication between drivers group and sales team.
Anyway, you have to understand that it is actually a "feature" for the greater good of gamers.

 

[3DVagabond]

Still waiting for your arguments. Do you have any evidence that contradicts what he says?

Yeah. He called him out for misquoting right away. That was the only evidence he presented. So post is invalid. The rest is just hyperbole and spin. Comes under the heading of "anything is possible" seeing as how there is zero evidence showing they weren't cheating in the reply.

 

[kraatus77]

And these apologists appeared when new gpus released. how convenient. i've seen new lows of an apologistic fanboy today. anything to defend nvidia. lol

 

[3DVagabond]

Can you explain where AMD got that press-only driver? Makes us wonder where that Geforce GTX 1080 came from, if a hardware website 'close' to them provided a review sample.

What difference does where they got the driver make?

 

[Bacon1]

Can you explain where AMD got that press-only driver? Makes us wonder where that Geforce GTX 1080 came from, if a hardware website 'close' to them provided a review sample.

Let me guess, you'd defend someone caught cheating by asking why the person came home early from work?

Nvidia's drivers are causing issues with incorrect rendering in games. They got busted because people cared to look at the images and not just the benchmark results.

Instead of asking Nvidia why the images don't look right, you are asking why AMD was able to use the drivers.

Stop trying to detract from the main issue here: Nvidia's card isn't outputting the correct image

 

[Magee_MC]

Can you explain where AMD got that press-only driver? Makes us wonder where that Geforce GTX 1080 came from, if a hardware website 'close' to them provided a review sample.

Does where AMD got it from have any relevance at all? It was the driver that NV released in order to benchmark the 1080. Once the NDA was lifted, I can't see that there was any reason not for AMD to have it since that was the driver that NV gave for reviews of their card. AMD was comparing the 480 to what had been reviewed in the press.

As for where they got it. Well, it was released to the public days before AMD's presentation at Computex. Instead of your insinuation that it was from some nefarious source, how about the simplest answer. They bought a few from a store.

 

[Sweepr]

Does where AMD got it from have any relevance at all? It was the driver that NV released in order to benchmark the 1080. Once the NDA was lifted, I can't see that there was any reason not for AMD to have it since that was the driver that NV gave for reviews of their card. AMD was comparing the 480 to what had been reviewed in the press.

As for where they got it. Well, it was released to the public days before AMD's presentation at Computex. Instead of your insinuation that it was from some nefarious source, how about the simplest answer. They bought a few from a store.

Retail comes with a newer version, so no. Where they got it? Was it from someone who received a card from NVIDIA to review? They didn't even bother to update it.

https://twitter.com/dankbaker/status/739880981612625920

Yeah. He called him out for misquoting right away. That was the only evidence he presented. So post is invalid.

No it's not, but let's pretend it is. Lots of valid points in his posts.

 

[Yakk]

Nvidia going back to lowering their Imagw Quality and raise fps for reviews again?!

Been a while... what's old is new again it seems.

Needs some more testing!

 

[Magee_MC]

Retail comes with a newer version, so no. Where they got it? Was it from someone who received a card from NVIDIA to review? They didn't even bother to update it.

That's one possibility. Another is that they got it from a member of the press since you can do a fresh install of any previous driver.

You are making accusations with absolutely zero evidence. When other possibilities are presented, you double down on your theories.

Once again, let me ask. Does wherever AMD got the driver from have any relevance to the benchmark comparisons that they did? Was it in any way unfair for AMD to use the driver that NV gave to the press in order to benchmark the 1080 in order to benchmark the 480 against the 1080?

Conspiracy theories aside, was AMD's comparison of the 480 in CF against the 1080 in ANY way an unfair benchmark?

 

[Sweepr]

Once again, let me ask. Does wherever AMD got the driver from have any relevance to the benchmark comparisons that they did? Was it in any way unfair for AMD to use the driver that NV gave to the press in order to benchmark the 1080 in order to benchmark the 480 against the 1080?

I don't know, are there any comparisons out? The fact that they weren't using up to date drivers tells they weren't paying too much attention at the NVIDIA system.

Conspiracy theories aside, was AMD's comparison of the 480 in CF against the 1080 in ANY way an unfair benchmark?

I wouldn't call it unfair, but considering how everybody knows this title favours their hardware and provides better than average multi-GPU scaling, it was very likely a best case scenario. Doesn't even make sense, they missed the opportunity to highlight the perf/$ by comparing it to Geforce GTX 960/970 instead.

 

[3DVagabond]

I wouldn't call it unfair, but considering how everybody knows this title favours their hardware and provides better than average multi-GPU scaling, it was very likely a best case scenario. Doesn't even make sense, they missed the opportunity to make it look like a great perf/$ by comparing it to Geforce GTX 960/970.

What's known is it's a title that both vendors are allowed to optimize for their hardware. Seems like best case would be a title that was only optimized for AMD.

 

[Abwx]

I don't know, are there any comparisons out? The fact that they weren't using up to date drivers tells they weren't paying too much attention at the NVIDIA system.

Lol...

So Nvidia sending this driver to the press means that they didnt pay much
attention to their own product systems..????..

To summarize AMD should pay more attention to Nvidia system than Nvidia themselves, that s what you are stating actually, re lol...

 

[littleg]

OK, so AMD got the driver from some site that had it. Great, that's settled.

Now what about those image quality shenanigans? What we really need to know is whether there's actually any performance difference and, if so, is it noticeable. Some hardware site needs to get this benched so we can put this matter to bed.

 

[Dresdenboy]

Retail comes with a newer version, so no. Where they got it? Was it from someone who received a card from NVIDIA to review? They didn't even bother to update it.

https://twitter.com/dankbaker/status/739880981612625920



No it's not, but let's pretend it is. Lots of valid points in his posts.

Did AMD actually spend time to analyze IQ while being interested in performance first? Or did they notice only after the fact appeared in discussion platforms?

 

[Magee_MC]

I don't know, are there any comparisons out? The fact that they weren't using up to date drivers tells they weren't paying too much attention at the NVIDIA system.

Once again, that's one possibility. Another is that they wanted to benchmark against the driver that was used in every single press review of the 1080 cards that they gave to the media. I've looked and I haven't found any updated comparisons yet, but given the amount of publicity that this issue is receiving, I can't help but expect that somebody will do them. Until we have those, we don't know the definitive answer. My guess is that that the difference in the driver performance will be negligible.

I wouldn't call it unfair, but considering how everybody knows this title favours their hardware and provides better than average multi-GPU scaling, it was very likely a best case scenario. Doesn't even make sense, they missed the opportunity to highlight the perf/$ by comparing it to Geforce GTX 960/970 instead.

I can't see how it's a best case scenario. AMD does do better on AOTS than NV, but from everything that I've read, that's not because the game is tilted to AMD, but because DX12 allows AMD's hardware to unlock it's full potential in a way that wasn't available in DX11. Additionally, NV had full access to the source code and was able to optimize it for their hardware. If anything, since both AMD and NV have been able to see and modify source code to optimize performance on their hardware since at least back to the Beta version, I tend to think of this as pretty much a good apples to apples comparison for what is possible on both brands of hardware under DX12.

 

[IllogicalGlory]

I can't see how it's a best case scenario. AMD does do better on AOTS than NV, but from everything that I've read, that's not because the game is tilted to AMD, but because DX12 allows AMD's hardware to unlock it's full potential in a way that wasn't available in DX11. Additionally, NV had full access to the source code and was able to optimize it for their hardware. If anything, since both AMD and NV have been able to see and modify source code to optimize performance on their hardware since at least back to the Beta version, I tend to think of this as pretty much a good apples to apples comparison for what is possible on both brands of hardware under DX12.

It may be capable in AotS, but that might have been Oxide's intention. "AMD can do AC, NV can't, so we'll put that in the game". Not necessarily true, but it is possible in a GE title.

On the other hand, this game seems to make good use of AC, so maybe they just wanted to test it out.

 

[Bacon1]

It may be capable in AotS, but that might have been Oxide's intention. "AMD can do AC, NV can't, so we'll put that in the game". Not necessarily true, but it is possible in a GE title.

On the other hand, this game seems to make good use of AC, so maybe they just wanted to test it out.

The developers have specifically said multiple times that they not only worked closer with Nvidia than AMD, but that AMD never requested them use any features:

Saying we heavily rely on async compute is a pretty big stretch. We spent a grand total of maybe 5 days on Async Shader support. It essentially entailed moving some ( a grand total of 4, IIRC) compute jobs from the graphics queue to the compute queue and setting up the dependencies. Async compute wasn't available when we began architecting (is that a word?) the engine, so it just wasn't an option to build around even if we wanted to. I'm not sure where this myth is coming from that we architected around Async compute. Not to say you couldn't do such a thing, and it might be a really interesting design, but it's not OUR current design.

Saying that Multi-Engine (aka Async Compute) is the root of performance increases on Ashes between DX11 to DX12 on AMD is definitely not true. Most of the performance gains in AMDs case are due to CPU driver head reductions. Async is a modest perf increase relative to that. Weirdly, though there is a marketing deal on Ashes with AMD, they never did ask us to use async compute. Since it was part of D3D12, we just decided to give it a whirl.

http://www.overclock.net/t/1575638/...able-legends-dx12-benchmark/110#post_24475280
Async compute is a big feature in DX12, its one of the reason to use DX12 in the first place. There are only a few architecture differences in DX11 and DX12 and Async compute and MGPU are the two major ones. Everything else you can already do in DX11.

Can we please get back on topic: Nvidia was caught with a driver that removed IQ from the game. So far they haven't apologised and instead are trying to blaming AMD for using it because they weren't supposed to have a copy of it.

 

[Elixer]

Can you explain where AMD got that press-only driver? Makes us wonder where that Geforce GTX 1080 came from, if a hardware website 'close' to them provided a review sample.

What exactly is wrong with this though?
If nvidia supplied those drivers to the press, and, the press used those drivers for all their reviews, then why is it foul for AMD to use those same drivers?

Conspiracy people would be saying that nvidia rigged those drivers to eek out extra performance for the launch.

Sane people would just look and see that everyone was using the same drivers.

If the issue is now fixed, I am sure that all those sites will now update all their reviews with the fixed drivers, right?

 

[Erenhardt]

I love how this has turned from IQ was incorrect due to rendering issue because of Nvidia's drivers into a conspiracy around how AMD got the bad rendering drivers in the first place. If there is a few % difference in performance no one would have noticed and all the reviews would have been done with the bad driver.

But because people were so die hard on making AMD look bad because they thought that AMD was rendering incorrectly they are spinning the issue to once again try to make AMD look bad.

QFT! It all started when some thrown rocks at amd for not rendering AOTS properly. Then it turner out to be nv not rendering properly. Perfect 180'. Ofcourse the same people trowing stones at amd are now hiding in their glasshouses downplaying the issue.

What the difference where did amd get those drivers. They could hack into nv datacenter for all I care. What is more important is why nv is sending one drivers for testing performance with results published in the press, while the consumer has no access to those drivers.

If the performance between consumer drivers and press drivers is not the same, we have an issue of false advertising.

 

[Magee_MC]

It may be capable in AotS, but that might have been Oxide's intention. "AMD can do AC, NV can't, so we'll put that in the game". Not necessarily true, but it is possible in a GE title.

On the other hand, this game seems to make good use of AC, so maybe they just wanted to test it out.

I don't think that was the intention. Remember, when it first came out that using AC significantly hurt NV's performance everybody was surprised since NV had said that Maxwell could do it and they exposed AC in their drivers.

Also, AMD may have had a marketing agreement with Stardock, however NV had complete access to the AOTS source code and could optimize it however they wanted as long as it didn't hurt AMD's performance.