No DX12 Software is Suitable for Benchmarking *spawn*

[CSI PC]

Is it "broken" or changed? I like my tinfoil hat so I am of the mind that nvidia did it to boost perf where fury x was getting too close. But Oxide are looking into it. I expect nvidia won't fix it till it no longer matters for sure, no matter the reason behind it.

https://www.reddit.com/r/pcgaming/c...ng_the_aots_image_quality_controversy/d3t9ml4



Can expect they will let us know whats going on. Very nice communication in the past

Thanks for the link.
Wonder how 'new' the 'new fp16' pipe is.
Surprised much more was not made about this game using fp16, and makes you wonder if part of this anomaly with 1080FE is due to the crippling of fp16 or the multi-precision fp32 cuda core design that is not playing completely nice with AoTS.

Anyway, makes me laugh because yet again AoTS is being a royal pain to Nvidia.
Probably would had been easier if they just engage with Oxide instead of continually ignoring them
OK the game has not sold much, but oh man it creates a shitstorm of publicity repeatedly for Nvidia.

Cheers

 

[Deleted member 13524]

Maybe it's just time to stop using AotS as the gold standard DX12 benchmark?

You're totally right, we should be using Hitman and TW Warhammer instead:

(That awkward moment when you realize your new $700+ graphics card is barely any better than last year's $500 offerings when using the new API)

 

[Freemantle]

Funny to see Fiji barely faster than Hawaii too.

 

[CSI PC]

You're totally right, we should be using Hitman and TW Warhammer instead:

(That awkward moment when you realize your new $700+ graphics card is barely any better than last year's $500 offerings when using the new API)

Need to be careful with Hitman because Chapter 1 was not optimised in any way for Nvidia.
PCGamesHardware also benchmarked chapter 2 and the performance is way better for Nvidia; in DX11 custom AIB 980ti comfortably ahead of a Fury X, custom AIB 980 just about equal to custom AIB Nano.
In DX12 the gap is closer but again Nvidia still ahead in terms of performance with Chapter 2: more so at 1080 and interestingly this is where DX12 is at its strongest in this game for NVIDIA, AMD falls back again a bit at 4k
http://www.pcgameshardware.de/Hitma...Episode-2-Test-Benchmarks-DirectX-12-1193618/

This goes to show just how much optimisation/designed for multiple architectures plays a part with any of these engines and how it can change between two iterations.
Cheers

 

[Kaotik]

Need to be careful with Hitman because Chapter 1 was not optimised in any way for Nvidia.
PCGamesHardware also benchmarked chapter 2 and the performance is way better for Nvidia; in DX11 custom AIB 980ti comfortably ahead of a Fury X, custom AIB 980 just about equal to custom AIB Nano.
In DX12 the gap is closer but again Nvidia still ahead in terms of performance with Chapter 2: more so at 1080 and interestingly this is where DX12 is at its strongest in this game for NVIDIA, AMD falls back again a bit at 4k
http://www.pcgameshardware.de/Hitma...Episode-2-Test-Benchmarks-DirectX-12-1193618/

This goes to show just how much optimisation/designed for multiple architectures plays a part with any of these engines and how it can change between two iterations.
Cheers

And how exactly are DX11 results in any way relevant for DX12 results?
I don't think anyone is thinking that Furys could offer 1080 performance even in DX12, the question is about the performance delta there

 

[Pressure]

It certainly makes the R9 Nano look great at 175 W and 28 nm.

 

[CSI PC]

And how exactly are DX11 results in any way relevant for DX12 results?
I don't think anyone is thinking that Furys could offer 1080 performance even in DX12, the question is about the performance delta there

You ignored my whole context that Chapter 2 is better optimised for both manufacturers and that Nvidia is actually top performer in both DX11 and DX12.
This is especially important when Nvidia HW also performs really badly in DX11 as it did with Chapter 1.

Putting that aside, as I said even in DX12 Nvidia performance is above that of AMD, more so at 1080p and 4k, at 1440p they are reasonably close.
Anyway Chapter 2 shows that there is a fundamental problem with Chapter 1 that was used in the example I responded to.

I do not know why you cherry picked Fury to 1080, when I explicitely mention 980ti to Fury X....
Look at both DX11 and DX12, and then look at Chapter 1 the other member posted.
http://www.pcgameshardware.de/Hitma...Episode-2-Test-Benchmarks-DirectX-12-1193618/
You need to scroll about half way down.

Anyway Chapter 1 is not really optimised/designed for NV hardware, same could be said about Quantum Break and its post processing.
Cheers

 

[Kaotik]

You ignored my whole context that Chapter 2 is better optimised for both manufacturers and that Nvidia is actually top performer in both DX11 and DX12.
This is especially important when Nvidia HW also performs really badly in DX11 as it did with Chapter 1.

Putting that aside, as I said even in DX12 Nvidia performance is above that of AMD, more so at 1080p and 4k, at 1440p they are reasonably close.
Anyway Chapter 2 shows that there is a fundamental problem with Chapter 1 that was used in the example I responded to.

I do not know why you cherry picked Fury to 1080, when I explicitely mention 980ti to Fury X....
Look at both DX11 and DX12, and then look at Chapter 1 the other member posted.
http://www.pcgameshardware.de/Hitma...Episode-2-Test-Benchmarks-DirectX-12-1193618/
You need to scroll about half way down.

Anyway Chapter 1 is not really optimised/designed for NV hardware, same could be said about Quantum Break and its post processing.
Cheers

Chapter 2 had broken DX12 implementation in those benches, those results are irrelevant

 

[CSI PC]

Chapter 2 had broken DX12 implementation in those benches, those results are irrelevant

Only the memory protection is causing issues with visual quality, and this would NOT had affected the DX11 that also improved.

But you prefer to argue with me about Chapter 2 being the flawed one when comparing NVidia to AMD, rather than Chapter 1 that cripples NVIDIA performance (in BOTH DX11 and DX12 when looking at 970/980) that another forum member posted?

In other words, Chapter 1 is more flawed for Nvidia, which was my point as you need to use both DX11 and DX12 to see the trend.
If you cannot see the issue with Chapter 1, well nothing else I can add.
Edit:
To be fair one may argue that Hitman should NOT be used at all as it may be flawed in different ways with each iteration, so another reason should be ignored in the original post.
Here is Chapter 1 I should had included as a reference: http://www.pcgameshardware.de/Hitman-Spiel-6333/Specials/DirectX-12-Benchmark-Test-1188758/

 

[Kaotik]

Only the memory protection is causing issues with visual quality, and this would NOT had affected the DX11 that also improved.

But you prefer to argue with me about Chapter 2 being the flawed one when comparing NVidia to AMD, rather than Chapter 1 that cripples NVIDIA performance (in BOTH DX11 and DX12 when looking at 970/980) that another forum member posted?

In other words, Chapter 1 is more flawed for Nvidia, which was my point as you need to use both DX11 and DX12 to see the trend.
If you cannot see the issue with Chapter 1, well nothing else I can add.
Edit:
To be fair one may argue that Hitman should NOT be used at all as it may be flawed in different ways with each iteration, so another reason should be ignored in the original post.

It may or may not affect other things, you and I don't know that, when there's benches with fixed version, we can see about it again

 

[Rootax]

Funny to see Fiji barely faster than Hawaii too.

I've a R9 290 , and that's what I'm thinking when i think about upgrading. Sure Fiji is more powerfull, but not by a big gap... Right now I'm playing most of my games nearly everything maxed out (fallout 4, doom 4, witcher3, gtaV, etc) at 1440p without AA. Sometime I'm in the low 30', but with freesync it's "ok".

Now, I want an amd card which can do 1440p WITH smaa 4x at 60fps in most games. Have not that yet, and Hawai is 3 years old... It was a beast, still is, in a way... Aged very well.

 

[Silent_Buddha]

You ignored my whole context that Chapter 2 is better optimised for both manufacturers and that Nvidia is actually top performer in both DX11 and DX12.
This is especially important when Nvidia HW also performs really badly in DX11 as it did with Chapter 1.

That doesn't even make sense. Both are using the same engine. It's the engine that will be optimized. A level may have greater or lesser assets, but is unlikely to have a differently tuned engine. At most it may use some shaders in Chapter 2 that aren't used in Chapter 1, hence why it is not rendering correctly on some cards.

Chapter 1 is just as relevant as Chapter 2. But if some assets in Chapter 2 are broken or not rendering correctly then that renders Chapter 2 irrelevant for whatever hardware isn't rendering it correctly. Just like the GTX 1080 results in AOTS aren't currently valid because the card is incorrectly rendering the scene. Or why the DX12 version of Tomb Raider is irrelevant.

Regards,
SB

 

[PlanarChaos]

why not use chapter 1.5

 

[CSI PC]

That doesn't even make sense. Both are using the same engine. It's the engine that will be optimized. A level may have greater or lesser assets, but is unlikely to have a differently tuned engine. At most it may use some shaders in Chapter 2 that aren't used in Chapter 1, hence why it is not rendering correctly on some cards.

Chapter 1 is just as relevant as Chapter 2. But if some assets in Chapter 2 are broken or not rendering correctly then that renders Chapter 2 irrelevant for whatever hardware isn't rendering it correctly. Just like the GTX 1080 results in AOTS aren't currently valid because the card is incorrectly rendering the scene. Or why the DX12 version of Tomb Raider is irrelevant.

Regards,
SB

I never said it was not the engine that will be optimised.
But it can be modified with the release of Chapter 2 that came awhile later, which it looks like it has, hence why I associate patched/optimised/changed/screwed with the chapter.
TBH it is currently not a good game to use as a comparison, same way no-one would use Quantum Break to show AMD performs better with DX12 as that game also has engine/optimisation issues with regards to post processing and Volumetric Lighting/AO on Nvidia hardware.

Cheers

 

[Psycho]

English forum, anything of note from the video?

It's my impression he only reports based on the public press conference.
He mentions that the AOTS CF test is with both lower power consumption and lower CPU usage than on the 1080 (which could explain the efficiency part - but more likely he just misunderstood the gpu utilization number).
And that the 29th is not just the review date, but also the date it goes on sale.

 

[Deleted member 13524]

OK, so AOTS can't be used because it doesn't look that good and Hitman Ch.1 can't be used because it's not optimized for nvidia enough.

What's the current excuse being used for putting aside TW Warhammer? Does someone in Creative Assembly eat pizza with pineapple?

 

[TomRL]

Is Ashes of the Singularity just super optimised for DX12 or does it lean toward AMD anyway?

 

[Alexko]

OK, so AOTS can't be used because it doesn't look that good and Hitman Ch.1 can't be used because it's not optimized for nvidia enough.

What's the current excuse being used for putting aside TW Warhammer? Does someone in Creative Assembly eat pizza with pineapple?

Now that would be a very valid excuse. F-ing animals!

 

[Kaotik]

Is Ashes of the Singularity just super optimised for DX12 or does it lean toward AMD anyway?

Hard to say, I don't think the game even supports DX11 and while the engine does technically, it was built for Mantle & Future APIs

 

[CSI PC]

OK, so AOTS can't be used because it doesn't look that good and Hitman Ch.1 can't be used because it's not optimized for nvidia enough.

What's the current excuse being used for putting aside TW Warhammer? Does someone in Creative Assembly eat pizza with pineapple?

Well I think it is fair to ridicule anyone who uses the excuse AoTS does not look good as a reason for it to be excluded
You forgot to say Hitman Ch. 2 can't be used because it may not be working correctly with AMD, while it does for Nvidia.

Cheers