[ComputerBase.de] Possible reason of Dirt 4 favoring AMD GPUs with CMAA

Status
Not open for further replies.

Muhammed

Senior member
Jul 8, 2009
453
199
116
We all have seen how AMD cards are so much faster than NVIDIA counterparts in Dirt 4, specifically when using CMAA as the AA method. While the situation becomes reversed with the use of MSAA.

I think the reason is revealed by the testing done by ComputerBase.de, they discovered that treed don't render correctly with AMD cards and CMAA, they a checkerboard/grid pattern appearance. The issue only resolves itself under high level MSAA like 8X MSAA, and NVIDIA cards don't suffer the same problem.

CMAA: (click to enlarge, look at the tree to the right of the center road sign)


8X MSAA: (click to enlarge, look at the tree to the right of the center road sign)


https://www.computerbase.de/2017-06/dirt-4-benchmark/#bilder

This is what causes the huge disparity in performance under CMAA in AMD's favor. Using 8X MSAA reverses the picture and NVIDIA cards are now faster. Here the GTX 1080 is always faster than Vega 64:

https://www.hardocp.com/article/2017/08/14/amd_radeon_rx_vega_64_video_card_review/12
http://www.babeltechreviews.com/rx-...e-overclocking-showdown-vs-the-gtx-1080-fe/3/


With 4XMSAA, Vega 64 is just 15% faster than 1080, according to PCGH.
http://www.pcgameshardware.de/Radeo.../Tests/Benchmark-Vergleich-vs-Nvidia-1235712/
 
Last edited:

PhoBoChai

Member
Oct 10, 2017
119
389
106
Source: https://www.computerbase.de/2017-06/dirt-4-benchmark/

OP, you're missing the full context and analyzed the issue wrong.

This is a known problem with the EGO engine version 3.0. While Dirt Rally has still provided some huge forests with this problem, there are only very few trees in Dirt 4 that are affected by the problem.

The problem exists in Dirt Rally as well, and in there, a lot more trees are affected by this bug, yet AMD GPUs do not run faster than NV, they tend to run worse in fact.

There is a major issue with performance loss on AMD GPUS with 8x MSAA, it is not just affecting RX Vega, but all the other cards.

https://www.hardocp.com/article/2017/08/08/dirt_4_real_gameplay_video_card_performance_review/10

No AA vs CMAA: Radeons take a very small hit.



>CMAA is only 2% from No AA here, hardly any difference.


No AA vs 8x MSAA: Radeons take a huge hit.



>No AA is 74% faster than 8X MSAA, again showing 8X MSAA is a bigger burden on Polaris.

You can look at GTX and CMAA & MSAA effects on it as well:

https://www.hardocp.com/article/2017/08/08/dirt_4_real_gameplay_video_card_performance_review/6

Basically CMAA runs identical performance within margin of errors on Radeons or GTX, while 8x MSAA runs much worse on AMD than NV (both lose perf big time though).

There is a graphics glitch with CMAA on some trees on Radeons. This bug exists in the engine since Dirt Rally. Nobody at Codemasters must have noticed.

The speculation in the OP is therefore proven wrong, CMAA is not the reason why AMD GPUs run Dirt 4 better because CMAA does not drop performance much (2-3%) for there to be a major difference between NV or AMD GPUs.

AMD GPUs run it better with No AA, CMAA, MSAA 2-4x, while they run it worse with MSAA 8x. It is not an RX Vega specific issue. It seems to be an optimization issue across the board for AMD GPUs with 8x MSAA.
 

Mopetar

Diamond Member
Jan 31, 2011
8,016
6,468
136
Yeah, I'm not sure what the hell I'm supposed to be looking for either. I swear that you have to be autistic to spot some of this crap. I feel stupid for spending two minutes just trying to look for something at all.

Also, the two pictures don't seem to be completely aligned either, so of course they'll look different. The HUD on the far left is overlaid on a completely different tree in both photos.
 

Muhammed

Senior member
Jul 8, 2009
453
199
116
You guys really really lack any sort of attention if you can't spot the difference, just enlarge the first picture, and look at the tress to the right, the most center tree appears pixelated like this:

 

Muhammed

Senior member
Jul 8, 2009
453
199
116
AMD GPUs run it better with No AA, CMAA, MSAA 2-4x, while they run it worse with MSAA 8x. It is not an RX Vega specific issue. It seems to be an optimization issue across the board for AMD GPUs with 8x MSAA.
I don't see it as an optimization issue, at 8X MSAA the IQ bug disappears thus causing performance drops. The reason AMD GPUs run the game so much better with no AA, CMAA, and 2X/4X MSAA is because the IQ issue is still there, and tress aren't rendered correctly which makes the Radeons run the game faster. At 8X MSAA they render correctly slowing the performance down.
 

UglyDuckling

Senior member
May 6, 2015
390
35
61
You guys really really lack any sort of attention if you can't spot the difference, just enlarge the first picture, and look at the tress to the right, the most center tree appears pixelated like this:


No offense but some of us don't put our faces right up to our screens to notice such small differences, especially when our monitors are literally only 21 inches (high 1080P PPI) or we use a TV.
 

Muhammed

Senior member
Jul 8, 2009
453
199
116

UglyDuckling

Senior member
May 6, 2015
390
35
61
Your link is very old, it's comparing CSAA to EQAA, this has nothing to do with 8XMSAA. And CSAA mode is now defunct on Maxwell and Pascal.


These AA methods still exist on Maxwell and Pascal and are still in use when a game uses MSAA.
 

UglyDuckling

Senior member
May 6, 2015
390
35
61

Besides that you can still use it, it's just hidden from the normal NVCP..

The thing i was showing was that 8xMSAA on Nvidia which is labelled (8xMSAA) is not the same as AMD's and not up to the same quality either which is why GeForce does better with "8xMSAA" people see the name for the method used and some how think they are the same.

Side by side, AMD's 8xMSAA solution does in fact look better, even in YT videos showing it.

I think the only recent game to show 8xMSAA looking VERY similar was Half Life 2.


"MSAA hasn't ever gone up to 16x unless you're running SLI. The problem is that NVidia lumped CSAA and MSAA together without making any real distinction between them - a practice I don't agree with, but one which technically isn't incorrect. In the control panel, with the old card, you had two different sets of options:

2x, 4x and 8xQ: these are all "pure" MSAA.
8x, 16x, 16xQ and 32x: these all use MSAA in tandem with CSAA."



There are of course other things..




"Totally forgot about this but I'm still a little annoyed about the missing 16xq, you say it should looks as good as 8xMSAA but I'm 100% convinced it doesn't. playing games like mirror's edge or cs go I remember not seeing a jagged edge in sight with 16xaa but with 8x it's pretty noticeable.

It's a shame maxwell doesn't support those options, I've used DSR before but it's not ideal, pretty big performance hit and the HUD often scales pretty poorly in some games. 16x seems to be my favorite AA option but I can't use it anymore!"
 
Last edited:
Reactions: Yakk

nurturedhate

Golden Member
Aug 27, 2011
1,762
761
136
You guys really really lack any sort of attention if you can't spot the difference, just enlarge the first picture, and look at the tress to the right, the most center tree appears pixelated like this:

Both screenshots exhibit the same issue, though the top one is worse. Also, as pointed out, you should have actually read the linked article as it told you why there's a difference. Also, the rest of everyone else's posts. Furthermore it seems your only purpose is to flame AMD at every turn. No basis, no objectivity. No nothing. Just trolling. Post after post. At least make them entertaining if that is how you wish to spend your time.
 

Muhammed

Senior member
Jul 8, 2009
453
199
116
Both screenshots exhibit the same issue, though the top one is worse. Also, as pointed out, you should have actually read the linked article as it told you why there's a difference.
Yeah the difference is a bug in the engine, we already know that. That bug is also responsible for increased performance on AMD GPUS in Dirt Rally too:

http://www.legitreviews.com/dirt-ra...geforce-gtx-970-versus-radeon-r9-390_167054/4

Furthermore it seems your only purpose is to flame AMD at every turn. No basis, no objectivity. No nothing. Just trolling. Post after post. At least make them entertaining if that is how you wish to spend your time.
Nice trolling try, just needs a little effort, if you need help try some online course! As for the basis, there are pictures, articles, and data to back it up, while all you have is trolling!
 

PhoBoChai

Member
Oct 10, 2017
119
389
106
I don't see it as an optimization issue, at 8X MSAA the IQ bug disappears thus causing performance drops. The reason AMD GPUs run the game so much better with no AA, CMAA, and 2X/4X MSAA is because the IQ issue is still there, and tress aren't rendered correctly which makes the Radeons run the game faster. At 8X MSAA they render correctly slowing the performance down.

The source you quoted was very specific on it.

>As long as no MSAA is used as edge smoothing, the trees on Radeon graphics cards are not displayed correctly and are provided with a grid-like pattern.

There is no visual difference between the tree with 4x MSAA vs 8x MSAA. Your implied 8x MSAA being massively slower because one tree in the forest isn't rendered properly is preposterous. No trees are rendered incorrect with 4x MSAA, yet Radeons are still faster.

Again, to clarify:

>While Dirt Rally has still provided some huge forests with this problem, there are only very few trees in Dirt 4 that are affected by the problem.
 

PhoBoChai

Member
Oct 10, 2017
119
389
106
This is 8x MSAA



4x MSAA:




CMAA:



The problem is non-existant with 4x MSAA. Yet Radeons are still faster.
 
Reactions: ZGR

amenx

Diamond Member
Dec 17, 2004
4,011
2,279
136
There are of course other things..

You can always find the odd anomaly with a certain game or two that affects one card vs the other. Always funny to see observations taken from one old test and not followed up with subsequent tests with newer drivers (and/or other cards) taken as absolutes that hold forever against one brand vs the other. Had to chuckle on the other link from 2011.. as if cards with much older arches handle the different forms of AA in the same exact manner as cards in 2017.
 
Reactions: Muhammed

UglyDuckling

Senior member
May 6, 2015
390
35
61
You can always find the odd anomaly with a certain game or two that affects one card vs the other. Always funny to see observations taken from one old test and not followed up with subsequent tests with newer drivers (and/or other cards) taken as absolutes that hold forever against one brand vs the other. Had to chuckle on the other link from 2011.. as if cards with much older arches handle the different forms of AA in the same exact manner as cards in 2017.

MSAA has never changed since ?? lol.
 

Glo.

Diamond Member
Apr 25, 2015
5,765
4,669
136
Both screenshots exhibit the same issue, though the top one is worse. Also, as pointed out, you should have actually read the linked article as it told you why there's a difference. Also, the rest of everyone else's posts. Furthermore it seems your only purpose is to flame AMD at every turn. No basis, no objectivity. No nothing. Just trolling. Post after post. At least make them entertaining if that is how you wish to spend your time.
Thank you.

It would be funny, if both screenshots were taken using Nvidia GPU.
 

Muhammed

Senior member
Jul 8, 2009
453
199
116
It would be funny, if both screenshots were taken using Nvidia GPU.
How cute, if you bothered to read the article, it clearly states AMD's IQ issue with Dirt 4, whicb presisted in Dirt Rally as well, takes half a dozen pictures of it, and proceeds to explain it in detail.

And yes, even at 8X MSAA the problem still exists though only in a minor form.
 
Status
Not open for further replies.
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |