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

[PhoBoChai]

Something for the technical minded people to appreciate..

The effect in Dirty Rally & Dirt 4 of checker-boarding on some trees is to simulate transparency during lod transition as frame transitions occurs.

http://www.adriancourreges.com/blog/2015/11/02/gta-v-graphics-study/

So why are some models drawn like this? Could it be to save on fillrate or shading calculation? Not really because GPUs don’t have such granularity: pixels are shaded in square of 2x2, not individually. It’s not about performance, it’s about LOD transition: this dithering pattern makes opaque meshes look a bit transparent when they transition between LODs.
This technique is actually called alpha stippling.

 

[PhoBoChai]

And yes, even at 8X MSAA the problem still exists though only in a minor form.

There is no checker-boarding pattern in 4x or 8x MSAA.

The pattern you're looking for is actually 1 pixel visible, 1 pixel empty, repeated. And yes, it will appear transparent during fast movement or at long distances.

 

[Hitman928]

Something for the technical minded people to appreciate..

The effect in Dirty Rally & Dirt 4 of checker-boarding on some trees is to simulate transparency during lod transition as frame transitions occurs.

So the effect is actually intentional? So does that mean that the game is not being rendered correctly if it's not seen?

 

[dogen1]

So the effect is actually intentional? So does that mean that the game is not being rendered correctly if it's not seen?

Maybe. It's possible the LOD ratios can change based on certain factors and are just visible further out on different cards. The "checkerboard" style lod transition is used by a lot of games.

 

[Muhammed]

Since it's actually load transitions, and since it happens sooner on AMD cards and at closer distances than NVIDIA's. This is a further proof that it actually helps in increasing performance on AMD cards.

 

[Krteq]

This is a further proof that it actually helps in increasing performance on AMD cards.

How? Additional alpha blend testing? It will decrease performance, not increase.

 

[Mr. President]

I'm not sure what the point of this thread is. Just turn off all forms of AA, see who wins, and then revel in your Pyrrhic victory.

It's honestly ridiculous to see GPU benchmarking be reduced to such contingents as AA method and lossy JPEGs. CMAA clearly isn't even antialiasing properly and that just means that these tech 'journalists' aren't doing their jobs.

 

[kawi6rr]

I'm not sure what the point of this thread is.

I sure hope Nvidia is paying these guys because they sure do put in a lot of effort to try to prove Nvidia cards aren't slower then AMD.

off-topic and attacking other members is not allowed
Markfw
Anandtech Moderator

 

[Mr. President]

It works both ways. I follow tech close enough to know that an unnamed poster in this thread owns AMD stock and has a vested interest in bumping them. Which also happens to be highly illegal.

So take everything with a grain of salt is what I'm saying. It's way too easy to pretend being knowledgeable by throwing around big technical sounding words.

off-topic and attacking other members is not allowed
Markfw
Anandtech Moderator

 

[Muhammed]

I'm not sure what the point of this thread is. Just turn off all forms of AA, see who wins, and then revel in your Pyrrhic victory.

I agree, however when one vendor is actually rendering less when AA is off then it isn't a fair comparison.

Some posters here took the Dirt 4 case and added them to the Forza 7 (now defunct) case to conjure up some unsubstantiated theory about Vega cards somehow on a trend to ascention to their mystical level. Whether this ascention is on the heels of reduced image quality or anomaly cases to be fixed through drivers.

A recent developments that comes to mind is the reduction in AF quality on Vega cards in the latest drivers, PCGH found out about the issue and contacted AMD who said that it's a bug to be corrected later.

http://www.pcgameshardware.de/Radeo.../Texturfilter-Bug-in-Radeon-Software-1240201/

 

[PhoBoChai]

It works both ways. I follow tech close enough to know that an unnamed poster in this thread owns AMD stock and has a vested interest in bumping them. Which also happens to be highly illegal.

So take everything with a grain of salt is what I'm saying. It's way too easy to pretend being knowledgeable by throwing around big technical sounding words.

The reason you "know enough" about me, is because I claimed so on the reddit thread when this user under a different name has been posting his false claims there. I was upfront in that I signed up here to address this claim at it's source, under the same username too. Nothing to hide.

I also own NV and Intel stocks among many others. But please, attack the man and not the argument...

off-topic is not allowed
Markfw
Anandtech Moderator

 

[Mr. President]

The reason you "know enough" about me, is because I claimed so on the reddit thread when this user under a different name has been posting his false claims there. I was upfront in that I signed up here to address this claim at it's source, under the same username too. Nothing to hide.

I also own NV and Intel stocks among many others. But please, attack the man and not the argument...

Good man to own up. It's incredibly important to be upfront about these things because guerrilla marketing and stock manipulation are very real things that happen on the Internet.

Keep in mind that we don't know you.

 

[Muhammed]

The reason you "know enough" about me, is because I claimed so on the [/SIZE]reddit thread when this user under a different name has been posting his false claims there.

Yeah false claims alright, the only thing you managed you prove is how AMD is still suffering from image quality problems with CMAA and how the need to test the game with 4XMSAA at least (in which performance is within the expected variances despite AMD's lead) or with 8XMSAA depending on who you ask.

I also own NV and Intel stocks among many others. But please, attack the man and not the argument...

Good God, I really wasn't sure of the rumors about posters owning stock, but man it seems there are a lot of people like you these days. And of courde I perume you own AMD stock as well?!

If you have to ask, no I don't even know how to sign in the stocks markets!

 

[zlatan]

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.

Nope. The rendering is correct. That's just the dithering pattern. And MSAA won't resolve it, just make it less visible. You need to use the advanced blending option, that use OIT, rather than dithering patterning. Also it doesn't matter how visible the pattern is, the calculation will be the same. The driver has a "performance" option in the sample quality setting to make the dithering pattern less visible. This will actually boost the performance, but not much.

 

[zlatan]

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.

This is not a bug, this is how dithering patterning should work. On the other hand the IHV has the right to control the sampling quality. AMD use different settings than NVIDIA. The performance option is the best for the AMD GPUs, to make the dithering pattern less visible. Personally I strongly recommend the advanced blending option in the game. It is much better than the dithering patterning, and todays GPUs are strong enough to use it.

 

[Muhammed]

The driver has a "performance" option in the sample quality setting to make the dithering pattern less visible

Don't you mean the quality or high quality texture filtering options?

 

[PhoBoChai]

This is not a bug, this is how dithering patterning should work. On the other hand the IHV has the right to control the sampling quality. AMD use different settings than NVIDIA. The performance option is the best for the AMD GPUs, to make the dithering pattern less visible. Personally I strongly recommend the advanced blending option in the game. It is much better than the dithering patterning, and todays GPUs are strong enough to use it.

That's what I posted once I found out about that dithering pattern. It's actually the intended effect by the game engine so calling it a bug like Computerbase.de did, which I assumed was right at first, is wrong.

 

[Tup3x]

AMD 8xMSAA = 4xEQ
Nvidia 8xMSAA = 4+4
\
Nvidia = inferior quality 8xMSAA.

http://www.tomshardware.com/reviews/anti-aliasing-nvidia-geforce-amd-radeon,2868-4.html

People cry about how Nvidia has bad AA in some games showing loads of shimmering yet no one does for AMD GPU's, it's one of the reasons.

The current 8xMSAA setting matches the old 8xQ setting. I did the pixel peeping.

I've noticed one bug where in old OpenGL games (actually not sure if this applies to all - Kotor and Kotor2 definitely have this problem) where 8xMSAA looks worse than 4xMSAA. Close to if not worse than 2xMSAA.

 

[stahlhart]

Some of you seriously need to get a life. Thread closed for moderation.

-- stahlhart

 

[ElFenix]

Good man to own up. It's incredibly important to be upfront about these things because guerrilla marketing and stock manipulation are very real things that happen on the Internet.

Keep in mind that we don't know you.

No one posting here can move the price of a publicly traded stock by posting here. This place just isn't that important. So quit witch-hunting.

AT Moderator ElFenix