x800 texture shimmering: FarCry video

[jvd]

Okay . I cna't get to that board and i'm having a ton of problems with fraps .

www.ooakalabeth.com/video

its being uploaded right now. But whenever i turn on recording it tanks my frame rates . I'm getting like 1 fps for some reason and it wasn't happening the 2 days ago. Anyone have any ideas ?

 

[tEd]

what you consider more solid games?

As I said earlier, UT2003 seems a good place to start. It's not ancient, but it's widely popular and has gotten a lot of attention from the IHV's. Perhaps a good Q3-based engine as well to take a look at how OGL/D3D may compare. In time, it might also show if there are optimizations not just in a "general" case looking at scenes, but with individual games taken into account.

FarCry gets a lot of attention as one of the better-looking games to come out, and while it's good to examine I don't think it makes a good ground state. It has other issues interacting with ATi's and nVidia's drivers right now, so it's hard to tell which results might come from the algorithm alone, and what might be a side-issue.

Max Payne 2 is a good start, though. Though I'm sure people would also be much happier knowing what driver tweaking you used so they can verify and validate the findings on their own.

But more games, more info (FPS comparisons), and more situations are the only way to sort things out properly. I'm not a big fan of over-analyzing one circumstance--and of course everyone likes a diversity of control factors as well, so the testing doesn't feel boxed in.

i've done some benchmarks in ut2004 in a map where i see problems(more texture shimmering,light mipmap banding)

res 1600*1200 using in game 16x AF (DM_Antalus):

w colored mipmap -> 67.3 -> default
wo colored mipmap -> 79.9 -> default
wo colored mipmap-> 77.3 -> forced trilinear
wo colored mipmap -> 72.0 -> disabled AF optimizations
wo colored mipmap -> 67,3 -> forced trilinear + disabled AF optimizations

 

[jvd]

sigh now fraps isn't working at all

 

[Ailuros]

Rather LOD differences and with a rather lucky guestimate lower than default "0" in some cases.

I wonder if that´s true if it´s really done on purpose or if it´s some sort of driver bug.

I'm sure it's not only LOD. If you do LOD lower 0 you may expect less performance right? Here it's the other way around you get a more performance. I can enable/disable the optimizations on will

You can disable the optimisations? What´s the performance difference then with and w/o.

And you´re right; they might actually pick a more negative LOD value in exchange with less AF samples. That way performance will of course be better.

***edit: your UT2k3 scores are from a X800PRO?

 

[tEd]

Rather LOD differences and with a rather lucky guestimate lower than default "0" in some cases.

I wonder if that´s true if it´s really done on purpose or if it´s some sort of driver bug.

I'm sure it's not only LOD. If you do LOD lower 0 you may expect less performance right? Here it's the other way around you get a more performance. I can enable/disable the optimizations on will

You can disable the optimisations? What´s the performance difference then with and w/o.

And you´re right; they might actually pick a more negative LOD value in exchange with less AF samples. That way performance will of course be better.

***edit: your UT2k3 scores are from a X800PRO?

yes x800pro

 

[Mr. Travis]

What... the... heck?

The shimmering is identical on BOTH videos, very evident in both, are you people even watching both?

I have yet to see a serious piece of evidence on this AF issue which is disappointing considering the negative momentum rumors like this are causing around the net...

blind? Def. not identical by far but if you don't see the difference the better for you i guess

er... i'm not kidding, my default vid player winamp, if you scale the vids SMALLER, they look identical, both have very severe shimmering and it happens exactly the same way on both, such as when it reaches that light part the shimmering cuts in at the exact same angle on both... this is how I first viewed them and couldn't see what you guys were talking about... this test just doesn't seem reliable to me... and to boot Far Cry has weird AF issues on some pixel shaded surfaces as well as others, and uses a weird detail texture scheme that does not smooth over a distance...

I'm no fan boy I've just seen a lot of bull around here without anything everyone will agree is conclusive... take a look at HardOCP's review today... extensive IQ AF, AA testing... no issue found http://www.hardocp.com/article.html?art=NjI4 ... and that's what I would tend to expect because the way this optimization works it is essentially cutting out a few strips of detail and making those area smoother instead... why would there be shimmering or other magnified effects? whatever, I'm not going to argue about it any more then this I'm just sick of the hunt for what is simply not been there so far...

(And personally I think NV wins this round with comparable speeds and PS 3.0, IQ is basically identical outside of this craziness)

 

[kemosabe]

Add to the growing list another investigator who considers image quality to be equivalent between Nvidia and ATI (already posted on another thread): http://www.extremetech.com/article2/0,1558,1607374,00.asp

If one neglects power supply considerations and availability issues (not that they aren't important), then it could essentially boil down to higher frame rates at high resolutions for X800XT vs. SM3.0 support for 6800U.

 

[bloodbob]

5 bit interpolator for the mip maps isn't a really big problem because it creates 64 bands instead of 256 bands not too much of a difference I would like a better interpolator I can see it in some of the mipmap transitions on XBIT black and white lines thing. I also believe it does dithering.

Any the problem with the shimmer is they are using a mipmap with too much noise 15ish% of the time ( trilinear fixes it by blending it with a lower noise mipmip ).

 

[tEd]

after trying farcry i can confirm that the x800pro does show more texture aliasing in that place. Pretty much i only see it when i use the flashlight though . After playing around 1-2 levels and besides that wall i didn't notice any problemes with filtering anywhere else.

After all that tesing i've done so far i would conclude whatever ati does in that new AF mode it can lead to more noticeable texture aliasing or light mipmap banding in some places. The question is whether you care about it . As long as i have enough performance room to get around it i'm certainly will do so. If there is a time i need the 5-15 fps boost it can give me i'm happy to take it for alittle IQ decrease here and there. For me that time is not here yet

 

[Ailuros]

My question would be if offsetting MIPmap LOD to quality is enough to cure the side-effects.

 

[Grestorn]

YUCK!!!

Ok, I downloaded 'em a while ago then forgot to check them out...the shimmering don't take any looking for, it jumped out at me and I'd find it annoying as hell to play with it flickering like that.

You sure it's a result of trylinear though? It looks like it ain't applying aniso to all levels, did you try forcing it to with rTool and checking it again?

I'm *not* sure that it's a result of "Trylinear". In the article I mentioned before (http://www.computerbase.de/artikel/hardware/grafikkarten/radeon_x800_texturen) the author found out that ATI is also fiddling with the LOD base. Maybe this is the cause of the shimmering.

What I'm sure of is that the stage optimization is off. There are two reasons: I don't use the CP to force AF but the in game setting. It's quite common knowledge, that the stage optimization is only applied when AF is forced through the CP.

And, just to be sure, I also use the "rtool" (with the option "R300") to force the stage optimization off. And to be even more sure, I deinstalled the rTool (maybe it's the cause of all the fuzz?) and checked again: same results.

 

[tEd]

My question would be if offsetting MIPmap LOD to quality is enough to cure the side-effects.

actually it can worsen the problem. I made some wierd findings in mp2. I reduced the texture shaprness to the lowest setting(you would think that will get rid of the aliasing) but instead of improving the situation the aliasing got even worse

update:

using quality mipmap setting actually does improve the situation but doesn't eliminate it.

using performacne mipmap setting makes it worse :?

strange behaviour

 

[Grestorn]

after trying farcry i can confirm that the x800pro does show more texture aliasing in that place. Pretty much i only see it when i use the flashlight though . After playing around 1-2 levels and besides that wall i didn't notice any problemes with filtering anywhere else.

After all that tesing i've done so far i would conclude whatever ati does in that new AF mode it can lead to more noticeable texture aliasing or light mipmap banding in some places. The question is whether you care about it . As long as i have enough performance room to get around it i'm certainly will do so. If there is a time i need the 5-15 fps boost it can give me i'm happy to take it for alittle IQ decrease here and there. For me that time is not here yet

Well, that's about exactly the same thing I said. It's not that you can see these effects left and right, but only at very specific spots.

Anyway, this is one game. There are other games where the effect is more evident and permanent (Deus Ex 1 and Max Payne 1 come to mind).

My point is: Brilinear can be a sensible optimization. But since there is currently no way for the game programmer to enable or disable it (through a DX flag), it has to be an option for the (advanced) user. There is just no other way.

It is neither acceptable that nVidia is forcing their brilinearity on us (currently only on the FX line, but that doesn't matter), nor is it acceptable that ATI is telling us "we have an intelligent alogrithm, it's doing the right thing, just believe us" when, in fact, they're just doing the same (maybe even more, with some changes to the LOD base) while pretending to be the good guys.

We, as the customers, have to stop to believe in either one of these companies. We have to look hard, discover these details and force them to acknowledge their schemes and, in the end, to put the user back in control.

I hope that, finally, some people here stop thinking of me as a fan-boy, only because I actually found a place where ATI's optimized filtering is definitely visible.

 

[Bjorn]

It is neither acceptable that nVidia is forcing their brilinearity on us (currently only on the FX line, but that doesn't matter),

I agree with the rest of the post but, i wouldn't say that forcing brilinear on the FX doesn't matter. Thankfully, i don't own a FX myself but if i would, i would like a full triliner option (and a non angle dependant option).

 

[Grestorn]

It is neither acceptable that nVidia is forcing their brilinearity on us (currently only on the FX line, but that doesn't matter),

I agree with the rest of the post but, i wouldn't say that forcing brilinear on the FX doesn't matter. Thankfully, i don't own a FX myself but if i would, i would like a full triliner option (and a non angle dependant option).

You misunderstood me (sorry, must be my english, I'm german after all ).

I meant, brilineariry is currently only forced on the FX line. The previous cards don't support brilinear and on the 6800 line it seems that it might become an option.

I didn't want to imply that in anyway it's acceptable that they force that optimization on any kind of card.

Hope that clears that up.

 

[tEd]

after trying farcry i can confirm that the x800pro does show more texture aliasing in that place. Pretty much i only see it when i use the flashlight though . After playing around 1-2 levels and besides that wall i didn't notice any problemes with filtering anywhere else.

After all that tesing i've done so far i would conclude whatever ati does in that new AF mode it can lead to more noticeable texture aliasing or light mipmap banding in some places. The question is whether you care about it . As long as i have enough performance room to get around it i'm certainly will do so. If there is a time i need the 5-15 fps boost it can give me i'm happy to take it for alittle IQ decrease here and there. For me that time is not here yet

Well, that's about exactly the same thing I said. It's not that you can see these effects left and right, but only at very specific spots.



My point is: Brilinear can be a sensible optimization. But since there is currently no way for the game programmer to enable or disable it (through a DX flag), it has to be an option for the (advanced) user. There is just no other way.

i can tell you one thing though the texture shimmering doesn't come from brilinear . It's the new AF mode they using on the x800. You mentioned that you tried to get around texture stage optimization with rtool well it doesn't work anymore. The new AF mode forces texture stage optimization and it doesn't matter whether you use rtool or use the ingame AF setting , you can't disable it that way.

From my expereince i would say brilinear is mostly fine. Having an option to disable it won't hurt though

 

[Grestorn]

I can tell you one thing though the texture shimmering doesn't come from brilinear . It's the new AF mode they using on the x800. You mentioned that you tried to get around texture stage optimization with rtool well it doesn't work anymore. The new AF mode forces texture stage optimization and it doesn't matter whether you use rtool or use the ingame AF setting , you can't disable it that way.

Well, actually I don't care how this optimization works.

If it causes visible artefacts it has to be an option.

Let's just replace all occurances of "brilinear optimizations" with "AF optimizations", ok?

But if you're right, it will be very interesting to see the same spot with an 6800, both with their brilinear opt. on and off.

Since I'm in the lucky position to be soon the owner of both cards, the x800 and the 6800 (as soon as the 6800 is available), I will make that comparison.

If what you're saying is correct, it would mean that at this very spot, the 6800 should look fine, regardless of their brilinearity. I actually hope that this is not the case, otherwise I would be the filthy nvidiot again here, just for finding that spot

 

[tEd]

I can tell you one thing though the texture shimmering doesn't come from brilinear . It's the new AF mode they using on the x800. You mentioned that you tried to get around texture stage optimization with rtool well it doesn't work anymore. The new AF mode forces texture stage optimization and it doesn't matter whether you use rtool or use the ingame AF setting , you can't disable it that way.

If what you're saying is correct, it would mean that at this very spot, the 6800 should look fine, regardless of their brilinearity. I actually hope that this is not the case, otherwise I would be the filthy nvidiot again here, just for finding that spot

well of course a to aggressive brilinear can give you other artifacts(mipmap banding). I'm just saying the brilinear ati use right now is mostly fine whether this is the case for nvidia too you we'll see then.

 

[cristic]

You also have to download a new devmode.lua file and replace your existing one before the cheats will work, including level unlock.

I'll try and dig up a link....

Since I didn't see any link here's the zip with all one needs to enable the cheats in farcry

http://cristi.caloghera.home.ro/farcrycheats.zip

if clicking on the above link does not work, than you should copy-paste the ling into your browsers address bar, or rightclick it and save it direclt; then it should work,...

 

[Evildeus]

Add to the growing list another investigator who considers image quality to be equivalent between Nvidia and ATI (already posted on another thread): http://www.extremetech.com/article2/0,1558,1607374,00.asp

And said in the other thread, the author of this article doesn't even know that Brilinear is forced on 6800s with 61.11 :?