When enough is enough (AF quality on g70)

[tEd]

/me wonders which was first company to made "fast" aniso... with noticeable problems also... and reviewers were happy to use non-equal-quality modes comparing ATi vs NVidia...
so why wonder if we get worse and worse quality on certain cases?

When nvidia only had a crappy ordered grid 4xMSAA mode and ATI a much better looking 4x sparsed let alone 6x MSAA mode , reviewers obviously couldn't do a fair comparison because nvidias AA hardware just couldn't do better than what it could do.

As long as reviewers point out that one implementation is superior than the other , with that we can live. It was the same case back then with ati's r8500 and nvidias gf4600 AF. r8500 couldn't do better than what it could do.

This case is different. G70/nv40 can do better than with reviewers usually bench it with , it's thus a unfair comparison by their choice and that's not acceptable.

There is also another point to this all. If we accept what's happening here IHV's won't put more effort in increasing quality of filtering,AA and so on. Why would they if everybody accepts the crappy quality.

 

[chavvdarrr]

i totally agree. I just think its next to impossible to turn the things /me hopes i'm wrong

 

[tEd]

wow i just read 2 replys from 3 on the hardcop forum regarding the subject and just realized how clueless they are.

It seems the mose people still believe that on ati cards you can't disable AF optimizations while you can do that since cat4.10

 

[Razor1]

Ted Apples to Apples, Benchmarks should turn off ATi's AI should be turned off and nV's should be put on High Quality. Over all ATi will lose more.

Now with the G70 hopefully nV decides to fix the shimmering, that would be very good since they have really nothing to lose since there is no competition at the moment.

 

[Hyp-X]

Ted Apples to Apples, Benchmarks should turn off ATi's AI should be turned off and nV's should be put on High Quality. Over all ATi will lose more.

Now with the G70 hopefully nV decides to fix the shimmering, that would be very good since they have really nothing to lose since there is no competition at the moment.

And when competition shows up they'll just turn the optimizations back on...

 

[tEd]

Ted Apples to Apples, Benchmarks should turn off ATi's AI should be turned off and nV's should be put on High Quality. Over all ATi will lose more.

Now with the G70 hopefully nV decides to fix the shimmering, that would be very good since they have really nothing to lose since there is no competition at the moment.

With nv40 yes , but with a g70 HQ doesn't have to look better than ati's AI low.

The problem is that in today reviews the only measurement in the end are fps numbers. We should have system where IQ can have an influence of the measurement of the product. This is very complicated though as to make it right you would need to adjust it on a per game or per benchmark basis.

 

[andypski]

Ted Apples to Apples, Benchmarks should turn off ATi's AI should be turned off and nV's should be put on High Quality. Over all ATi will lose more.

Except that judging from the videos that's still not remotely apples to apples image quality wise, so claiming that that's a fair comparison seems strange in light of the facts.

And if, as it appears, we're managing to maintain higher image quality with our optimisations enabled than the competition do with all their optimisations supposedly disabled then shouldn't we be rewarded for doing a good job and giving good image quality options to our users, rather than penalised for someone else apparently failing to do the same?

And why is everything still shimmering on 7800GTX in "High Quality" mode if all optimisations are off, anyway?

If we do better with CatalystAI enabled than they do in "High Quality" mode then why shouldn't we get to keep our optimisations on until they at least match our image quality?

 

[Moloch]

I made the videos and I couldn't find any satisfying compression. If there were artefacts because of the compression fan--boys (and probably nvidia) would have say "These videos are meaningless, artefacts come from the compression."Losless was required.

Sorry, but you need to read www.doom9.org
With the right compression and settings, i promise you can get results that are indistinguishable from the original.
I repeat, the video quality of the videos I saw was far from what I see playing the same game same res same settings.
It just doesn't look right.

 

[Skrying]

Sorry, but you need to read www.doom9.org
With the right compression and settings, i promise you can get results that are indistinguishable from the original.
I repeat, the video quality of the videos I saw was far from what I see playing the same game same res same settings.
It just doesn't look right.

Any type of compression at all, no matter how great of a quality it is, would be dismissed by people who dont want to believe it. He honestly had no choice in this, he had to keep it lossless, even if their was a codec that made it 70MB smaller and looked almost perfectly the same.

 

[Moloch]

Any type of compression at all, no matter how great of a quality it is, would be dismissed by people who dont want to believe it. He honestly had no choice in this, he had to keep it lossless, even if their was a codec that made it 70MB smaller and looked almost perfectly the same.

He could have provided an alternative for people who aren't anal.

 

[Tridam]

He could have provided an alternative for people who aren't anal.

And a coffee ?

 

[Moloch]

And a coffee ?

It's not too hard to do nor does it take alot of time.
Let's face some facts here, your site isn't the fastest site I've been to(my loc is so cal, so it may be much faster for others) and those videos are 70MB a piece.
I was pulling about a megabit per second from you on a 4 megabit connection.
Based on calculation it would take me about 42 minutes to download them all.
Is it worth it?
By adding an alternative link you will greatly reduce the time it takes to download so people might be more willing to look at the videos and it will obviously use less bandwidth from your host.

 

[Kanyamagufa]

He could have provided an alternative for people who aren't anal.

I think it's pointless to keep arguing about this. If you don't like how he handled the situation, that's fine. But continuing to complain about it cannot help the situation.

As for the rest of the thread...I'm sick of all the strings these companies are pulling and their desperate need to find loop holes to remain competitive.

 

[Tridam]

It's not too hard to do nor does it take alot of time.
Let's face some facts here, your site isn't the fastest site I've been to(my loc is so cal) and those videos are 70MB a piece.
I was pulling about a megabit per second from you on a 4 megabit connection.
Based on calculation it would take me about 42 minutes to download them all.
Is it worth it?
By adding an alternative link you will greatly reduce the time it takes to download so people might be more willing to look at the videos and it will obviously use less bandwidth from your host.

3DCenter is not my website. I recorded these videos and provided them to some people who were free to publish them or not. I think it's great that 3Dcenter was able to publish the full quality videos.

I did not publish these videos on Hardware.fr because they were too big and because I did not find an acceptable compression and of course I tried many different codecs. I can't spend a week finding an acceptable codec for that purpose... if it exists.

 

[Moloch]

3DCenter is not my website. I recorded these videos and provided them to some people who were free to publish them or not. I think it's great that 3Dcenter was able to publish the full quality videos.

I did not publish these videos on Hardware.fr because they were too big and because I did not find an acceptable compression and of course I tried many different codecs. I can't spend a week finding an acceptable codec for that purpose... if it exists.

I like the videos I'd just like an alternative since they take a long time to download.
An acceptable codec is xvid, you just need to know how to use it, which is admittidly not a walk in the park since it's so configureable.

I think it's pointless to keep arguing about this. If you don't like how he handled the situation, that's fine. But continuing to complain about it cannot help the situation.

What more is there to argue about?

 

[Razor1]

Except that judging from the videos that's still not remotely apples to apples image quality wise, so claiming that that's a fair comparison seems strange in light of the facts.

And if, as it appears, we're managing to maintain higher image quality with our optimisations enabled than the competition do with all their optimisations supposedly disabled then shouldn't we be rewarded for doing a good job and giving good image quality options to our users, rather than penalised for someone else apparently failing to do the same?

And why is everything still shimmering on 7800GTX in "High Quality" mode if all optimisations are off, anyway?

If we do better with CatalystAI enabled than they do in "High Quality" mode then why shouldn't we get to keep our optimisations on until they at least match our image quality?

Because I don't see that much shimmering in a game situation. In that demo fine it has a good deal of shimmering. Plus the LOD fix was never used either. Which nV lets you choose the LOD settings in the control panel now.

 

[wireframe]

This shimmering business (and moire) is an issue I will be keeping a close eye on. It really bothers me that these basics are being trampled upon so late in the game. I really hope that ATI will not follow suit and that we will see deteriorated filtering quality in the R520 for the sake of speed.

About the videos:

The Fraps FPS1 format is already compressed somewhat, but I cannot find anything about it being lossless. The only hint for it being lossless I could find was in ffmpeg/Mplayer, which stated something to the effect of lossless decoding. Not sure what they mean by that. It's unfortunate that the Fraps team cannot clearly state what compression methods are used, epecially for situations like this. (BTW, I tried converting the video to HuffYUV, which is completely lossless and this increased the file quite a bit. Therefore, I believe the FPS1 format is lossy.)

As for arguing about wanting more highly compressed videos. I fail to see what the point of that. With a lossy codec you will always lose detail and in this case we are talking extremely fine details/high frequency. If anything, a problem with the existing video is that they are not lossless in their current state (a guess I make from my previously mentioned experiement). Sure, Xvid compressed videos could show you what is going on, but this is about degrees. You either accept that it is happening and then you need to look at the actual output in uncompressed form. Lossy compression is therefore pointless.

One thought about capturing this phenomenon using Fraps or other tools. Fraps uses a strict frame rate format when capturing. These videos, for example, seem to have been captured at 30fps. Now, this is not necessarily the frame rate that the video card is outputting frames. I am not familiar with how Fraps deals with this, but wouldn't it be a safe guess that Fraps gathers frames at strict time intervals. What's my point? That if you are running at 90fps 3D output and capturing at 30fps video, you are discarding 2/3rds of the data and the captured frames are not consequtive as seen from the 3D output side. This means that the faster the 3D hardware is running the more lost frames and therefore the 'blending" effect in time, over frames, will be distorted. This is important because shimmering is only seen in motion and not in the static frame. If you discard 3D frames you can lose or create a shimmering effect merely by what frames you discard when you "stitch" your video output.

A solution that would at least make comparison fair is to restrict the frame rate output of the 3D application when capturing to match the capturing rate of Fraps (using v-synch).

 

[PeterAce]

Tridam, what the program used (in the videos) to show these filtering modes? Is it freeware? If so can you provide me a download link?

 

[BRiT]

restrict the frame rate output of the 3D application when capturing to match the capturing rate of Fraps (using v-synch).

I believe this is what FRAPS already does when capturing...

 

[Tridam]

I used UT2003 (with lod set to 0.0 of course).

Wireframe, you're right. But I tried to avoid that problem by slowing down UT (slomo command). So these videos show what you can see on the screen in real time.
Fraps reduce the framerate. But if the movement between 2 frames is too important (that happens with default speed and 30 fps) the videos don't look good.