OMG HARDOCP REVIEW OF UT2003 AND FILTERING

[Dave H]

I'm curious why this sort of thing has been rejected in the past.

Wonder no more, DaveH - it was done by ATI in the not so distant past, and they got slammed for it, upon the release of the 8500, iirc, there was a driver set that didnt do "full trilinear" and ATI got hammered for it.

If you're referring to the "quack" scandal, what was going on there was that the 8500 was for some reason fetching really high mipmaps (i.e. mipmaps with far too little detail) for a handful of textures in Q3 which were nonetheless heavily used enough to make a big difference in both performance and IQ. Unlike the current issue, the IQ loss was extremely noticeable when it existed.

Also unlike the current issue, there is at least a reasonable chance that this "optimization" was in fact a bug on ATI's part. (Or perhaps a working optimization that for some reason failed for those particular textures. Or an optimization that worked fine on the R1xx but for some reason did not on R2xx. That is, there was definitely active app-detection going on, but one seriously doubts that the IQ results were what ATI was expecting.)

Meanwhile, the 8500 didn't do trilinear at all when AF was enabled. Obviously that was almost two chip generations ago (another way for saying, "just last chip generation") but it does go to show how far we've come to be worrying about problems with partial trilinear that apparently can't even be seen in most situations.

EDIT: And if you're referring to the fact that the 8500 didn't do trilinear at all when AF was enabled...it didn't do trilinear at all when AF was enabled! It wasn't a driver bug but a hardware limitation; AFAIK, the 9200 you can go out and waste your money on buy today will have the exact same behavior. And straight bilinear is way way more noticeable than this partial trilinear NV3x is using in UT2003.

 

[Reverend]

Here's the page with Reverend's post, to save giving Kyle a few hits.

Please, I only want clarifications from Kyle. There is no need to be nasty (anymore) towards Kyle because most of you have already voiced your opinion about him/his site.

 

[Althornin]

Wonder no more, DaveH - it was done by ATI in the not so distant past, and they got slammed for it, upon the release of the 8500, iirc, there was a driver set that didnt do "full trilinear" and ATI got hammered for it.

<Buncha stuff i already know>

nope, nice rant about stuff i already know though!
i thought about it some more, and i just cant remmeber which card it was that did the same thing (not quite trilinear). It just seems like it was longer ago, but the 9700 launch was a year ago! However, i am sure that some of the other peiople here remember what i am talking about.
Again, thanks for the long story on stuff that was known, and appologies for my mistake (tho i said "iirc", when apparently i did not recall correctly).

 

[digitalwanderer]

Here's the page with Reverend's post, to save giving Kyle a few hits.

Please, I only want clarifications from Kyle. There is no need to be nasty (anymore) towards Kyle because most of you have already voiced your opinion about him/his site.

Sorry, I did not post that for people to flame on...and I've been banned from posting there for a while now so I'm pretty harmless.

 

[Reverend]

The point I was trying to get across is that if you personally do not feel like you want to visit a website for whatever reasons, I don't think there is any need, nor is it appropriate, to suggest to people to feel/do the same. I would appreciate it if this doesn't happen in our forums. I do not like the possibility of Beyond3D's forum to appear to be "at war" with another site's forum.

 

[K.I.L.E.R]

No one is at war with anyone, nor will these forums be at war with [H]'s forums. People are attacking the credibility of Kyle and Brent. People feel that they are being lied to. People know they are being mislead.
Hence the attacks.

Rev, If I were to mislead you will you be okay with that?

The point I was trying to get across is that if you personally do not feel like you want to visit a website for whatever reasons, I don't think there is any need, nor is it appropriate, to suggest to people to feel/do the same. I would appreciate it if this doesn't happen in our forums. I do not like the possibility of Beyond3D's forum to appear to be "at war" with another site's forum.

 

[OpenGL guy]

If you're referring to the "quack" scandal, what was going on there was that the 8500 was for some reason fetching really high mipmaps (i.e. mipmaps with far too little detail) for a handful of textures in Q3 which were nonetheless heavily used enough to make a big difference in both performance and IQ. Unlike the current issue, the IQ loss was extremely noticeable when it existed.

How do you know it had a large impact on performance? Once the problem was fixed, performance stayed the same, and even improved later.

Meanwhile, the 8500 didn't do trilinear at all when AF was enabled. Obviously that was almost two chip generations ago (another way for saying, "just last chip generation") but it does go to show how far we've come to be worrying about problems with partial trilinear that apparently can't even be seen in most situations.

EDIT: And if you're referring to the fact that the 8500 didn't do trilinear at all when AF was enabled...it didn't do trilinear at all when AF was enabled! It wasn't a driver bug but a hardware limitation; AFAIK, the 9200 you can go out and waste your money on buy today will have the exact same behavior. And straight bilinear is way way more noticeable than this partial trilinear NV3x is using in UT2003.

Except that when you compare a part that is doing no or partial trilinear to a part that is doing full trilinear (AF or no) then it's not a fair comparison, is it? I wouldn't expect people to compare bilinear-AF on an 8500 to trilinear-AF on a GeForce card, or, if they did, then the image quality differences need to be pointed out.

 

[digitalwanderer]

The point I was trying to get across is that if you personally do not feel like you want to visit a website for whatever reasons, I don't think there is any need, nor is it appropriate, to suggest to people to feel/do the same. I would appreciate it if this doesn't happen in our forums. I do not like the possibility of Beyond3D's forum to appear to be "at war" with another site's forum.

My apologies, I'll shut up.

 

[Reverend]

No one is at war with anyone, nor will these forums be at war with [H]'s forums. People are attacking the credibility of Kyle and Brent. People feel that they are being lied to. People know they are being mislead.
Hence the attacks.

Rev, If I were to mislead you will you be okay with that?

No, it will not be okay with me and I will shoot you an email saying "You sonofab*tch, you suck so bad for misleading me... I am never going to visit your site again and I will tell my buddies via emails to do the same"... while posting in Beyond3D's forum "In this article by K.I.L.E.R. he said these things... here are my point-by-point reasons why I don't think K.I.L.E.R's words expresses the truth of the situation and appears misleading to me..."

That is what I would do, mostly out of being able to act rationally and also out of respect for the wishes of one of B3D's site partners in his efforts to improve the content quality of the site's forums.

 

[Althornin]

No, it will not be okay with me and I will shoot you an email saying "You sonofab*tch, you suck so bad for misleading me... I am never going to visit your site again and I will tell my buddies via emails to do the same"... while posting in Beyond3D's forum "In this article by K.I.L.E.R. he said these things... here are my point-by-point reasons why I don't think K.I.L.E.R's words expresses the truth of the situation and appears misleading to me..."

That is what I would do, mostly out of being able to act rationally and also out of respect for the wishes of one of B3D's site partners in his efforts to improve the content quality of the site's forums.

I would also like the personal attacks on kyle and brent to cease.
They are unproductive, and all they do is lower this site, and cloud the very real issue brought up with FUD - and they give reason for all the fanpeople to point out "bias". We all have our favorite forum that allows and is cool with rants and insults, so lets put em there and not here.

Not that anyone would not listen to Rev. but then listen to me

 

[K.I.L.E.R]

That is what I would do, mostly out of being able to act rationally and also out of respect for the wishes of one of B3D's site partners in his efforts to improve the content quality of the site's forums.

I hope you are not implying that I'm being disrespectful towards the B3D site partners.

My intentions never were to be disrespectful.

 

[Miksu]

The entire Trilinear issue was brought up with NVIDIA in face-to-face meetings over two months ago. Long before the B3DPolice came to the scene of the crime. Myself and the owner of another hardware site that is in the spotlight often talked specifically about Bi/Tri/AF with NVIDIA and how it was handled in the drivers.

If this is actually the case, is sitting on this information and not mentioning supposed to be a good thing for everyone? And why get so annoyed by the "B3DPolice" for doing so?

It seems that Kyle has known about all of these issues in nVidia's drivers way before those were posted here. He says "I've know that for months" to every new thing you post here. So a) He has really known about these but didn't report them to public b) Doesn't want to be the second and that's why claims that he has known about these issues for a long time.

I'll say B is more likely.

 

[Reverend]

That is what I would do, mostly out of being able to act rationally and also out of respect for the wishes of one of B3D's site partners in his efforts to improve the content quality of the site's forums.

I hope you are not implying that I'm being disrespectful towards the B3D site partners.

My intentions never were to be disrespectful.

I am not implying anything against any one specific person -- I have simply given an example of the type of forum participation I/B3D prefers while taking into account its forum participants wish to "vent", which we prefer to be done via private correspondences, separate from B3D forums. As much as possible, B3D prefers criticism of content here rather than criticism of the author of the content. Look at Dave -- he has more reasons than anyone to criticize Kyle here due to the attacks (not criticisms but actual written attacks) on B3D Kyle has made in his own forums (and AFAICT Dave has not attacked neither Kyle nor [H] anywhere) but Dave is above this kind of thing. So am I, although I have had to practice incredible self restraint. Yes, sometimes we forget, or simply cannot stand it anymore, and let slip a criticism or two but we are almost always conscious of the fact that we should not be doing this all the time. Many of you here will not let this go and you just want to continue criticizing Kyle/[H] again and again in the same thread or let this permeate into other threads. I hope you'll respect my wish for this to discontinue being the case.

 

[zurich]

Rev,

I won't cut and paste Kyle's reply, but wow, you sure do have some restraint.

(and FYI, it took massive restraint not to cut and paste it )

 

[Dave Baumann]

If you're referring to the "quack" scandal, what was going on there was that the 8500 was for some reason fetching really high mipmaps (i.e. mipmaps with far too little detail) for a handful of textures in Q3 which were nonetheless heavily used enough to make a big difference in both performance and IQ. Unlike the current issue, the IQ loss was extremely noticeable when it existed.

No, he wasn't referring to that. 8500 also had a Bi/Trilinear mix functionality. Read [H]'s 8500 Revisited review fo their views on it then and the use of mipmap colorings.

 

[Bjorn]

Except that when you compare a part that is doing no or partial trilinear to a part that is doing full trilinear (AF or no) then it's not a fair comparison, is it? I wouldn't expect people to compare bilinear-AF on an 8500 to trilinear-AF on a GeForce card, or, if they did, then the image quality differences need to be pointed out.

Well, [H] pointed out the differences, didn't they ?

They showed some screenshots, played the game and decided that they couldn't see the difference, at least not while playing. They also specifically mentioned that UT was a fast paced FPS which made it even more difficult to see the "problem". Maybe not all here would agree with them, after all, we're talking about a subjective thing. But that's another issue.

Now, i tend to trust B3D a bit more when it comes to subjective opinions though which means that i won't really trust [H] in this matter unless i tried both cards and could verify it myself. But i don't see what they did wrong. At least not to such an extent as to what some people here think they did.

Maybe i should add that i agree with most of the stuff Dave H has posted here

 

[Reverend]

Rev,

I won't cut and paste Kyle's reply, but wow, you sure do have some restraint.

(and FYI, it took massive restraint not to cut and paste it )

I'd been refreshing that thread for an hour earlier and the page just won't load, so I didn't know if Kyle replied or not. Now that I have seen his reply, I have just made mine.

 

[Reverend]

But i don't see what they did wrong

They didn't do anything wrong in my opinion.

They just have a website with personnels that have different interests than B3D's. Leaving aside "trusts of websites", it is ultimately down to readers to gather information, not from just one website, but from many others, to make an informed decision on a particular subject matter.

Whether that subject matter matters to that one or many representative of the general public is one thing we, as reviewers, cannot determine. Presenting facts should be the agenda of reviewers. Presenting subjective opinions as a result of knowing the facts should not be attempted by reviewers, but only by the general public.

 

[Dave Baumann]

DaveH. WRT to your comments as to whether this is noticable or not, I think my earlier comment that "once you know its there you can see it easier" really has affected me now.

I looked at a few more maps, at random to see if I could actually see any mipmap issues. Here are a few images that I believe I can see issues with. Note: I've done mip colour images as well, but I wont put them directly in the thread to pinpoint them.

LostFaith Map

http://www.beyond3d.com/previews/nvidia/nv35/images/utmip/lostfaith-mip.jpg

Citadel Map

http://www.beyond3d.com/previews/nvidia/nv35/images/utmip/citadel-mip.jpg

Antalus Map

http://www.beyond3d.com/previews/nvidia/nv35/images/utmip/Antalus-mip.jpg

Note: Png's of the images are available here if you really want download them.
http://www.beyond3d.com/previews/nvidia/nv35/images/utmip/utmip.zip

Now in these images the effects are subtle, but I can actually spot the different mipmap levels, which you shouldn't be able to do with full Trilinear enabled. Its quite interesting that once you know its there it becomes even more clear there are different mip levels on the uneven levels, as all of these are. You can see that one some polygons there are clearly two different mip levels being used, which you shouldn't be able to see if full Trilinear was in operation. I'm also going to do 100% turaround on my comments on the Antalus map, since it is quite evident to me that there are different mip levels in use once you come out from a shadowed.

Now, is this a representative sample of the entire game? No, I only looked at a few maps. Its ture to say that I looked at as many maps that I could see an issue with as one that I couldn't.

To my mind, yes the effect is subtle, but I can see an IQ difference that would wouldn't expect to be there if full Trilinear was used. From what I've seen its also not true to say that this is either always an issue or never an issue.

(FYI - No I haven't done a full side-by-side analysis to look if there is increase texture aliasing, thats just a theoretical premis. I think I've wasted enough time on this now as it is and really need to crack on with a review or article!)

 

[Hanners]

Thanks Dave - I think all of those screenshots show that the mip-map transitions are obvious to the naked eye without the need for colouring them. I'm not a great one for spotting these things myself at times, but it seems very apparant in all those shots to me.