PDA

View Full Version : A way to force trilinear filtering on a GeForce4 Ti?


phial
08-13-03, 05:25 AM
is there a registry entry that one can use to force standard trilinear filtering in games? not trilinear AF, just standard filtering



the reason is AF gives my card too much of a performance hit .. even at 2xAF. so id settle for Trilinear of the standard , horrible looking bilinear


so..... does anyone know how to do this?

oh and while your here whats yoru favorite tweaking software for Gf4TIs ? =)
actualy gimme a few different ones ify ou dnot mind... so i can experiment. ive gotten this Gf4 recently because i fried my 8500le... nvidia is kinda new ot me

thanks guys

StealthHawk
08-13-03, 06:05 AM
The answer is no :( Unless the game supports trilinear, there's no way for you to get it.

phial
08-13-03, 09:13 AM
not trying to jump on teh "hate nvidia" bandwagon ,but this video card has been nothing but disappointments


if i can sell it for half of what i bought if for last month, im gettin a 8500LE to hold me off until 9700pros come down

phial
08-13-03, 09:19 AM
ok... well does anyone know of a site where i can get alot of info on tweaking these cards? preferrably registry settings and what not


i want to at least try a little more before i burn this thing

saturnotaku
08-13-03, 04:01 PM
http://www.guru3d.com/rivatuner

yoladude
08-13-03, 07:03 PM
Originally posted by phial
not trying to jump on teh "hate nvidia" bandwagon ,but this video card has been nothing but disappointments


if i can sell it for half of what i bought if for last month, im gettin a 8500LE to hold me off until 9700pros come down

actually, i'd keep that geforce4, because in most situations, the 8500le is slowler, and they dont oc to regular 8500 levels (almost on par with gf4s).

here's a review of a geforce4 ti4200:
http://www.hardocp.com/article.html?art=Mjky

notice how it's faster than the 8500 in about all respects. i'm not anti-ati or anything like that, cause their higher-end cards rock, i just dont want to see you waste more money on a card that isn't an upgrade at all.

phial
08-13-03, 10:02 PM
i guess i didnt explain that i owned an 8500le


i had it previous to the Ti4200...


what happened was i volt modded my 8500le and fryd it. my voltmeter was cheap and didnt give me an accurate reading on the voltage


so i bought the TI4200


i agree.. if you use bilinear filtering and quincunx (which IMO isnt usealbe because it looks like CRAP) then the TI4200 is faster


but i use a minimum of 2x AF. just by enabling this one feature causes the TI4200 to drop tojust about how fast my 8500le was


sinse i could use 16x AF (8x equivilant on GF4TI cards) with a mere 10fps drop... i consider the 8500le faster. if i use 8x AF on the Ti4200, my framerate drops from 110 to 50 in BattleField 1942

i averaged 60fps or so with 16xAF and the highest texture/mipmap settings on the 8500le

phial
08-13-03, 10:03 PM
you have to remember that benchmarks usually have AF and AA off.....

so thats why the TI4200 appears faster. when you use features that are sometimes needed (AF) its just simply not the case anymore, as i have found out first hand


thanks saturnotaku , but ive tried Rivatuner =)

i guess its the best one eh?

yoladude
08-13-03, 11:05 PM
fair enough; the 4200 is significantly faster than the 8500 in AA situations (but at lower quality; uses multisample AA, 8500 uses slower supersample), but the 8500 leads in AF situations.

The Baron
08-13-03, 11:07 PM
But the 8500 looks like crap with AF.

phial
08-14-03, 12:15 AM
Originally posted by The Baron
But the 8500 looks like crap with AF.


no.. its just named different.


as far as i, and quite a few other people can tell, 8x AF on the 8500= 4x AF on teh Ti4200

teh 8500 goes up 16xAF... and really, it looks damn nice. the TI4200 IS very clear with 4xAF .. but not better than my 8500

phial
08-14-03, 12:21 AM
Originally posted by yoladude
fair enough; the 4200 is significantly faster than the 8500 in AA situations (but at lower quality; uses multisample AA, 8500 uses slower supersample), but the 8500 leads in AF situations.

well.. theres somehting funny about that actually


in a recent thread, me and a few other people were trying to find drivers that properly did AA on GF4TI cards.

the reason is in 42.xx or higher Dets, teh AA sampling patterns are totally screwed. one person used "FSAA tester" and discovered that 2xAA and 4xAA have basicallyteh same sampling pattern!! wtf is up with that!


and ive noticed this first hand as well. with 2xAA in BF1942, q3, and Anarchy Online vertical jaggies are virtually untouched. horizontal, you can see if you look close where teh AA engine is blending the edges but on virtical edges theres no improvement whatsoever. whats even more astounding is at 4xAA jagges are STILL untouched on virtcal edges!!!!!

4XAA! i mean COMMON. that setting should remove all but the most faint of jaggies... ona moving screen they shoudl be unnoticeable. but they are. they stick outas if your not even using AA at all



4xS solves the virtical jaggy problem but its a mix of supersampling AA and therefore much too slow to use in any but the oldest of games


im currently using 41.07 Dets because they do proper AntiAliasing. 2x looks like 2x should.. and 4x is very , very nice. only problem i have is some bugs in Anarchy Online .. but not major ones


i dont find teh multisampling AA on this card that much faster anyways. its still too slow to use in BF1942... unless you go with 2xAA or Quincunx (crap.. makes the whole screen blurry) and NO AF. but i cant go with no AF, its hard to sniper

StealthHawk
08-14-03, 05:00 AM
Originally posted by phial
well.. theres somehting funny about that actually


in a recent thread, me and a few other people were trying to find drivers that properly did AA on GF4TI cards.

the reason is in 42.xx or higher Dets, teh AA sampling patterns are totally screwed. one person used "FSAA tester" and discovered that 2xAA and 4xAA have basicallyteh same sampling pattern!! wtf is up with that!

I can tell you that I've never had that problem with any driver revision I've used. I've used every official NVIDIA release along with some leaked betas.

I've also never seen anyone prove that the sample patterns for 4x FSAA were anything close to 2x FSAA using a sampling pattern tester.

4x FSAA has always provided minimal quality gains over 2x FSAA on NVIDIA cards that use MSAA(and probably SSAA too, since 4x SSAA looks almost the same as 4x MSAA).

and ive noticed this first hand as well. with 2xAA in BF1942, q3, and Anarchy Online vertical jaggies are virtually untouched. horizontal, you can see if you look close where teh AA engine is blending the edges but on virtical edges theres no improvement whatsoever. whats even more astounding is at 4xAA jagges are STILL untouched on virtcal edges!!!!!

4XAA! i mean COMMON. that setting should remove all but the most faint of jaggies... ona moving screen they shoudl be unnoticeable. but they are. they stick outas if your not even using AA at all

Unfortunately you're going to have that problem with any NVIDIA card.

4xS solves the virtical jaggy problem but its a mix of supersampling AA and therefore much too slow to use in any but the oldest of games

4xS isn't that much slower than 4x. I think it's worth it. It provides the best edge quality out of any official supported FSAA modes.


As to the thread author's complaints, gf4 was never a good card for AF. AF is bugged in D3D. gf4 was still a good card for its time though. Still a pretty decent card.

edit: re: 4x vs 2x FSAA quality. Here are some screenshots from nV News' gf3 preview. 2x FSAA (http://www.nvnews.net/previews/geforce3/images/antialiasing/gf3_q3_1024x768_2xaa.shtml)
4x FSAA (http://www.nvnews.net/previews/geforce3/images/antialiasing/gf3_q3_1024x768_4xaa.shtml)

Then you can take a look around the net at reviews that show NVIDIA 4x FSAA not looking much different from 2x FSAA- it's the same picture. There is no magic driver which had made NVIDIA's 4x FSAA look significantly better than NVIDIA's 2x FSAA. Seeing as how MSAA is done in hardware and is not programmable, that isn't surprising.

phial
08-14-03, 08:15 AM
well, its funny that the 41.07 Dets did make my FSAA look better on 4x and 2x



im not posting just to contradict you... it really did happen.


and isnt AA implimented thru an algorithm that can be changed?

StealthHawk
08-14-03, 08:11 PM
Originally posted by phial
and isnt AA implimented thru an algorithm that can be changed?

NVIDIA's MSAA is not flexible. Otherwise we would have seen 4x RGMS by now, and gf3's QCA would look as good as gf4's. There is no hack that can change gf3 QCA to gf4 QCA either. Again, suggesting that sample patterns are not programmable to any reasonable degree.

Think of it this way. 2x, or 4x is doing a certain number of samples. Where you take those samples should not affect performance. So, if NVIDIA could increase quality by changing sampling position, why have they not done so? The answer is that they can't.

phial
08-14-03, 08:30 PM
thread (http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=15064)



check out this thread


you said the same thing at the beginning, about NV MSAA not being programmable, but didnt post after that



how come so many of us ARE finding improvements with different driver verions? if what your saying is true, then we should not be able to see any difference at all. but i can, its quite noticeable in fact

The Baron
08-14-03, 08:55 PM
how come so many of us ARE finding improvements with different driver verions? if what your saying is true, then we should not be able to see any difference at all. but i can, its quite noticeable in fact
Of course, it never occurred to you that it might be other settings that have nothing to do with AA. LOD, placebo effect, etc. al.

StealthHawk
08-14-03, 09:04 PM
Originally posted by phial
thread (http://www.nvnews.net/vbulletin/showthread.php?s=&threadid=15064)



check out this thread


you said the same thing at the beginning, about NV MSAA not being programmable, but didnt post after that



how come so many of us ARE finding improvements with different driver verions? if what your saying is true, then we should not be able to see any difference at all. but i can, its quite noticeable in fact

Prove it. I don't see any proof in that thread except stuff that corroborates what I said. Nobody posted screenshots showing the magical improvement of 4x FSAA.

OTOH, we have seen this:
1) all GeForce cards using gfFX era drivers have the same FSAA.
2) not much difference between 2x and 4x on a gf3 with the very first drivers.

Contrary to what you are claiming, ie, that quality used to be good and somehow is bad now. I don't know what you are seeing. I don't know whether you installed drivers correctly. I can tell you that pure FSAA speed has almost never gone up, so I don't see how/why NVIDIA would be decreasing quality.

Why do some people say that FSAA is broken for them with some drivers(not working at all)? Again, I have no clue. I don't have that problem.

phial
08-14-03, 09:05 PM
i have had mipmap and texture detail both set to highest possible thru nvtweak, with all the drivers ive tested

and of course ive tested them with these settings on default. the default of nvtweak, not the default that the drivers set. so even if one version sets it higher than the another version, i have set them all equal



i didnt know that LOD had such an impact on aliasing edges? seemed to make no difference EVER when i messed with it in the past, with no FSAA enabled.



look... i may not be the most knowlegeable about the details of this stuff, but i have been dabbling in it for a while and have a rudimentary knowledge. i can spot mipmap transition, and can tell teh difference when 2xAA is enabled and when its not

with the 40.72, i notice a VERy big difference on verticle edges. you guys know what i mean.. how AA blurs the jaggies together if you look close and magnify the screen. with 44.x drivers those verticle edges are not even touched.



there has to be more to it than you guys know.

phial
08-14-03, 09:10 PM
ok i have no proof except what i seen with my own eyes.


admit i was assuming when i said the guy tested with FSAA Tester. i believed him but perhaps he didnt have a clue as to what he was talking about

i have done a clean install for my current drivers, and used 2 different cleaning programs for the testing i did. when i have used these drivers on an "old" install where i have used several different drivers, the results are the same as now with the fresh install

ill take some screen shots when i go home soon. the power outages in Ontario (teh ones affecting New York as well) have put the server we use for out billing offline, so were going home early "woot!" as they say :afro:

might not have them up until tomorrow. i dont think ill feel like cleaning drivers 3 or 4 times tonight. but iwill have them up within a day or so



i have 2g of smoe good ganga waiting for me, halleluja

StealthHawk
08-14-03, 09:34 PM
I uninstalled 44.03, then installed 40.72.

2x (http://www.mrclam.com/~stealth/2x4072.jpg)
4x (http://www.mrclam.com/~stealth/4x4072.jpg)

As you can see, 4x is not doing any better AA at vertical and horizontal edges. If you look at my big FSAA post, you can see that 2x is the same with the two driver revisions. I haven't checked 4x yet.

edit: I checked 4x, and it is identical.

I will check any other driver revision you want me to as well.

The Baron
08-14-03, 09:38 PM
I will check any other driver revision you want me to as well.
And complain to me about you doing it.

I wonder... should they advertise "For forum mods--avoid answering silly driver questions about AA and AF quality!" on ATI boxes? ;)