DaveB3D states that the V5 is as fast as the GTS...

[*kjm]

I have an Athlon 550 and the V5 5500 and used the settings. I dont have Q3 so I tested it with Tribes and MDK2. My FPS took a big jump and with the LOD settings things look much better......thanks dave and Wingznut PEZ!

 

[hans007]

i've got a v4, the new drivers make it almost as fast as a gf2 mx, and its kills the mx in 2xfsaa which i use so its a real benefit. THat and the 2d quality is way crisper with way sharper text in web browsing etc. My 2 geforce2s i tried sucked for that and its real important, my old v5 was good too, but i didnt really need it so i sold it, the 3dfx cards are great cards, just not the fastest, but i'm not in this for a dick measuring contest anyways

 

[GL]

My Geforce2 MX 2D quality is great (and I've used everything from an ATI Rage Pro, to a TNT, to a Matrox G400 MAX). I believe the 2D quality of the Geforce 2 MX is better than standard Geforce2s. Someone back me up on this....I believe it had something to do with a good side-effect of hardware used for twinview.

-GL

 

[ElFenix]

GL- it has almost everything to do with the quality of the capacitors used as filters between the video out of the ramdac and the output port of the card, if i've read correctly. some manufacturers use cheap caps, so the quality is poor. nvidia can't control it. since 3dfx and matrox are the sole soure of their boards, they can control it.

 

[Wingznut]

You are most welcome, *kjm!

I posted some benchmarks that I ran today, on another thread. I figure I could throw them up here, as well.

My system
P3-700e, overclocked to 933mhz
Asus CUSL-2
256mb generic RAM
V5-5500 AGP, 1.04 drivers with the previously mentioned (by Dave) advanced option adjustments

Q3, High Quality
1024x768x32: 80.8 fps
1280x1024x32: 51.1 fps
1600x1200x32: 33.7 fps

 

[Weyoun]

those are some impressive numbers people are quoting here. dave i seriously think if these options don't harm visual quality, then they should be turned on by default. this could give all those people who couldn't be bothered tweaking (ok, no hard task i know, but still....) can have their performance anyway. just send a memo to a few friends at 3dfx who write drivers, I'm sure they'll welcome the thought....

 

[mikelou26]

haha

 

[Volenti]

Ok I did some testing with my V5 and came up with these results;
System;

p3 600@800 on a bx
192meg cas2
V5@183, latest DX 8 drivers, 1.04 or whatever they are

Q3 verson 1.17, default config, standard HQ setting at 1024x768 with the texture slider all the way to the right.

refresh optimization off, depth precision disabled;

001, 74fps
Quaver, 65.8fps

refresh on;

001, 75fps
Quaver, 67fps

depth precision fastest;

001, 81.6fps
Quaver, 73.1fps

both enabled;

001, 81.7fps
Quaver, 73.4fps

above settings with texture compression off;

001, 68.7
Quaver, 55fps

From watching the demos run it seemed to me that the depth precision thing seemed to make the faster running parts of the demo run even faster, rather than improving the speed of the slower parts (showers of blood ect) but's that's just the impression I got, I could be wrong...

I don't actually play Q3, so I don't have any real "gameplay" experence to tell weather it's smoother/better in a "real world" suitation, most of the other games I play (apart from UT) are pretty cpu limited and are played with 32bit + 4xFSAA.

 

[GT1999]

GeForce/TNT Image Quality Problems Due to RFI Filtering

 

[Ahriman6]

Reading this thread really makes me wonder how Sharky could call the V5 a "dog". Hmmmm. . . .

 

[DaveB3D]

drop him an email and ask him.

 

[Wingznut]

I have a couple theories...

Most websites seem very content just to run a few benchmarks when comparing video cards. They rarely look into things such as compatibility, stability, image quality, etc... And they certainly don't look into the advanced options in the drivers. Which is kind of odd, since these same websites will spend a significant amount of time going over how well a nVidia card will overclock, and the results of it. Yet, they can't seem to find the time to look deeper into the 3dfx options.

These are also the same websites the delve deeply into each nVidia driver release, even the betas... Yet don't spend a minute with the very significant 1.03 V5 driver release.

It's very "in style" right now, to jump on the "bash 3dfx" bandwagon. Maybe it's just to fit in with the others. It's really quite overwhelming the number of people who'll cry "3dfx sucks", yet they've never used a V5. Hell, many of them have never used a GTS either. I guess they've heard it enough times, that it MUST be true.

Maybe if a popular website would really do their homework and say that the V5 is an excellent card, they'd stand out. Maybe people would think, "That website must not know what they are doing. Other websites say the V5 sucks." So, maybe they don't want to go out on a limb.

 

[DaveB3D]

Well I tried to be as fair as possible when I reviewed the V5. I think I did that pretty well. Unfortunately, at that time these options were not present in the V5 drivers so there wasn't much I could do.

I've been talking with some people at websites lately about the stuff though. Basically trying to raise awareness. That is why I can here and talked about them. I know a lot of people come here, so guys can raise awareness at other places. I talked to Anand the other day about it, and I'll be sure to catch up with him again when he gets back from Comdex.

 

[Wingznut]

Way back at the beginning of this thread, xtreme2k threw out some of AnandTech's UT benchmarks. I still have no clue as to why AnandTech thought it'd be a good idea NOT to use the V5 to the best of it's ability, but they did use D3D with the V5 instead of Glide.

Btw xtreme2k, have you changed your position any regarding the V5 being competitive with the GTS?

 

[ElFenix]

back when anand first tested the v5 in UT there was a bench that showed that glide didn't actually help much, D3D was faster at high res, and the 32 bit scores were just wrong on glide.

 

[Wingznut]

Well Glide doesn't truly support 32bit. And from what I understand (and what I've seen), UT D3D doesn't really support 32bit color.

But do trust me when I say that it runs a lot better in Glide. I highly doubt there's a V5 owner who plays UT in D3D.

 

[KarlHungus]

I have to say that if these advanced settings are doing what I think they're doing you'll definitely see a drop in visual quality. Depth precision fastest sounds like 3dfx is forcing the game to use a 16 bit z-buffer. While I am all for the performance increase I really think that comparing these results to a card that uses a 32 bit z-buffer is a little loaded...

 

[Wingznut]

Seriously. I cannot see any difference whatsoever.

 

[DaveB3D]

No, 3dfx is not forcing a 16-bit Z at all. 3dfx is very clear about this: In 16-bit color you get a 16-bit Z, in 32-bit color you get a 24-bit Z and 8-bit stencil.

 

[Topshelf]

Me either. I've played all the normal games I play with the new settings and there is absolutley no difference i can tell. Maybe if I took a screenshot and blew it up 300x I could see something, but I'm not going to do that. I couldn't be happier with my V5. Thanks.

Oh, and what is the best LOD setting with 2xFSAA?

Oh, and Dave, could you possibly if you get the time post these settings you mentioned before?

1) best performance
2) best image quality
3) best balalnce

Thanks again

 

[Wingznut]

I prefer -1.5, but it's up to your own personal preference.

 

[DaveB3D]

yeah, sure.. give me 10 minutes.. I'm so nice, I'll pause what I was doing for a minute.

hehe

 

[KarlHungus]

No, 3dfx is not forcing a 16-bit Z at all. 3dfx is very clear about this: In 16-bit color you get a 16-bit Z, in 32-bit color you get a 24-bit Z and 8-bit stencil.

I know 3dfx took this position in the past. Can you confirm it again with the new drivers? I only ask because everyone knows you work there.

 

[DaveB3D]

There is NO question in my mind about in being 16/16 and 32/24+8 as I said.

For the settings.

Leave the setting at default unless I otherwise say:


Performance:

VIA Optimization: Performance
Refresh Optimization: Enabled

D3D:

3Dfilter Quality: Normal
LOD BIAS: +2.0 (will reduce image quality a lot in lower resolutions)
Z-buffer Opt: Enable


GL/Glide


3Dfilter: Normal
Depth Precision (16 and 32): Fastest
Legacy Texture Compression: Enabled
LOD Bias: +2.0
V-sync: disabled


---------

Image quality:

(D3D/Glide/GL)
3D Filter Quality: Auto (could do high.. I say stick with auto)
Alpha Blending: Adjust according to each app.. whatever gives best quality
Anti-aliasing: 4x
LOD bias: adjust accordingly
Mip-map dithering: adjust accordingly to app (I like it on typically)
TC: disabled (though the quality loss is minimal with it enabled)

---------
My settings:

I use a mix of performance and quality. I enabled all the performance features, and then adjust image quality features based on each app.

---------
So I think that gives you a general idea. This is just my personal view though, so don?t take it as coming from 3dfx or anything official like that.

 

[KarlHungus]

Thanks.