3dfx Voodoo 5 vs LeadTek Geforce 3 (4x FSAA)

[Daemon_UK]

Surprises ME! :Q

IMHO, the screenies look much better on the V5. Much more detailed. :Q

Heres hoping on the NV25 or the NV30 3DFX engineers are given a free hand on the FSAA implemenation.

 

[EMAN]

Voodoo 5 has the best FSAA hands down but the card lacks the fill rate and memory bandwidth to compete with current video cards.

 

[Rand]

Havent even glanced at the article yet, but personally I've always felt the V4/5's FSAA offers clearly superior image quality then that of ATi/Matrox.
In the vast majority of non-GLide games, the V5 simply cannot perform even remotely close to the GF3 though... and I'd unhesitatingly rather the GF3 over a V5.

As a side note, IMHO I still feel the GF3 has the worst FSAA I've ever seen- including that of the older GF1/2.
Quincunx is worse then no FSAA it blurs everything so ridiculously badly, and even the 2X and 4X MultiSampled AA on the GF3 manages to look slightly worse then the classic 2X and 4X SuperSampled AA nVidia used on the GF1/2.
Of course one can remedy much of that with anisotropic filtering, albeit at a heavy performance loss.

 

[usas]

Your right, that is suprising, the Voodoo 5 still does seemingly has better FSAA quality than the Geforce 3. this is a testimate to the engineering team who came up with the idaea of the RGSS FSAA implimentation used. Wow, and the voodoo even sometimes performs better than the Geforce 3 (albiet sometimes in glide)

 

[Innoka]

I don't see what you can tell from the first pictures, the compression effect is awful.
How does one take screenshots with the V5500 anyway? I was informed the onscreen image is not in the framebuffer the way it is with other supersampling cards, it is combined with a duplicate on the way to the screen.
It was hard for me to believe how close the V5 is to the GF3 in the benchmarks.
The V5 2x setting was surprisingly useful in taming jagged lines in games not normally associated with the need for FSAA, like a Serious Sam. I would take it over 32 bit color in that case.

 

[bpt8056]

The Voodoo5 has been untouched in terms of AA quality and I remembered the first time my jaws dropped when I saw the aliasing lines disappear in Madden 2000.

I read a bit of the article and I can tell you firsthand that the benchmark scores in Nascar 4 is not accurate. I have a 1.2 GHz Athlon and I've been running better scores than what the article posted. The screen shot for the GF3 definitely does not have any anisotropic filtering or it is set at 8-tap. I usually run 1024x768 with Quincunx and 32-tap aniso and it really does look sharp and runs in the high 30s in the pack. I've seen Voodoo5's FSAA and an LOD bias of -1.5 and it looked frickin' great, but the performance hit was too much for my liking (21-23fps). It's amazing what 3dfx accomplished and no other chip manufacturer has implemented the same AA quality.

 

[BFG10K]

The number of errors in that article is appalling:

But Nvidia seems to have some significant problems getting this feature to work in Dx apps.

Err, no. Anisotropic filtering works just fine in Direct3D apps. In fact that feature, just like trilinear, is really completely transparent to the apps. They don't even need to know that the card is doing it.

Also why is he using anisotropic filtering when the Voodoo5 doesn't even support it? It'll completely ignore that setting. Also the Voodoo5 only does approximated trilinear filtering while the GF3 is doing true per-pixel trilinear. In otherwords the GF3 is working much harder in the same benchmarks so all of the results are completely invalid. Also who knows what other features those games are using which aren't even supported on the Voodoo5 (the various mapping modes come to mind).

Using the same settings the GF3 should be at least three times as fast as a Voodoo5 across the board yet in some benchmarks the Voodoo5 is beating the GF3.

That review was a total joke. I wish these "experts" would get a clue before posting so much misleading data.

 

[Orbius]

<< In otherwords the GF3 is working much harder in the same benchmarks so all of the results are completely invalid. >>



And yet the V5 manages to look better in 4x FSAA using 'approximated trilinear' than the Geforce 3 using 'per-pixel trilinear' so why are you complaining?

Perhaps if you dropped the GF3 down to bilinear it would totally look like ass.


People use the same arguments as ammo against the Radeon, and it was stupid then and its stupid now. I dont care if the Radeon 8500 is using connect-the-dots filtering and Nvidia is using Trilinear, all I know is that the Radeon produces a better image with a less costly method of rendering.

If you werent so biased you would be congratulating the Voodoo5 for looking so good without having to resort to 'per-pixel trilinear' and chastise the GF3 for producing an inferior image using that costly form of rendering.

In short yuo=fanboy. No surprise there for regulars of this board though.

 

[EMAN]

Orbius, I couldn't have said it better.

 

[EdipisReks]

who cares what the voodoo5 uses. the best looking FSAA on a consumer level card didn't do 3dfx any good in the end. as an aside, with 64 tap anisotropic and qunicunx anti-aliasing my geforce3 suffers from no blurriness or any other image problems.

--jacob

 

[EMAN]

who cares what the voodoo5 uses. the best looking FSAA on a consumer level card didn't do 3dfx any good in the end. as an aside, with 64 tap anisotropic and qunicunx anti-aliasing my geforce3 suffers from no blurriness or any other image problems.

Still voodoo 5 FSAA is better no matter how you look at it. Yeah yeah your geforce 3 looks good too.

 

[rogue1979]

I miss my V5, why couldn't a faster version have came out before the fall of 3dfx? I am having dreams about Nvidia incorporating Voodoo 5 anti-aliasing, could it really happen? I don't want to get my hopes up too much!

 

[Innoka]

"the GF3 is working much harder in the same benchmarks so all of the results are completely invalid"
aw come on BFG, the man is trying to get the best image quality from both cards, presenting the screenshots and benchmark comparisons. He's trying if the GF3 is a beneficial upgrade to him as a simmer. The problem is really that the screenshots are way too compressed and the benchmark figures look wrong.
At the moment I'd say the GF3 appears the main upgrade to consider from the V5 because of the antialiasing qualities. Right now you could perhaps live without it, wonder what the GF4 will bring...

 

[chizow]

<< who cares what the voodoo5 uses. the best looking FSAA on a consumer level card didn't do 3dfx any good in the end. as an aside, with 64 tap anisotropic and qunicunx anti-aliasing my geforce3 suffers from no blurriness or any other image problems.

--jacob >>



Product and image quality, driver stability and compatibility, performance, overclockability, included software, heatsinks.....and pretty much everything else that makes a card awesome to the normal consumer had nothing to do with 3dfx's downfall. It came down to the green, pure and simple. 3dfx ran out of deniro and went belly-up. You can fault their business strategy in making their own cards (leading to inventory costs they could not sustain or turn-over fast enough) or their attempt at targeting the enthusiast/retail channel (compared to nVidia/ATI who dominated the OEM market and had a nice piece of the retail market as well).

Point is, I own a V5, and a GF3 Ti200...the image quality for the V5 in FSAA crushes the GF3. That being said, I would never go back to a V5 for any FPS type game, as they move so fast that the 32-bit with all Max details at 1280 always look blurry due to the speed For older games however, I switch to my V5 (its a PCI in dually config) and wratchet up the FSAA. Turns a normally ugly game like Pool of Radiance into total eye-candy and makes a game that looks good normally like Fallout Tactics look great.

Chiz

 

[SCSIfreek]

As for 2d quality 3dfx voodoo5 has the best, but was limited by the fps


--SCSI

 

[AA0]

Why did this idiot benchmark in 16bit? Its not 1998 again is it?

Its never been a secret how bad nvidia's FSAA sucks, GF3 looks awful.