Hello everyone,

I hope that you enjoyed christmas and had a good transition to this year.

It's 2018 now, more than 17 years after 3dfx's demise. But we're still here! And there are some German geeks that are still writing articles about 3dfx. They were even doing brand-new benchmarks with Voodoo cards and their competition. Please have a look at this retro special, which has been translated to English for your convenience:

http://www.pcgameshardware.de/3dfx-Voodoo-Hardware-255488/Specials/3dfx-Voodoo-Retro-Special-1246582/

The article contains some history, an interview with Oscar Barea (you all should know him) and, what is the most exciting thing, new benchmarks with games from 1999 to 2005. To make it even more special, a modified Voodoo 5 5500 AGP with 128 megs @ 210 MHz is included in the benchmarks. The results and analysis tell us what AGP and Hardware TnL were really worth - and much more.

Have fun! And please spread the word. Take all people in the world on this retro trip. 3dfx may be gone, but the audience is still listening.

Greetings from Germany,
Raff

QuoteOriginally posted by m14radu

Nice to see you back Raff !
Btw: Do you plan to update your voodoo5 6000 biggest test, using the new SFFT releases ?

Heya!

Hm, I don't think so. But there's still a big comparison of maaany old cards on the agenda, starting from Voodoo Graphics to ... well, the top of AGP.

___

Concerning the drivers: The SFFT 1.9 popped in, yeah! I've done some quick tests of the SFFT 1.6 in comparison to 1.9 on my V5 5500 AGP @ Windows XP x86. Focus: Direct3D, since that's one of SFFT's main goals.

UT2004: Fine all the way
Unreal 2: Graphical corruption starting in the main menu (SFFT 1.6 and 1.9), couldn't start a game
Flatout 2: Some objects are green, the rest is fine with SFFT 1.6. Version 1.9 shows some lining corruption in the starting videos, looks like a scanline bug.
Max Payne 1 & 2: Perfect!

What does the Command FIFO thing in the SFFT configurator do?

Greetings from Germany,
Raff

PS: Release notes, pleeeease.

Thanks for your push.

I'm familiar with many details about Rampage/Spectre because I've done much research for an article. Unfortunately, the greatest problem is getting a card. Running and writing a DX8 shader in comparison is an easy task ... There are some more games running (see NextHardware for that).

But it seems that nobody is interested in providing hardware to clear some of Rampage's myth.

Greetings from Germany,
Raff

SFFT, great to see you're back!

But, please, with sugar and a sweet cherry on it: Try to show us release notes of new drivers. It's impossible tu guess what you're doing. We could give much better feedback if you'd tell us what you're tweaking on right now.

Thank you very much and please keep it up!

Greetings from Germany,
Raff

Hey guys,

I know it won't be easy, but I'm completely serious about it: I want a Rampage.

Not any Rampage, a working sample. One of these famous Spectre cards in A1 revision that are said to be running. I've already asked Gary, but I can understand if he's not willing to sell his legend. Unfortunately I can understand anyone doing this. Your all are fans and collectors, it won't be easy at all.

I have a simple but "mighty" purpose in mind: testing this card. Since I'm editor at a quite big magazine and 3dfx fan for many years, I want to tell the world about this myth. The "goddess" V5 6000 is through, we know about the quality of this legend, but Rampage/Spectre still is a mystery. I just want to run some benchmarks and at least one DX8 demo on it!

This may sound crazy, but: If nobody is willing to SELL, maybe someone can LEND his working sample? Are there any sources to get one of these cards? I repeat: This is no joke.

Greetings from Germany,
Raff

PS: If anyone owns a WORKING Daytona card - that's a bit easier - it could get my attention (and money) too.

Yeah, I'm also hoping that version 1.5 won't be the last release ... it's already 6 months ago!

SFFT, are you still developing your code?

Greetings from Germany,
Raff

The game is PCI limited, I assume. Try raising the clock.

Greetings from Germany,
Raff

http://www.pcgameshardware.com/aid,670546/News/A_tribute_to_3dfx-_Eight_years_out_of_business/



MfG,
Raff

I hope it's not your last one, especially concerning driver deleopment.

Greetings from Germany,
Raff

What abou newer drivers? Do they solve the crashing?

Greetings from Germany,
Raff

Well, 3DMark2000 boosted with Alpha 59. But then I started Max Payne 2 again - and saw this:

http://www.keepmyfile.com/download/9e3f212332563 (MaxPayne2.AVI)

Rainbows! Extreme artifacts. Interestingly, "only" the static objects look psychedelic, while Maxey is textured correctly. TrackMania Nations forever shows comparable texture corruption ... just to mention other games. It's definitely an omnipresent problem. But in these cases, SSAA doesn't solve the problem, the games only drop Fps (and jaggies).

SFFT 1.3 returns to the old fps level, but it doesn't solve the rainbow bug - no driver does it. I don't know if this is possible, but, SFFT, you should try to "migrate" Alpha 58 and 1.3. Only you know what you did in the last Alphas. You were on the right path, on the way of killing one of the last major bugs!

And please: Keep up your great effort! I bet you can kill this "trouble maker" in the driver. Nobody else could do it!

Greetings from Germany,
Raff

SFFT, I don't know what you changed in the last drivers, but I think it's only partially the right way. Let's recall the symptoms of the SLI f*uckup: Every Direct3D game became cruelly slow when you set the resolution to 1.152x864 or higher. 1.024x768 and less ran fine. Always - the bug only occurred in high res and only under XP with the V5 6000. The V5 5500's Dual-SLI doesn't suffer from this bug.

With the last SFFT Alpha's you did some serious changes that affected the bug in a positive way. Using the Alpha 57, I already noticed some apps scaling in high res i.e. the 3DMark2000. It was amazing: The 6k finally surpassed the 5500's fps by factor 2,x! But the bug didn'd vanish in many cases. Max Payne 2, to mention one game, still creeped at ~5 Fps.

With Alpha 58 there was another change. With this driver, Max Payne 2 initially ran extremely fast - SLI worked the way it should! But the mysterious "square" that 6k users know from older drivers returned. This is it:




A blue-coloured artifact that sits "in front" of the picture. You can't grab it on a screenshot. It's only shown in resolutions above 1.024x768, so I assume it has something to do with the SLI bug. What is this? Could you kill it again?

Then Alpha 59 came and even 1.920x1.440x32 and 2.048x1.536x32 in 3DMark2000 ran fluenty - while Alpha 58 freezed them. The square was still there ... as it is with every new release since then. Enabling SSAA makes this artifact vanish! Then it must have something to do with the SLI communication, I think. But I'm a lay man.

I did some investigation this weekend. The aim of my driver testing was the answer for this question: "Which driver handles Quad-SLI the best way?" SFFT, as we noticed for some time, you're trying to kill the V5 6000's SLI f*ckup. That's great! I'll try to give you some information, whether you're on the right way or not.

I've tested the last 6 SFFT releases with my V5 6000 (Rev. A-3700) using Windows XP x86 SP2. Let's look at some Fps first. It's Max Payne 2 in 1.600x1.200x32, no SSAA:

SFFT 1.3: Avg: 55.133 - Min: 37 - Max: 88
SFFT 1.2: Avg: 48.700 - Min: 35 - Max: 76
SFFT 1.1: Avg: 47.667 - Min: 34 - Max: 76
SFFT 1.0: Avg: 55.233 - Min: 36 - Max: 87
Alpha 59: Avg: 55.533 - Min: 36 - Max: 87
Alpha 58: Avg: 52.733 - Min: 18 - Max: 87
Alpha 57 and older: <10 Fps
As you can see, SFFT 1.1 and 1.2 lose the comparison, while SFFT 1.3 regains the speed of older drivers. The Alpha 58 seems to be one of the worst drivers. But that's not true, 'cause these numbers don't tell the truth.

Ah nice, SFFT now also works on Glide drivers?

Greetings from Germany,
Raff

According to Entech and fillrate tests done by me, Powerstrip is also able to overclock all Voodoo GPU's.

And yep, you're right - it's the PCI bus that keeps the card from rocking. Try a higher resolution or with SSAA enabled, then the card should show a nice scaling. Higher PCI clock doesn't kill the card either.

Greetings from Germany,
Raff