retro games 100 wrote:

an AGP fx5200 graphics card.

Kill it with fire!

leileilol wrote:

retro games 100 wrote:

an AGP fx5200 graphics card.

Kill it with fire!

Hehe!! Is it really that bad? I thought those pcpbench speeds were quite impressive. And it plays Quake 2 quite well.

retro games 100 wrote:

leileilol wrote:

retro games 100 wrote:

an AGP fx5200 graphics card.

Kill it with fire!

Hehe!! Is it really that bad? I thought those pcpbench speeds were quite impressive. And it plays Quake 2 quite well.

So does 3dfx Voodoo 1.

And yes, it is that bad.

And nobody runs t-bred 1700+ @ default 😊

How many times did you run those tests to find a mean score? If their one-shots I suppose the variation could be within a margin of error. If I clock my Barton (T-Bred with double cache) at 1466mhz it gets a speedsys result of 1712. Maybe you could do that and compare the results. I know that T-bred's can clock much higher on the FSB where the Palomino it starting to to out, it's logical to think they tweaked the micro-architecture in such a way that yes, the Palomino is more effective at these speed.

The most telling difference between the two is/was better efficiency -- and there were two Thoroughbreds, an A and a B. (I'm quite certain that the Barton was on much different silicon than anything before it, coming out as it did, after the K8s were released.)

I recall a Palomino 2100 needing a half again larger cooler size than a T-bred B 2200 (the latter of which is still in use as a loaner, on a Biostar MB with a Via KT-266 chipset).

prophase_j wrote:

How many times did you run those tests to find a mean score? If their one-shots I suppose the variation could be within a margin of error. If I clock my Barton (T-Bred with double cache) at 1466mhz it gets a speedsys result of 1712.

I tried running pcpbench 3 times, for each CPU. The 3 scores for each CPU were the same, plus or minus < 1.0

BTW, you say you got a score of 1712 - that's huge. Do you mean 171.2?

I did my test using speedsys. It's okay though, when I first read your post and tested it with mine, I was wondering how you got such a small result 🤣. Since I don't have a comparable card to test with, the PCP scores don't directly compare. If you have a Voodoo 5 AGP I can test mine and then we'll be "apple to apples".

I'll dig out an AGP V5, and redo my tests...

In the meantime, I just ran speedsys (windows 98 -> reboot to dos), and it gives me a score of 1752.

I ran winquake (non GL quake 1, 1024x768 video res, win98), and using the "timedemo demo2" console command option, it gives me 60.4 fps which ain't bad.

BTW, this fx5200 AGP fanless contraption isn't that bad, cos it's 128-bit and doesn't seem too sluggish.

Lastly - whenever I quit Quake, it always freezes the PC. In fact, I have to switch off at the mains power. Sometimes this action "scrambles" the mobo's BIOS, and I have to reset the mobo's FSB value back to 133, from it's default setting of 100.

I just swapped out the AGP FX 5200 card, and replaced it with an AGP Voodoo 5500 card. Its BIOS is 1.06. I am using the 'basic' Voodoo CD for its driver and 3dfx tools application. I haven't messed about with the card or 3dfx tools in any way - so no "overclocking".

pcpbench gives me a score of 219. (Like all my pcpbench tests, I run it in win98, and like most of these tests, the graphics look "heavily corrupted".) I noticed in the mobo's BIOS setup area, that the AGP "speed" had been reduced to 2x. (The FX 5200 card allowed 4x.) Naively, I increased the "megabyte thing" from 64 to 128 to see if that helped matters, and then re-ran pcpbench. The score dropped to 217.

BTW, speedsys gives me a score of 1762.

Cool, I'll get my Voodoo going tonight and I'll do the same, even though your board is slightly different I don't expect there to be much difference, at least when it comes to the main chipset and the way it is initialized/programmed. I was also thinking there could be differences in the way the chipset is programmed by the BIOS, I wonder if there is an easy way to dump the config registers in plain DOS, as that would be an easy way to compare BIOS settings and the others you don't have access to unless your doing manually with WCPedit or a PCI patch or something. My result by the was in Pure DOS 7, Windows 98 + F8 + Command Prompt. I'm sure the Voodoo will fare pretty good though, I had tested it once before and it was one of the best as far as VESA results go, although my TNT1 based card was the leader by a sizable margin.

Quick follow up -

I ran winquake again, this time using the V5, and in the same manner as before -- non GL quake 1, 1024x768 video res, win98 -- and using the "timedemo demo2" console command option it gives me 35.3 fps.

BTW, inside the mobo's BIOS setup area, I have tried to "max everything out" by setting all of its crazy-sounding settings to "Enabled". However, I haven't touched the CPU settings - they are simply set to "133" (not 100), and all the other associated variables (voltage stuff) are set to "Default".

Incidentally, I've never tried to overclock a CPU. Perhaps it's time I give it a go! 😀

Sorry to return to Quake 1 yet again -

I DL'd a 3dfx patch from id software, and after installing it, when I launch Quake inside Win98, I see the 3dfx splash screen logo. I've created a .bat file to launch 3dfx glQuake using these parameters: -width 800 -bpp 16, and my timed demo test gives me a 75.2 fps score. I would really like to test the 3dfx glQuake game using -width 1024 -bpp 16 parameters, but I get an error saying: "Could not initialize GL". The opengl32.dll file inside the Quake directory seems very much up-to-date (8/3/01), so I don't understand why this error occurs.

Try nuking opengl32.dll from the Quake folder. It should then pass to 3dfxvgl.dll.

AGP slot voltage may play a factor in your screen corruption. If your newer AGP card doesn't corrupt but the older one does, you'll know.

Solved - I removed the current up-to-date opengl32.dll file inside the local Quake folder, and replaced it with an older version from the "mini gl version 1.47" package, available off the net. I can now get 3dfx gl quake to run in 1280 resolution, and the "timedemo demo2" console command gives me 58.4 fps (using 16 bpp), which is very good. (1024 resolution gives me 75.1 fps)

I don't seem to have any 3dfxvgl.dll on my system.

Also, both AGP FX 5200 and Voodoo 5500 cards display corrupted graphics using pcpbench - garbled pixels, constant flickering, and you can see the stars behind the ground.

retro games 100 wrote:

I don't seem to have any 3dfxvgl.dll on my system.

Really? 3dfx has been shipping that DLL with their drivers since early 1999 when q3test came about.

leileilol wrote:

retro games 100 wrote:

I don't seem to have any 3dfxvgl.dll on my system.

Really? 3dfx has been shipping that DLL with their drivers since early 1999 when q3test came about.

I reinstalled the 3dfx drivers from the "Quick Install: Voodoo Driver installation CD" - (c) 2000, version 1.3. (However, this CD probably came with a PCI V5, and not with the AGP V5 I'm currently testing.)

I then did a search for 3dfxv*.dll, and no files were found on my system.

With the corrupted screen in PCPbench thats normal, I had the same problems when I ran my tests. Just a quirk of the program. I do remember that the V5 exhibited the least amounts of flickering. Sill getting the test setup toether, should be done tonight.

I just want to draw people's attention to this thread -

Most demanding DOS game, resource wise? (see half way down, and then more results continue throughout this thread)

, as it has some interesting & relevant pcpbench test results.

didnt the v5 have a full opengl icd hance no 3dfxvgl.dll ?