noshutdown wrote:

2. the cyrix 6x86 failed all opengl tests with nvidia cards, it worked with an ati rage128 but i don't want to post results with a different card.

Which driver versions did you try? NVIDIA detonator version 12.41 was the latest version I could get working on a Am5x86-160, so perhaps it will work on a Cyrix 6x86. What ATI Rage 128 versions card did you try? 128 GL, VR, Pro?

I did not know winrar had a benchmark. How does this work?

Is Winquake the same as GLQuake?

feipoa wrote:

Which driver versions did you try? NVIDIA detonator version 12.41 was the latest version I could get working on a Am5x86-160, s […]
Show full quote

noshutdown wrote:

2. the cyrix 6x86 failed all opengl tests with nvidia cards, it worked with an ati rage128 but i don't want to post results with a different card.

Which driver versions did you try? NVIDIA detonator version 12.41 was the latest version I could get working on a Am5x86-160, so perhaps it will work on a Cyrix 6x86. What ATI Rage 128 versions card did you try? 128 GL, VR, Pro?

I did not know winrar had a benchmark. How does this work?

Is Winquake the same as GLQuake?

i used driver 30.82, it managed to run 3dmark01 and a few other d3d games, its the opengl icd that failed. i also tried 8.05 and its opengl icd failed the same way, with 3dmark01 also crashing. rage128 was rage128vr, i think rage128gl and rage128pro would work too.

winrar has its own benchmark, just in the menus.

winquake uses software rendering, which is just an official win32 version of dos quake, no 3d acceleration involved.

Do you want to give version 12.41 a shot?

Which version of WinRAR are you using? I've been using WinRAR version 3.0 for over 10 years, but do not see a benchmark option in the menus.

feipoa wrote:

Do you want to give version 12.41 a shot?

Which version of WinRAR are you using? I've been using WinRAR version 3.0 for over 10 years, but do not see a benchmark option in the menus.

i don't think it would do any good. did your nvidia card just load windows on the 486, or tried some opengl games? what card is it?
on the 6x86 with geforce2, anything that invloves opengl icd, including just checking icd info in everest, leads to bsod.
i used winrar 3.80. the benchmark option is the last item in "tools" menu, not sure if its in 3.0 though.

I was able to find working OpenGL drivers on my 486 using the GeForce2 MX400. GLQuake and Quake II both finish to completion. I have found working Win9x and WinNT4 drivers. Only one or two versions worked. I have found that the the original Cyrix 6x86 had issues with a lot of graphic card drivers and usually the driver version which worked on a 486, also worked on the 6x86.

WinRar 3.0 doesn't have a tools menu.

The K6-3+ 550MHz 3DMark2000 score of 2220 is lower than I expected.
I used to get about 3000 on an MVP3, K6-3 450MHz, Geforce2MX. Unless my memory is off by 1000pts, but I don't think it is.
The differences I see are that my system back then was Win98SE, yours is Win2k SP4. I was also using an older nVidia driver, v5.32 I believe.
I do remember seeing slower performance with a later nVidia driver, but by later I mean something like 6.x or 7.x. I never tried a version remotely close to the one in this test.
Do the drivers and Windows version possibly explain the difference? I don't think I ever tested that system under Win2k, so I don't know what it would have done there.

shamino wrote:

The K6-3+ 550MHz 3DMark2000 score of 2220 is lower than I expected. I used to get about 3000 on an MVP3, K6-3 450MHz, Geforce2MX […]
Show full quote

The K6-3+ 550MHz 3DMark2000 score of 2220 is lower than I expected.
I used to get about 3000 on an MVP3, K6-3 450MHz, Geforce2MX. Unless my memory is off by 1000pts, but I don't think it is.
The differences I see are that my system back then was Win98SE, yours is Win2k SP4. I was also using an older nVidia driver, v5.32 I believe.
I do remember seeing slower performance with a later nVidia driver, but by later I mean something like 6.x or 7.x. I never tried a version remotely close to the one in this test.
Do the drivers and Windows version possibly explain the difference? I don't think I ever tested that system under Win2k, so I don't know what it would have done there.

i tried the 5.32 driver(last version in 5 series anyway), but the result is abysmal. cpu is k6-2+550, just compare with my previous results:
3dmark2000: 1420pts
3dmark01: 590pts
winquake: 48fps
quake2: 70fps
quake3: 6.1fps
now i wonder if you still have the rig and can post a screenshot of k6-3-450 and gf2mx scoring 3000pts, or that you have taken 3dmark99 score as 3dmark2000. 😈

feipoa wrote:

I was able to find working OpenGL drivers on my 486 using the GeForce2 MX400. GLQuake and Quake II both finish to completion. I have found working Win9x and WinNT4 drivers. Only one or two versions worked. I have found that the the original Cyrix 6x86 had issues with a lot of graphic card drivers and usually the driver version which worked on a 486, also worked on the 6x86.

WinRar 3.0 doesn't have a tools menu.

i have read you reports of 3d cards on 486 rigs, and you said that while some nvidia drivers managed to run with amd486, they still failed with cyrix5x86, i think thats how it goes.

noshutdown wrote:

i have read you reports of 3d cards on 486 rigs, and you said that while some nvidia drivers managed to run with amd486, they still failed with cyrix5x86, i think thats how it goes.

The outcome has a high probability to be the same with 6x86, but it is still not certain. It could be that the NVIDIA drivers took the 6x86 into consideration, especially if it is an older card like the original RIVA TNT. It is possible that the drivers just do not work with the Cyrix 5x86. To be conclusive, you would have to test for this specific scenerio.

noshutdown wrote:

i tried the 5.32 driver(last version in 5 series anyway), but the result is abysmal. cpu is k6-2+550, just compare with my previ […]
Show full quote

i tried the 5.32 driver(last version in 5 series anyway), but the result is abysmal. cpu is k6-2+550, just compare with my previous results:
3dmark2000: 1420pts
3dmark01: 590pts
winquake: 48fps
quake2: 70fps
quake3: 6.1fps
now i wonder if you still have the rig and can post a screenshot of k6-3-450 and gf2mx scoring 3000pts, or that you have taken 3dmark99 score as 3dmark2000. 😈

It was definitely 3DMark2000, that much I'm sure, but I could be misremembering the score. I wonder if the OS version could also be the difference (I was on Win98SE back then). I remember needing a particular combination of drivers to get the best performance (and stability) out of it. Needless to say, my first impression of AGP and VIA wasn't very pleasant.

I remember the 3D2000 score was in-line with most of the K6-3/GF2MX systems that had been submitted to Madonion's online database at the time. Unfortunately that database got pulled off their site like 6 months later so no way to check it now.

That system was long ago disassembled, but I still have the motherboard and CPU. I also have a Geforce2 MX card but not the same one. The card I had back then was a Hercules model with 183MHz SDRAM. The cards I have now are the standard 166MHz type, but it probably shouldn't make a big difference.
I'm a bit logjammed with incomplete projects, so I'm not sure when I'd be able to put the K6-3 together, but I'll keep it in mind and I'll definitely get a screenshot if I get it put together and tested.

hello
i put up another platform: vp3 motherboard(512kb onboard cache disabled because it can only cache for 128mb anyways), 256mb sdram, k6-2+ running at 366, and geforce2mx. i installed win98se and tested 4 driver versions: 5.32, 8.05, 23.11 and 30.82.
in 3dmark2000, 23.11 and 30.82 both scored around 1350pts, which is inline with win2000 30.82 performance(k6-2+550 at 1980). 5.32 and 8.05 are indeed very fast, and both scored over 1900pts, so it seemed that k6-3-450 getting around 3000pts is reasonable.
however, 5.32 and 8.05 both scored around 610pts in 3dmark01, 23.11 scored ~720, and 30.82 scored 800. so using 30.82 is not a bad idea either.
5.32 for win2000 performed badly because its among the earliest versions to supported win2000, so it was far from mature. by the time of 30.82 win2000 performance had surpassed that of win98.

results updated! most socket7 cpus are included now.

noshutdown wrote:

i tried the 5.32 driver(last version in 5 series anyway), but the result is abysmal. cpu is k6-2+550, just compare with my previ […]
Show full quote

i tried the 5.32 driver(last version in 5 series anyway), but the result is abysmal. cpu is k6-2+550, just compare with my previous results:
3dmark2000: 1420pts
3dmark01: 590pts
winquake: 48fps
quake2: 70fps
quake3: 6.1fps
now i wonder if you still have the rig and can post a screenshot of k6-3-450 and gf2mx scoring 3000pts, or that you have taken 3dmark99 score as 3dmark2000. 😈

i put up another platform: vp3 motherboard(512kb onboard cache disabled because it can only cache for 128mb anyways), 256mb sdram, k6-2+ running at 366, and geforce2mx. i installed win98se and tested 4 driver versions: 5.32, 8.05, 23.11 and 30.82.
in 3dmark2000, 23.11 and 30.82 both scored around 1350pts, which is inline with win2000 30.82 performance(k6-2+550 at 1980). 5.32 and 8.05 are indeed very fast, and both scored over 1900pts, so it seemed that k6-3-450 getting around 3000pts is reasonable.

I can't believe it's been so long, but I finally have my old super socket 7 system back together, and I've been doing some benchmarking with it.
I think I've basically recreated the setup I had back then. It's not technically the same video card - my card back then was a Hercules with faster RAM, but what I tested here is a standard MX.
I don't remember what version of the AGP driver I had back then, so I made a best guess.

Tyan S1590 (MVP3 chipset, 1MB cache)
K6-3 450MHz
256MB PC100 CL2
Geforce2 MX AGP (default clocks: 175MHz core, 166MHz 128-bit SDRAM)

Win98 SE
VIA AGP driver 4.17 installed in "normal" mode (which I believe just means it stays in AGP1X, not 2X)
nVidia driver 5.32

3DMark 2000 v1.1 default benchmark (1024x768x16, hardware T&L)

2873 points
My card in those days had faster RAM, but this is a standard model and it comes close.

I remember that back in the day, the score tanked by about 500pts if I installed a later driver, so I stayed with 5.32.
5.32 had some graphical glitchiness at first - the solution for that was to install the VIA AGP driver in "normal" mode. That option was only offered in older VIA drivers, so that's why I used an old VIA driver along with the old nVidia driver.
Later nVidia drivers would fix the graphical glitches, but they also slowed down the card significantly so I preferred to get it working with 5.32.

That info is from a long time ago, I haven't tried experimenting with driver versions on the current build. I might still experiment. Interesting that you had good results with 8.05, I might try that out.
I'm suspicious that maybe nVidia 5.32 was optimized better for the K6, and then they changed something shortly after that when they didn't care about the K6 anymore. Or another possibility - maybe K6 optimization has nothing to do with it and it's an issue with how they addressed the glitchiness of VIA AGP.

The Hercules "3D Prophet MX" that I had back in ~2001 ran the GPU at standard 175MHz but had faster RAM clocked at 183MHz.
I don't have that card anymore, but I overclocked my current card to those settings to match.

2954 points - it's only a modest difference but that's the scoring I remember getting.