I did not see any differences during the first 2 tests with 1.72 and 1.82.
But i had stability issues in test 4. Turned out it wasn't the agp driver but the nvidia driver 7.58, after installing 7.76 the problems were gone.

Ok @ 550MHz I got 4381 points. And no difference in score when I run whole test. I am using 3DMARK version 1.1 as used on HWBOT

Guys, if you're looking for a decent cooler for your K6s, you may want to try this one:

https://www.youtube.com/watch?v=XLOI3_2-_Ug

I've used it with great success, it is very small in size and will fit all motherboards. It is a little noisy though, so try to get a fan controller along with it 😁.

Will start fresh, this time on a p5a-b 1.04 that i need to test anyway 😎

Also by luck i found 2 gf2 gts cards from another brand so hopefully this combo works well, otherwise the gf3 will be back in.

Will not be breaking any records with a k6-2+ but i'll do my best to get closer and atleast go way over 4000 points.... if i can't get that with this board i will try the Aopen ax59pro.

meljor wrote:

leileilol wrote:

What about 3Dmark 2001?

I think that is a bit too heavy on a k6?

With a Geforce3 in play, it definitely isn't.

Well I am going away for a week. When I get back one of you better have broken the world record 😎 😊 😀

Some screens from my system:

Settings:

For the other K6, I don't have made benches there with 3DMark. Also by requiring CPU-Z you force anyone to have at least Win98SE...
This is not so suited for such old systems.

isnt that weird how even l3 cache has a rather low throughput here? or is 320 mbps good for systems this old?

and why is fastwrite disabled? 😮

I don't remember, likely because it got slower.

No screenshots as long as i'm not breaking records (still working on it).

Highest score for now is p5a-b, 5,5x 110fsb, k6-2+, 128mb pc133, geforce3 ti200, agp 1.82, nvidia 7.76
4190 3dmarks 3dmark 2000 stock.

Tried 115fsb (slower!), 120fsb without l3 cache was slower, agp 1.72 no difference, nvidia 7.58 was a tiny bit slower (so gts was faster with 7.76!), gf2 gts was a little slower as the ti200.

With the same cpu, memory, graphics card, drivers and psu this particular p5a-b is a little bit faster as the p5a that i tested. Both with same bios settings and latest beta bios.

Next up is the Aopen ax59pro... Once i know what the fastest board is i might take another shot with the k6-3+.

meljor wrote:

No screenshots as long as i'm not breaking records (still working on it).

Ah sorry, I didn't aimed for the stars. I thought there was some general interest in K6-2+ system speeds.

there is; the k6 was one of the more interesting machines, first a 350, later a 550; but then again i always kinda grew attached to machines when they became too weak to use contemporary software

elianda wrote:

meljor wrote:

No screenshots as long as i'm not breaking records (still working on it).

Ah sorry, I didn't aimed for the stars. I thought there was some general interest in K6-2+ system speeds.

Sure! That was about myself... as long as i am not breaking records i do not provide the screenshots as i will try and do better anyway. 🤣

Hmmm, with the same components i scored 3800 on an Aopen ax59pro. Have to look for the best via agp driver and see if that helps any.

Agp is a bit more finicky on this one as the gf3 ti200 started to act a bit weird on colors on this board, but finished the benchmark anyway. Gf2 Gts was fine.

mrau wrote:

and why is fastwrite disabled? 😮

So I rechecked this and could not see any performance influence by FastWrite On/Off with the Quadro2 Pro.

kanecvr wrote:

Skalabala wrote:

Yes, Geforce 2 GTS Install in this order. […]
Show full quote

Yes, Geforce 2 GTS
Install in this order.

1: Win98se
2: Ali 1.82
3: DirectX 7A
4: Detonator 7.58
5: 3dmark2000
6: CTU tweak

128mb ram.
Ram timing fastest in bios.
AGP turbo mode on.
CTU tweak "No write ordering"
You will get 3200+ score.

My P5AB rev is 1.04 With the beta 5 bios.

I think that VIA motherboards are the ones needed for world record.
AOPEN AX59-PRO and EPOX MVP3G2
But also AUSUS P5A/B

At one point I got over 4K on my Aopen AX59 PRO board + Geforce 3 Ti 500. I'll re-run and post the results here.

Ok - my previous statement was wrong. That score was for 3dmark 99. In 3dmark 2000 I get under 2k.... I tried several configurations:

Asus P5A + K6-III @ 550MHz, 256MB ram, Geforce 3 Ti200 - 1824 pts
Aopen AX59PRO + + K6-III @ 550MHz, 256MB ram, Geforce 3 Ti200 - 2081 pts
FIC PA2013 + K6-III @ 550MHz, 256MB ram, Geforce 3 Ti200 - 1877 pts

What am I doing wrong? I did set all ram timings to their lowest settings in BIOS. What exactly is your system configuration OP? (Ram amount, driver version and where to find it, DX version, 3dmark version, bios settings)

I have to admit that it is also a bit unclear for me, how more than 5K 3DMarks 2000 can be reached.
My system is not too far off from the clocks and config and the Quadro2 Pro is running at 295/530 MHz where I get about 3.2K 3DMarks.
Your system seems to perform 60% faster...

elianda wrote:

I have to admit that it is also a bit unclear for me, how more than 5K 3DMarks 2000 can be reached.
My system is not too far off from the clocks and config and the Quadro2 Pro is running at 295/530 MHz where I get about 3.2K 3DMarks.
Your system seems to perform 60% faster...

I said a few posts back:

Highest score for now is p5a-b, 5,5x 110fsb, k6-2+, 128mb pc133, geforce3 ti200, agp 1.82, nvidia 7.76
4190 3dmarks 3dmark 2000 stock.

This is on a k6-2+ and he uses a k6-3+. I used 110fsb he got 115fsb. And maybe with some fine-tuning of the agp driver etc. it is possible to score even more. This is purely the old graphics driver making the biggest difference. On my ''normal'' k6-3+ system with much later driver i also get much lower scores.

@kanecvr Running a clean win98se, ali agp driver 1.72 or 1.82, Nvidia 7.76 graphics driver and directx7.0a you should be able to reach 3500-4000 points at 5.5x 100fsb with a geforce3 ti200. This is ofcourse on an Ali V board and a k6-3+, running 128mb pc133 at fastest timings. In my case an Aopen is little bit slower, but still managed 3800 points.

Edit: i see you use 256mb. The performance hit on a k6-3 is much lower than on a normal k6-2 ofcourse but going over 128mb is still a bit slower. For the highest scores use 128mb.

I have been running tests on my Jetway J-542B ss7 ali motherboard but it is no overclocker: can't get it above 100mhz fsb as there are no settings (and tried a few non documented ones as well).
Beeing lmited to max. 6x100fsb i couldn't score as high as with the other boards.... P5a-b still on top.

Next up is a mvp3 DFI board, but i don't think that will be a big overclocker.

elianda wrote:

Some screens from my system:
*Quadro2 screen*

OT: Thanks for inspiring me to write a nvidia postprocess shader 😜