Nice score 😉
With 120 MHz it probably could crack the 11 FPS barrier...
What chipset is on the board?

And I'm supporting your call for benchmark results. Come on, 486 owners, let's find out what combination is fastest!

5u3 wrote:

Nice score :wink: With 120 MHz it probably could crack the 11 FPS barrier... What chipset is on the board? […]
Show full quote

Nice score 😉
With 120 MHz it probably could crack the 11 FPS barrier...
What chipset is on the board?

And I'm supporting your call for benchmark results. Come on, 486 owners, let's find out what combination is fastest!

My plan is to put a beefier CPU on it using my voltage adapter but I think the ancient Ami colour BIOS which I think is the last week/month it was used will cause problems with for example an AMD 5x86

Chipset? uhh....look below 😀

They must have been proud to finally have boards with PCI slots even though you had to configure everything manually in the BIOS

Have an ASUS PCI/VLB board were you set the PCI IRQ's by jumpers, 🤣

Testec the same Matrox card on this motherboard + same CPU as I had on the MSI

http://stason.org/TULARC/pc/motherboards/G/GI … -REV-2-136.html

The board has UM8881F chipset

Just got a score of 7.5 in pcplayer bench

A Diamond Stealth 64 with VRAM gave 7.7 😜
On the MSI system it gave me a poor result of 6.7

interesting stuff with this PCPBENCH:
AMD K6/II 450Mhz with an ALI based mobo with AGP:
Voodoo3/3000 AGP : 22.5
Diamond Stealth 3D 3000pro : 30.4 !!! (VirgeVX based)
wth? is the agp implementation of ALi so crap? I run some 3D games in win98 they are fine... and AFAIR I have tried a pci voodoo3 and it was crap in win98se , I think I tried Unreal in glide mode...

keropi wrote:

interesting stuff with this PCPBENCH: AMD K6/II 450Mhz with an ALI based mobo with AGP: Voodoo3/3000 AGP : 22.5 Diamond Stealth […]
Show full quote

interesting stuff with this PCPBENCH:
AMD K6/II 450Mhz with an ALI based mobo with AGP:
Voodoo3/3000 AGP : 22.5
Diamond Stealth 3D 3000pro : 30.4 !!! (VirgeVX based)
wth? is the agp implementation of ALi so crap? I run some 3D games in win98 they are fine... and AFAIR I have tried a pci voodoo3 and it was crap in win98se , I think I tried Unreal in glide mode...

Ahhh...finally someone who hooks on to this thread and does some testing 😁

Amigo, maybe it's like the early PCI motherbords...the AGP implementation might be crap

Tried another AGP card?

nope, not yet
I have a virge GX2, will try it now 😀

damn, although AGP key is correct, the mobo won't boot with the GX2 card... and now it will not boot with a PCI vga, not even a voodoo3/pci or the same Stealth3000 ... it will only boot with the AGP/voodoo3 now... WTF!?!!!?!?

Check the BIOS for 'Init Display First' setting - it sometimes will mess things up. If it won't work, try to reset the bios settings (clear CMOS).

OK progress

I noticed PCI cards default to normal 640x68x8bit modes, where ALL AGP ones default to 640x68x8bit LFB modes...

So I disabled Linear Frame Buffer with the /nolinear switch.

Voodoo3/3000 AGP : 46 !!!
Virge GX2 AGP : 42

edit: I suspect that SVGA games, like Shadow Warrior, Blood etc that have VESA2.0 modes result to this default LFB modes for AGP cards... How can I disable this for a voodoo3?

Use the NOLFB program?

is the version of NOLFB found here OK?
http://www.advsys.net/ken/build.htm

@Amigaz: The ALI M1429G is indeed one of the older PCI chipsets (and most certainly the PCI bus is bridged to the VLB), so the good performance is somewhat surprising. I'm wondering if the later M1439 and M1487 are even faster.

@keropi: The AGP implementation in AladdinV is a bit wobbly, but that doesn't affect DOS VBE modes that much.
Also, LFB modes should be faster than banked modes, not the other way around. Looks like some chipset features didn't get enabled by the BIOS...

BTW, my scores for the K6 system are a bit outdated since I forgot to enable write combining on the CPU. Here are my new K6-3+ scores:

AMD K6-3+, Asus P5A, Voodoo5 5500

1      MHz | PCPBench score LFB (noLFB)
2-----------+-------------------------
3  66x2 133 |  25.1 (20.8)
4  66x4 266 |  45.6 (35.9)
5  66x5 333 |  53.9 (42.6)
6 100x4 400 |  67.9 (53.5)
7 100x5 500 |  80.7 (63.9)

And more scores coming up: When sorting through some old stuff I found an AMD 5x86! 😁
Of course I had to try it out on my 486 board.

AMD 5x86, Asus PVI-486SP3, S3 Trio64V+

1     MHz | PCPBench score LFB (noLFB)
2---------+-------------------------
3 25x3  75 |  5.7 (5.1)
4 25x4 100 |  6.5 (6.0)
5 33x3 100 |  7.6 (6.9)
6 33x4 133 |  8.7 (8.0)
7 40x3 120 |  9.2 (8.3)
8 40x4 160 | 10.5 (9.6)

Meh. The weak FPU is really holding it back. 160 MHz and no chance for a big score 🤣

BIOS features NOT enabled? enlighten me 5u3!

OK, here are the 2 BIOS screens I think are relevant:

I have changed
Primary Frame Buffer to ALL
Data Merge to ENABLE
and IO recovery time to 0us

now pcpbench is 45.3 for LFB and 46 for normal... any ideas?

Ok, tested pc player on my K6 rig.
Using a Geforce FX5200 AGP

Got lousy 33.2 but with the /nolinear I got a whoppin' 54.4 😁

could this be some K6 "feature" ?

5u3 wrote:

@Amigaz: The ALI M1429G is indeed one of the older PCI chipsets (and most certainly the PCI bus is bridged to the VLB), so the g […]
Show full quote

@Amigaz: The ALI M1429G is indeed one of the older PCI chipsets (and most certainly the PCI bus is bridged to the VLB), so the good performance is somewhat surprising. I'm wondering if the later M1439 and M1487 are even faster.

@keropi: The AGP implementation in AladdinV is a bit wobbly, but that doesn't affect DOS VBE modes that much.
Also, LFB modes should be faster than banked modes, not the other way around. Looks like some chipset features didn't get enabled by the BIOS...

BTW, my scores for the K6 system are a bit outdated since I forgot to enable write combining on the CPU. Here are my new K6-3+ scores:

AMD K6-3+, Asus P5A, Voodoo5 5500

Copy code to clipboard

1      MHz | PCPBench score LFB (noLFB)
2-----------+-------------------------
3  66x2 133 |  25.1 (20.8)
4  66x4 266 |  45.6 (35.9)
5  66x5 333 |  53.9 (42.6)
6 100x4 400 |  67.9 (53.5)
7 100x5 500 |  80.7 (63.9)

And more scores coming up: When sorting through some old stuff I found an AMD 5x86! 😁
Of course I had to try it out on my 486 board.

AMD 5x86, Asus PVI-486SP3, S3 Trio64V+

Copy code to clipboard

1     MHz | PCPBench score LFB (noLFB)
2---------+-------------------------
3 25x3  75 |  5.7 (5.1)
4 25x4 100 |  6.5 (6.0)
5 33x3 100 |  7.6 (6.9)
6 33x4 133 |  8.7 (8.0)
7 40x3 120 |  9.2 (8.3)
8 40x4 160 | 10.5 (9.6)

Meh. The weak FPU is really holding it back. 160 MHz and no chance for a big score 🤣

Actually have a board with that chipset, another MSI board...4143 instead of 4134 which the VLB/PCI model is
Sadly the board seem to be dead..haven't completely given up on it though..still some more CPU models + memory combinations to try etc

free image host

Seems that this is the last model the used the ALi chipset on...Swaaye has the model that came after and it's equipped with the SiS 496/497 chipset

keropi wrote:

I have changed Primary Frame Buffer to ALL Data Merge to ENABLE and IO recovery time to 0us […]
Show full quote

I have changed
Primary Frame Buffer to ALL
Data Merge to ENABLE
and IO recovery time to 0us

now pcpbench is 45.3 for LFB and 46 for normal... any ideas

Yeah, that's definitely better than the 22.5 FPS you posted earlier.

Now try one of these K6 write-combining/MTRR utilities and check if it makes a difference. I've attached a DOS utility called MXK6OPT.

will try it 5u3 ! Thanks!
I wonder why there is no BIOS setting for that thing!