no bus divider is not fully stable (yet at least).
so far the best i got is 3x66, 1:2/3, bios on max.

How are you setting a 1:2/3 bus ratio on the LSD? I thought JP19 sets either 1:1 or 1:1/2. This assumption is based on an old thread, I’ve never measured it for certain.

Ack.
My bad.
Everywhere i wrote 2/3 should be 1/2.
I guess it wrote that in my first respond to Feipoa and then just kept repeating myself. 😖

Dang, was really hoping you were aware of a jumper combo for it.

Don't forget that the LSD doesn't work with the on-board floppy controller at 66 MHz. Perhaps this can be circumvented by using an I/O card or LS-120. One of the issues i've run into with LS-120 drives is that once the DOS driver is loaded, the LS-120 drive doesn't always work correctly once booted into Win9x.

All the SiS 496 boards can only do 1:1 or 1:1/2 to PCI. Later UMC 8881 boards have the 1:1, 1:1/2, and 1:2/3. I'm going off of a 10-year old memory, but I recall that the PCI bus divisor comes from the UMC chipset while SiS 496 requires the 1:1/2 to come fro the PLL clock generator chip on the motherboard.

Yep.
Over the last 2 days i spent time testing 200MHz Am5x86 chips on VLI UUD and other boards.
I guess i was at UUD while responding here and the 1:2/3 divider was fresh in my head.

feipoa wrote on 2022-05-09, 03:23:

Don't forget that the LSD doesn't work with the on-board floppy controller at 66 MHz. Perhaps this can be circumvented by using an I/O card or LS-120. One of the issues i've run into with LS-120 drives is that once the DOS driver is loaded, the LS-120 drive doesn't always work correctly once booted into Win9x.

All the SiS 496 boards can only do 1:1 or 1:1/2 to PCI. Later UMC 8881 boards have the 1:1, 1:1/2, and 1:2/3. I'm going off of a 10-year old memory, but I recall that the PCI bus divisor comes from the UMC chipset while SiS 496 requires the 1:1/2 to come fro the PLL clock generator chip on the motherboard.

lucky star used a couple different super IO chips - this may not be true of every single board

Well, here is my first try 😀 . I got 3 x 64 = 192 MHz. 3.9V stable with perltier.
Unfortunately I am a bit late to the party, and a bit unlucky with my CPUs.
My M/B is also an older revision and does not support EDO ram.
Got some relative good results (I think).

Video:
https://youtu.be/PZ82WlMuG7k

Benchmark results:
https://docs.google.com/spreadsheets/d/1hUfLI … w6Dlv7xmWyDDUFT

PS: I have a month now a custom PCB for 1MB L2 mod. Waiting for chips.
The first packet got returned 🙁

Fantastic results! Where can I learn more about the custom clock generator? Looking forward to updates!

Chadti99 wrote on 2022-05-19, 17:43:

Fantastic results! Where can I learn more about the custom clock generator? Looking forward to updates!

Thank you, Here... 😜
https://youtu.be/odcH80wiiPs

CPUGalaxy's 486 Quake record has been beaten for some time now by users pshipkov and Chadti99. I think they are at 22.6 - 22.8 fps.

The custom any clock device is interesting. I'd personally like to see a plug-in replacement for DIP-14 PLL's found on many of the socket 3 PCI systems, e.g. UM9515-01. My interests would be to fine tune a Cyrix 5x86-133/4x chip. We know 150 Mhz is too much for it, so some tuning in the 135-145 MHz range would be fun. I recall running into fully programmable PLL's in the past, but never used them. Swapping the MB's crystal that feeds the PLL doesn't work well in this case because it renders other systems inoperable, like the floppy diskette.

Binning / Testing 14 Am5x86 CPUs. The 190% Overclock on Am486-DX2-66 and the Unreal 70MHz 486 Bus!
Enjoy...
https://youtu.be/dwYBq7K7h-Y