https://datasheet.octopart.com/FW82371EBSL37M … heet-117403.pdf

Pg28

Tldr: it's fixed to 1/4pci

Regarding VIA 686 southbridge:

https://theretroweb.com/chip/documentation/vt … b8269986481.pdf

Page 56 (62 of PDF) states that the default ISA clock speed is PCICLK/4 - I've never seen any BIOS that will allow you to change it.

Even then, I don't recommend the VIA 686 southbridge (and the VIA 586 southbridge to a lesser extent) for ISA benchmark testing anyway - not only is it worse overall compared to Intel chipsets, but it also cripples the performance of video cards that can normally hit the top speed of the ISA bus. I've been collecting benchmark data for the past week across various motherboard+CPU setups specifically investigating this:

fdoom13h benchmark was done with FastDoom 1.0.7. For some reason, Quake refused to run on the VIA Apollo+586 motherboard. Benchmarks for fdoom13h and Quake are not included from the 486 benchmarks due to CPU bottleneck.

Unfortunately, my reliance on my Apollo 133T+686B benchmarks made me believe that the Oak OTI087 and OTI087X are top-speed ISA cards for DOS which I've stated in the past, when further evidence now shows that it is not the case.

tl;dr - don't trust the ISA benchmarks listed at http://vgamuseum.info/index.php/benchmarks as those were done with a VIA 686B southbridge.

I have noticed the much lower ISA performance on my VIA boards with 686B, and I made a little util that overclocks the bus and it gained some of the performance back. It should make ISA speed PCI/3 from default PCI/4. There's a PCI/2 option in the chipset but I think I experienced issues with it, which is why I didn't go for it in the util.

I took the liberty---

https://www.vogonsdrivers.com/getfile.php?fileid=2240

Let me know if you want it here: https://theretroweb.com/chips/2733#driver

Sure, no problem with it being on TRW too ~

Thanks Tiido. My aim is to overclock isa card on win98se, I edited my first post accordingly.
Will your utility work on win98se?

It is for running from pure DOS before Windows starts, something you put in AUTOEXEC.BAT. It might do stuff while Windows has already started but I have never tested it that way and don't expect it to work then.

Tiido wrote on 2025-01-20, 16:48:

I have noticed the much lower ISA performance on my VIA boards with 686B, and I made a little util that overclocks the bus and it gained some of the performance back. It should make ISA speed PCI/3 from default PCI/4. There's a PCI/2 option in the chipset but I think I experienced issues with it, which is why I didn't go for it in the util.

Hi,
Unfortunately it does not work with my VT82C686B PCI to ISA Bridge. It says VT82C686 PCI to ISA Bridge not found. The bridge's device id is: 1106-0686.

Interesting, vendor 1106 + device 0686 is exactly what it seeks for, but only at bus 0, device 7, function 0, which is where it resides on Chaintech 6VIA5T mobo. Perhaps it lies at different locations in different motherboards, in which case the detection mechanism needs to be made more flexible.

Tiido wrote on 2025-01-21, 17:14:

Interesting, vendor 1106 + device 0686 is exactly what it seeks for, but only at bus 0, device 7, function 0, which is where it resides on Chaintech 6VIA5T mobo. Perhaps it lies at different locations in different motherboards, in which case the detection mechanism needs to be made more flexible.

It's exactly on the expected place (bus 0, device 7, function 0) yet the utility cannot find it:

Interesting, and you run it from pure DOS ?

Tiido wrote on 2025-01-21, 22:37:

Interesting, and you run it from pure DOS ?

Yes, of course.
BTW, I tried it in real mode without EMM386/JEMM386 started.

Very interesting, then I'm unsure what is going on, especially since it works perfectly on the Chaintech board I made it for in the first place... There shouldn't be any kind of board specific things getting in the way and I'm only doing standard PCI conf mechanism IO ports accesses which should work the same way regardless of motherboard or chipset. The error does suggest that something goes wrong and the program is unable to read back the device and vendor IDs from the PCI conf mechanism.

I do have couple other 686 bearing motherboards but they are packed away and I don't really have a good testing setup for the time being, perhaps once me and the GF move to an actual house where there's actual place for my stuff 🤣.

Thank you for testing, I will try to investigate once I am done with my YMF71x card stuff.

Falcosoft wrote on 2025-01-22, 00:05:

Tiido wrote on 2025-01-21, 22:37:

Interesting, and you run it from pure DOS ?

Yes, of course.
BTW, I tried it in real mode without EMM386/JEMM386 started.

@Edit:
Test video:
https://youtu.be/4iyJRjSaF9M
You can see that device is properly listed by PCISleep yet Viaisa cannot find it.

There's a clue, the FFFF FFFF are vendor and device figures it found. It seems that for some reason PCI conf mechanism isn't functioning, not the way the program assumes it would... It is using IO port based method instead of any BIOS calls etc.
After I'm done with my current project, I'll investigate things proper.