Did you try posting with only a video card? And did you try various graphics cards in different PCI slots?

Did you confirm, by way of other motherboards, that the PCI Post Write Buffer feature has any measurable impact on benchmarks? If so, how much impact?

Maybe the SMD capacitor was intentionally not soldered onto that spot? I see several locations with empty SMD solder pads.

What is the molex power connector for near the CPU? Is it necessary to connect the PSU to this connector?

Perhaps try an AWARD-based BIOS instead of AMI? The layout and components of this board are similar to the MSI MS-4144 and the Chaintech 486SPM.

If you use an AWARD-based BIOS, you should be able to modify the board for PS/2 mouse support. Also, you have 1024K cache as an option. I really like working on these brand named PCI 486 boards such as this.

There is a user manual for this particular board right here on Vogons. I'd have a look at it.

http://www.vogonsdrivers.com/getfile.php?fileid=662

[EDIT]

The onboard molex connector is a power source for chassis fans.

Hi,

I tried three different cards, in the three different pci slots. They give either a blank screen or a garbled screen after POST.
I don't know the impact of the setting, and I don't have another working board to test it.
It's possible that the cap wasn't there in the first place, but I doubt it: there were noticeable 'blobs' of solder left, unlike the other unpopulated pads. Also, the board in the picture isn't mine, and it does have a capacitor there. And last but not least, after soldering a capacitor there the board went from not working to working 😉

That's cool that I could mod the board for PS/2 mouse support, I didn't know that! Do you have a link on how to do that maybe?
Currently I have 512kb of cache ram, I looked it up and it doesn't seem like there would be a benefit in adding another 512K, unless I wanted to install a lot of RAM? I have 16MB of RAM in there now, that's already enough for a 486-DX2/66 running dos 6.22 + WfW 3.11 I think. And I don't have bigger ram sticks anyway 😀

I looked at the user guide, it has this to say about the CPU-to-PCI Memory Post Write Buffer:
Enable this parameter to control the posting of the CPU-to-memory write data in the posting buffers. Disable the parameter to deactivate the buffering function.

edit: I also ran memtest86 for a while and it gave no errors.

Its a bit of work, but you can start by looking through this post,
Re: DTK PKM-0033S +5V pin overheating + PS/2 mouse implementation
Re: Native PS/2 mouse implementation for 386/486 boards using the keyboard controller

Not much measurable benefit of 1024K vs. 512K unless you want to cache 128 MB in WB mode, or 256 MB in WT mode. I do not much care for 512 KB single-banked cache because it does not tolerate as fast of cache timings compared to double-banked.

Not too long ago I compared cache sizes using CPUMark99. I will copy/paste a section of that thread here,

1024K - Write-back (Write-through)
16 MB = 5.67 (5.47)
32 MB = 5.82 (5.61)
64 MB = 5.79 (5.62)
112 MB = 5.83 (5.61)

32 MB (write-back)
1024K = 5.83
512K = 5.52
256K = 5.33

FYI, cacheable limits
1024K (write-back) = 128 MB
1024K (write-thru) = 256 MB
512K (write-back) = 64 MB
512K (write-thru) = 128 MB
256K (write-back) = 32 MB
256K (write-thru) = 64 MB

Interesting how CPUMark99 shows a 10% improvement when the cache jumps from 256K to 1024K. I think the real-world improvement is around 2%. Shows how sensitive this benchmark is in regard to cache. Similarly, the benefit of write-back over write-through cache was 4%.

Thanks for the information 😀 that mod seems doable, but a bit too involved for me atm.. This is my only working 486 board and I'm going to play it safe. I do want to do some tests on it, to compare the performance and see if it remains stable under stress. What programs can I use? Does this CPUMark99 work in DOS and where can I get it?

CPUMark99 is only for Windows.

There are CPU stability testing programs, burn-in programs, etc, which run the CPU at 100% for as long as you want. Personally, I like to let Quake autoplay for at least an hour and run MemTest for 24-hrs.