Hello,

Before I continue with my story, let me just give you the goods up front. Oh, and I've only tested this BIOS on a V3.4 / REV:3.4 board. I haven't tested this on any other revision of the M729. I have also not tested this with 100MHz FSB, only with 66MHz.
Use this hacked BIOS at your own risk, I cannot be held responsible for anything that may or may not happen as a result of you (mis)using this edited BIOS. Don't even *think* of using this on anything that has a critical role (such as a hospital system).

Okay, now that that's out of the way...

Last Sunday I parted with $10 to wind up with an old Celeron 300 (Covington, natch) based box with this motherboard in it. HDD was shot and the cheap PSU had a swollen cap, but other than that the system worked. (I replaced the power supply with one that didn't threaten to burn my house down.)

Anyway, after swapping the HDD and getting a Mendocino Celeron 333MHz, I was dismayed with the numbers I was getting:

I figured that this board was woefully underperforming so I went into the AMI BIOS setup... only to find a very paltry selection of options available, some of which didn't even work (Delay Transaction)! Unable to get much of an improvement out of what was available to me, I searched for an AMI BIOS editor - and found an old DOS version, AMIBCP76, which fortunately was good enough to allow me to edit the BIOS. What met my eyes was a rather large number of options that were hidden and disabled by PC CHIPS, presumably to fake a better return rate by disabling what they saw were the most problematic features.

Anyway, I re-enabled almost all of the hidden options (what remained were relatively non-pertinent to performance and wouldn't otherwise be worth enabling) and flashing the modded BIOS to the board, I discovered that PC CHIPS had forced the IO Queue Depth (IOQ Depth in BIOS) to *1*. This would explain the board's miserable memory throughput. 😜 I set it up to 4, and noticed a large improvement in memory throughput, from 109MB/sec to 152MB/sec under SPEEDSYS. The option allows 1, 2, 4 or 8 depth, but the 8 depth didn't seem to affect anything so I just left it at 4 in my case.

I also enabled the options for Passive Release, PCI-33 Deferred Mode and PCI-66 Deferred Mode, and their associatedposted write options. All three of these sets of options work ed to some extent on my board, but depending on the VGA card you're using you may run into issues with the PCI-66 Deferred Mode. The beforementioned Matrox G450 really liked them, but the SiS 6326 AGP that came with the computer was much less enthusiastic, and gave me a lovely garbled DOS screen. Also, Windows 98SE may not like the PCI-33 or PCI-66 Deferred Modes either - it sure didn't like them on my board, but maybe you'll be luckier?

I enabled the PCI IDE BusMaster option, but unfortunately it doesn't work correctly on my board. (Hence why I am using a Promise Ultra/66 card in this system - it outperforms the onboard IDE in a dramatic fashion anyway so no big deal for me.) Oh, there's also an option to disable the onboard sound chip through BIOS. I didn't test this though.

Anyway, after a hearty amount of fuddling about with the options that I've unlocked (and replacing the SiS 6326 with the Matrox G450 card), here's where I am at now with this system:

Not a bad improvement overall if I may say so myself - DOOM realtics went from around 1600-ish (on a 333MHz Mendocino, no less) to 898 after all was said and done. (I have my DOOM benchmarking setup calibrated to COMPLANG's list: https://www.complang.tuwien.ac.at/misc/doombench.html)

Thanks for putting up with my sordid adventures and I hope this helps out anyone else stuck with a PC Chips M729 V3.5,

Eep386