Hey all, wanted to share a little something that may be helpful.
I have a M912 v1.7 board (possibly the Amptron version, idk) which came with AMI WinBIOS 12/01/1995. I noticed that the Turbo button doesn't work on that BIOS or the 12/02/1995X BIOS that some people use (only the Ctrl+Alt+Plus/Minus keyboard combinations work, and they seem flaky), and I just can't stand having a case with a non-working Turbo button, so I switched over Award with the J.2 patched BIOS.
Oddly, despite many folks reporting success with this BIOS (apart from the cosmetic issue with 5x86 speed detection), I actually experienced stability problems. Notably, I could not complete the extended memory test in Speedsys (crashing out with no message), and 3DBench would freeze a few seconds after showing a score. It looked like a memory problem, but swapping memory didn't seem to help. I also tried slower memory timings in the BIOS to no avail.
I started thinking about why this may have been happening, and realized that the source Award BIOS used for the J.2 patch is the 2.21R BIOS from the v1.4 board. v1.4 and v1.7 BIOSes are separate on The Retro Web, and possibly for good reason. On the AMI ones the POST string ends in GREEN or DGREEN for the v1.4 boards, but on v1.7 most end with GREEN-H or DGREEN-H. In addition, 2.21R uses the Award BIOS string contained "2C4X6B13" but the v1.7 Award BIOSes use "2C4X6H01". This suggests that the BIOSes are actually different to some extent and I probably need to use one specifically for v1.7. Unfortunately, the newest Award BIOS for v1.7 is from 11/03/1994 vs the v1.4 BIOS 2.21R from 09/08/1995, so it's a good bit older, so not exactly a good option and it would need modded to properly support 5x86 and writeback CPUs anyway.
So I decided to open both the v1.7 11/03/94 BIOS and the J.2 BIOS in Modbin and see if there were any settings differences. Turns out there were! There were a lot of differences in the auto tables. I'm not super knowledgeable in this yet, but I understand that the auto tables in particular are related to memory timings. I also noticed in the speed pin settings that the keyboard controller type for the '94 BIOS is KB-200X and the J.2 BIOS it's KB-200.
I went ahead and modified the settings in the J.2 BIOS to match the settings in the 11/03/94 BIOS. Running this, I'm no longer seeing the stability problems I had before, and can even push the memory timings up a bit successfully. I'm not sure if this was just a special "me" issue, but it seemed to help. I'm attaching my modified J.2 BIOS if anyone else with a v1.7 board sees similar issues and wants to try it.
Bonus, no idea if these are any good or not, but some benchmark scores with this BIOS, Am5x86@160mhz (4x40), 256k cache, 32MB 60ns FPM, and 2MB VLB Mach32 DRAM...
(BIOS settings: DRAM Wait State Select: 1WS, DRAM Page Mode: Fast, L2 Cache Read Wait State: 2-1-1-1, L2 Cache Write Wait State: 0WS, L1 Cache Update Scheme: Writeback)
NSSI: 71658 Dhrystones, 22578 Whetstones
3DBench 1.0c: 86.7
Topbench: 327
Speedsys: 59.99, Memory Bandwidth: 84.16 MB/s, L1: 128.36 MB/s, L2: 46.51 MB/s, Memory Throughput 28.89 MB/s
Doom demo3, max details: 2134 gametics in 1494 realtics = 49.99 fps
Quake demo1, max details: 15.4 fps (side note: on the WinBIOS, Quake runs extremely slow, scoring around 6.5-7.5fps depending on the memory settings, but all other benchmarks return "normal" results. I wonder why?)
The OPL Archive - Preserving MS-DOS music in a unified format!