Hi Nexxen,

Thanks for your detailed benchmarking of the M912 J.2 BIOS against the AMI versions!
I have a comparable UMC498F board with VLB IDE now (Aquarius/BCOM MD-4DUVC Ver 2.11) to help with further testing and compare scores with. 😊

However at the moment I’m busy finishing the 25th anniversary update of my CHKCPU CPU Identification utility and I expect to continue with the M912 Award BIOS project in a week or two.

Jan

@Chkcpu, any news? I've just got this board with 5x86 P75 and now see that it will require your patch 😀

There was another board from that seller of v1.7 with not so fancy black ISA slots that detected that CPU as DX4-486-100S (or something like that) but I preferred this one that doesn't even boot with 5x86 (though I see post activity). I'll try some other BIOS versions for sure but seems like your solution is going to be the best one.

Edit. Interesting - I have exactly the same mobo as BastlerMike did (the first comment author). It has BIOS ver 12/02/1994 from retroweb. I updated it to the BastlerMike's BIOS (ver 12/01/1995 from retroweb) and it works. And detects my CPU as 5x86 with 133MHz frequency. Jumpers were set to the Cyrix settings from the first try. Will check the cache settings later.

Figured I would chime in on this as well.

I have v1.7 of this board with a 120 mhz AMD (Am486DX4-100SV8B). The BIOS has an option for write back and write thru (though I later noticed it says 'L1'). This is definitely supported by the CPU.

The L2 cache has been upgraded to 256 K with appropriate sized tag, and that all works in write through mode.

However the system won't seemingly work with write back enabled. I've tried 3 different jumper settings for CPU type (as per documentations) and turning on WriteBack in BIOS, but they all result in either the system straight up freezing at boot, or cachechk will report extremely high results for the megabyte #1 and claim there's no cache installed.

I've tried tweaking every BIOS setting and also 'auto' mode to no avail.

So at this point I can only think I need an updated BIOS, unfortunately I don't have an eeprom tool.

I'm open to other suggestions though 😀

Ehm.... To get an eeprom tool? 😀 It's useful even for arduino flashing not to mention any bios related stuff. And a bunch of W27C512 chips, they are pretty often used on cards of this era and can be erased without UV-lamp.

rkurbatov wrote on 2023-03-30, 21:43:

Ehm.... To get an eeprom tool? 😀 It's useful even for arduino flashing not to mention any bios related stuff. And a bunch of W27C512 chips, they are pretty often used on cards of this era and can be erased without UV-lamp.

Well yeah that's a possible solution, I don't know if an updated BIOS would even fix the issue, hence my request for other ideas 😁

rkurbatov wrote on 2023-03-28, 19:56:

@Chkcpu, any news? I've just got this board with 5x86 P75 and now see that it will require your patch 😀

There was another board from that seller of v1.7 with not so fancy black ISA slots that detected that CPU as DX4-486-100S (or something like that) but I preferred this one that doesn't even boot with 5x86 (though I see post activity). I'll try some other BIOS versions for sure but seems like your solution is going to be the best one.

Edit. Interesting - I have exactly the same mobo as BastlerMike did (the first comment author). It has BIOS ver 12/02/1994 from retroweb. I updated it to the BastlerMike's BIOS (ver 12/01/1995 from retroweb) and it works. And detects my CPU as 5x86 with 133MHz frequency. Jumpers were set to the Cyrix settings from the first try. Will check the cache settings later.

Hi rkurbatov,

Due to lack of time and new ideas, my M912 Award BIOS project has been dormant these past months. But a week ago I pulled my UMC498F board (Aquarius/BCOM MD-4DUVC Ver 2.11) out of storage, installed an Am586-P75 and figured out how to set the jumpers for x4 multiplier and L1 cache WB mode.

This worked fine with my M912 patch J.2 BIOS, however the incorrect CPU speed indication by the BIOS (100MHz instead of 133MHz) was present here as well. So this is definitely a BIOS bug and not a problem of the M912 board. 😉
I’m still investigating this bug and I hope to have a fix soon.

Okay, you got it working with the AMI 1995 BIOS! But you may have to change some jumpers to get it working correctly in L1 cache WB mode.

@drosse1meyer, reading about your L1 cache WB problem with the Am486DX4-120 SV8B, I’m convinced a BIOS update will fix this. When you get hold of an EEPROM programmer, the AMI 1995 or 1995X, or my Award M912_J2 BIOS will fix the issue.

Cheers, Jan

Chkcpu wrote on 2023-04-02, 14:41:

rkurbatov wrote on 2023-03-28, 19:56:

@Chkcpu, any news? I've just got this board with 5x86 P75 and now see that it will require your patch 😀

There was another board from that seller of v1.7 with not so fancy black ISA slots that detected that CPU as DX4-486-100S (or something like that) but I preferred this one that doesn't even boot with 5x86 (though I see post activity). I'll try some other BIOS versions for sure but seems like your solution is going to be the best one.

Edit. Interesting - I have exactly the same mobo as BastlerMike did (the first comment author). It has BIOS ver 12/02/1994 from retroweb. I updated it to the BastlerMike's BIOS (ver 12/01/1995 from retroweb) and it works. And detects my CPU as 5x86 with 133MHz frequency. Jumpers were set to the Cyrix settings from the first try. Will check the cache settings later.

@drosse1meyer, reading about your L1 cache WB problem with the Am486DX4-120 SV8B, I’m convinced a BIOS update will fix this. When you get hold of an EEPROM programmer, the AMI 1995 or 1995X, or my Award M912_J2 BIOS will fix the issue.

Cheers, Jan

thank you

@Chkcpu thank you!

One more question - are there some better explanation of what do the processor jumpers really do? That's so messy, there are two of multiplier sets (one for AMD, one for Intel), one unknown jumper and just a set of somewhat random settings like 1-2, 3-4 or 2-3 in the main jumper block. The only clear part is voltage (though why 3 jumper blocks for it?)

rkurbatov wrote on 2023-04-02, 22:28:

One more question - are there some better explanation of what do the processor jumpers really do? That's so messy, there are two of multiplier sets (one for AMD, one for Intel), one unknown jumper and just a set of somewhat random settings like 1-2, 3-4 or 2-3 in the main jumper block. The only clear part is voltage (though why 3 jumper blocks for it?)

Ahh, jumper hell. 😊
The reason behind all those jumpers on 486 boards is best explained by taking a look at this website:
http://ps-2.kev009.com/eprmhtml/eprmx/h12203.htm

Socket 3 boards can take a lot of different 486 CPUs. The above resource shows the wild variety of pin-outs on these CPUs that needs to connect to the correct chipset pins. That’s the job of all those jumpers.
In addition, many 486 chipsets need trapping jumpers to match its functionality with the installed CPU model.

The 3 Vcore voltage jumpers actually carry the current for the CPU and are just connected in parallel to divide the load. Using only one jumper there would be unreliable.

Jan

Thank you, Jan!

Apologies for the necroposting.

I have PCChips M912 ver.1.4 mobo with real cache chips, but the BIOS version doesn't support Am5x86-P75 which I want to use with the board. And of course the factory 1 Mbit BIOS chip is OTP EPROM, so I can't flash it with other BIOS versions (combined AMI + Award BIOSes into one chip).

I'm pretty sure the SST27SF010 chip will be a suitable replacement for the 1 Mbit factory OTP EPROM, but it's not easy to find and not so cheap. Please suggest me some suitable BIOS chip substitutes for this board. What about SST29EE010 or Winbond W29EE011 EEPROMs?

I would also like to know which are the latest versions of AMI and Award BIOS dumps from TRW site that would be suitable for Am5x86 CPUs.

Thank you!

I bought Winbond W27C512-45, and they work well.

0.30€ each, now they are around 0.90. 🙁

Prolly 95X is latest P-75 friendly. You'll just test em to chrck, no problem at all.

Thank you for the info, Nexxen, but I think this Winbond W27C512 is 512 kbit (64 kB) in size. I want to keep the dual AMI-Award BIOS option, so I need 1 Mbit (128 kB) EEPROM chip. Maybe Winbond W27C010 will be compatible too.

Ok, for AMI BIOS I'll go with 40-P301-001437-00101111-072594-DGREEN-H - 12/02/1995X, but what about the Award one?

analog_programmer wrote on 2024-07-10, 16:36:

Thank you for the info, Nexxen, but I think this Winbond W27C512 is 512 kbit (64 kB) in size. I want to keep the dual AMI-Award BIOS option, so I need 1 Mbit (128 kB) EEPROM chip. Maybe Winbond W27C010 will be compatible too.

Ok, for AMI BIOS I'll go with 40-P301-001437-00101111-072594-DGREEN-H - 12/02/1995X, but what about the Award one?

Yep, my mistake. I didn't read that part. I never tried dual BIOS.
I have 1.7, I don't know if it's the same but it was both the latest available that worked with P-75.

Scrolling back into this thread I saw that 09/08/1995-UMC-498GP-2C4X6B13-00 - 2.21R Award BIOS is the latest for my ver. 1.4 board, but this BIOS dump from TRW site is with modified/broken L2 cache size report routine and it lacks any Am5x86 CPU support. Our BIOS-mods expert Jan Steunebrink managed to fix this in his 2.21R-J1 modified version, but this version also lacks Am5x86 CPU support. And later he also added 5x86 support in 2.21R-J2 version which works fine, except for the wrong frequency report (maybe this bug is related to x2 multi for 486DX2-50/66 CPUs).

So there is no Award BIOS dump on TRW site, which is suitable for Am5x86 CPUs, thus I'll just use Jan's 2.21R-J2 modified version.

Finally I got a couple of Winbond W29EE011-15 EEPROM chips to change the original OTP-EPROMs on this and some other old boards, so I can update it with the BIOS version that supports Am5x86 CPUs.

Now the Award BIOS version 2.21R-J2 (the modified one from Jan Steunebrink posted in this thread) works flawlessly with my M912 1.4 mobo, but the latest AMI BIOS (12/02/1995X) downloaded from TRW site does not.

I made a combined BIOS file from these two BIOSes (first Award 2.21R-J2, then AMI 12/02/1995X), but on first test the motherboard POSTed only with Award BIOS activated by jumper J2. AMI BIOS stops at POST codes "B7, B6". Before I switch the active BIOS by jumper J2 I always clear the CMOS with J1 shorted on positions 3-4. Then I made another combined BIOS file - first AMI 12/02/1995X, then Award 2.21R-J2 and still the board POSTs only when Award BIOS is activated.

So maybe AMI BIOS version 12/02/1995X is not compatible with my revision 1.4 of the board or the BIOS dump I've downloaded from TRW is corrupted. I'm just guessing as I don't know where the problem with latest AMI BIOS comes from. Is there another source for AMI BIOS 12/02/1995X dump-file to be downloaded from?

Also I will try AMI BIOS Release 07/02/1995 version, but does it support 5x86 CPUs?

analog_programmer wrote on 2024-08-17, 12:31:

Is there another source for AMI BIOS 12/02/1995X dump-file to be downloaded from?

Try this:

Thank you, Nexxen!

I tried your dump copied two times in one file just to be sure that there's no matter which address space will be used from the 1 Mbit EEPROM chip. And still the POST process is stuck, this time at codes "D3, D1".

I think there are two variants now: The AMI (win)BIOS is somehow "incompatible" with this Winbond W29EE011 EEPROM chip, or this AMI BIOS version is not compatible with my revision of the board. (I've checked this - see next post)

I'm using i486DX2-66 CPU for the tests, so I'll flash some of the older AMI BIOS versions starting with Release 07/02/1995 to see if they'll work.

P.S. Well, I always read old longer threads backwards and maybe this is a kind of mistake. Now I see that on the first page of this thread there are plenty of AMI (win)BIOS dumps attached. And also this part of an old post tells me enough:

cyclone3d wrote on 2021-08-16, 18:06:

I have multiple M912 boards and got frustrated with them as certain BIOS versions will work on certain boards but not on others.

All of mine have real cache.

So, I have to get all of the available AMI (win)BIOS dump-files here and on TRW, compare and sort them out and then test them one by one from newest to oldest, 'til I find some working AMI BIOS that's newer than my original one 😳

P.S.2: I think during '90s AMI lost their positions for the Award/Phoenix in (s)s.7 times up to UEFI BIOSes era, because of all these windows-like "fancy" (but actually c*appier) BIOSes for late 486 boards. I still prefer AMI BIOSes for s.370 boards over Award's ones, but this changes nothing, especially for the long past times 😀 Just a thought.

It turned out that that the AMI (win)BIOSes somehow really are"incompatible" with this Winbond W29EE011 EEPROM chip. Once I've seen similar problem in some Necroware's YT-video.

How I found this? I just copied two times my original working version of AMI BIOS (the one I dumped from the factory OTP-EPROM) and flashed the resulting 128 kB file just to find out that the POSTing process is broken with this "combined" AMI BIOS x 2. After this I did the same thing with Award BIOS and everything works fine even without clearing CMOS when selecting low or high EEPROM address space by jumper J2 (just as I expected).

So, I have to search for another 1 Mbit EEPROM chip that will be compatible with those AMI (win)BIOSes. And I don't have many ideas except for Winbond W27C010 or compatibles.

P.S. I still hate these AMI windows-like BIOSes, even more than before 😁