Manual for the ATC-1442G:
http://www.elhvb.com/mboards/a-trend/1442g/index.html

Manual for the UM4981 AIO (it's for the rev 2.1 but probably still generally relevant)
http://www.elhvb.com/mobokive/Archive/Ecs%20- … M4981_Rev_2.PDF
Or TH99 of course:
https://stason.org/TULARC/pc/motherboards/E/E … UM4981-AIO.html
Oddly enough the manual for rev 2.1 refers to an optional VRM, but this rev 1 board appears to already have VRM onboard...

Not sure which one I'd prefer - both are very much B-label brands, neither board is perfect. I'd agree with the layout comment about the ISA slots - but I'm very suspicious of the huge distance between chipset/cache controller and the L2 cache on the ATC board. Maybe it works fine, but I'd want to be sure of stability before using that one. On the other hand the ECS board is older and doesn't refer to 5x86 (Am or Cx) anywhere. Am5x86 generally works with Am486DX4-WB settings, but Cx5x86 tends not to work at all if not explicitly supported. Which CPU are you intending to use here? The Cx5x86, Am5x86 or that PODP-83 (which will probably run at 33MHz due to missing fan).

VLB with integrated IDE, Floppy and COM and 4 SIMM72.

dionb wrote:

Manual for the ATC-1442G:
http://www.elhvb.com/mboards/a-trend/1442g/index.html

That helped. Thanks!

dionb wrote:

Am5x86 generally works with Am486DX4-WB settings, but Cx5x86 tends not to work at all if not explicitly supported. Which CPU are you intending to use here? The Cx5x86, Am5x86 or that PODP-83 (which will probably run at 33MHz due to missing fan).

I plan to use pretty much all these CPUs and move between them to see what they can do. Obviously all of them, incl. the PODP with the fan 😀

Neither of those boards seems to be supporting the Cyrix 5x86 out of the box. But I've been lucky with ATC-1442. I ignored (or couldn't find) the manuals. So I just wired HITM, INV and WB CPU signals to the chipset via jumpers using multimeter. I couldn't be bothered with SMI signals. The UMC board reported 486DLC-100. The Opti board reported something DX4-100, but worked. I confirmed the frequency was correct 100 MHz and even write-back was supported for both L1 and L2. I was able to even overclock it on the Opti to 120 MHz via FSB increase and a bit of L2 latency increase in BIOS.

With the Am 5x86 the UM4981 worked fine with L1 in WT/WB. However, on the ATC-1442G it doesn't POST in WB mode at 133 MHz. At 120 MHz, it is fine, even in WB model

Also the UMC seems to be caching only first 16MB if I put two 16M sticks in. Not sure if TAG chip upgrade would help..

I will keep playing with them and I guess I am going to keep the ATC-1442G... The onboard IO on the UM4981 will be missed though...

Which 486 VL-Bus motherboard or chipset in general do you prefer BTW?

Anything with good documentation for starters 😉

I really like FIC motherboards for build quality and documentation. Very sad my 486-PVT is dead. That would fit your requirements perfectly. It has a Via chipset.

That aside, I like my Biostar MB-1433UIV with UMC8498F chipset, although its jumpers are baaaad - but that's no different on most 486 boards. Only a few Asus ones seemed to have been made with easy config in mind.

mpe wrote on 2019-10-20, 21:03:

With the Am 5x86 the UM4981 worked fine with L1 in WT/WB. However, on the ATC-1442G it doesn't POST in WB mode at 133 MHz. At 120 MHz, it is fine, even in WB model

Hello, I wondered if you got the UM4981 working with a 5x86-133 in internal cache WB mode, including running himem.sys and DOS=HIGH or if you just tried POST.
With mine, it locks up after running config.sys and before giving the A:\> prompt, but if I switch to WT mode it works fine.
If I don't load himem.sys, or specify dos=low in config.sys, it works fine in WB mode.
If I swap the Phoenix BIOS chip for an AMIBIOS one from a PC Chips M912 (same chipset), it works fine in WB mode.
If I use a custom copy of Himemx.exe (from FreeDOS), that I assembed with an extra WBINVD instruction added before it toggles the A20 line, it also works fine in WB mode with the original Phoenix BIOS.
So it's kind of a bizarre problem and I wonder if anyone's run into it on another UM4981 v1.0. It's not a hardware problem since the M912 BIOS makes everything work right. I suspect the Phoenix BIOS is missing some of the steps in programming the UMC chipset for a WB CPU since, aside from this broken A20 line, it seems to work fine?
The board has markings on it for a P24D (486DX2-WB) so internal cache WB should be fully supported.

I tried with both the original BIOS chip as well as the 4981.22D BIOS floating around online. The 4981.22D bios actually has 5x86 support in it unlike the original, but no difference in it working. I've also tried jumpering it to be a 486DX4-100.

I got working AMD 5x86 in writeback mode on UM4981 AIO. Their BIOS is old
Only DX4-100 Mhz on screen, whatever i set (JP29) 3x or 4x multipler.
I place 5x86 and config it correct (config same as P24D ) , it runs with 100 Mhz
Already, i set Jumper JP29 to 2-3 , it run on 133Mhz. 😀, despite "DX4-100" on screen.

Here a bios image of my board. It doesn't know 5x86, it runs (not so) good in WB-mode. (Win98SE crashed, if 5x86 is jumpered as P24D , it run stable with P24C stetting)

Memory throughput is not good. 36MB/s by 40MHz FSB.
But, i have crack 60 FPS barrie on DOOM bench , with Trio64V+ VLB (homemade )

Madao wrote on 2020-11-08, 09:15:

Here a bios image of my board. It doesn't know 5x86, it runs (not so) good in WB-mode. (Win98SE crashed, if 5x86 is jumpered as P24D , it run stable with P24C stetting)

Here's the hacked version. I wonder if it will help yours work in WB mode as it did mine. I also had stability problems, but for me, they showed up as soon as I attempted to boot DOS=HIGH. The problem appears to be that the BIOS does not initialize the chipset correctly for a write-back CPU. Good luck.

It help ! Thanks

5x86 WB run stable , but a bit less performance than old BIOS in WB Mode. Win98SE stable, DOSbench stable, no freeze.
A interessing notice: CPU type DX4 and 5x86 is dected as DX2-100

Greetings
matt

Madao wrote on 2020-11-09, 15:46:

It help ! Thanks […]
Show full quote

It help ! Thanks

5x86 WB run stable , but a bit less performance than old BIOS in WB Mode. Win98SE stable, DOSbench stable, no freeze.
A interessing notice: CPU type DX4 and 5x86 is dected as DX2-100

Greetings
matt

If you haven't already you may want to go into the bios setup and disable autoconfig and set your cache and memory timings manually. Part of the hacking of the bios was that is missing CPU detection for a 5x86. If a 5x86 is present, it detects it as whatever CPU was in the system prior. This causes issues where, if you reset the CMOS memory, the 5x86 can be mis-detected and it may fail to boot unless you put a DX4 in first and then swap it back. The hacked BIOS is setup to detect a 486DX4-WB and a 5x86 as a DX2WB. It may think your bus speed is 50 MHz in autoconfig mode and therefore de-tune the system.

Careful with stability if you try and bump the cache faster than 3-2-2-2 however, especially on my system where I'm running at 40 MHz. I suspect the UM8672 VLB IDE may play a role in that. A good test for stability is the Doom shareware timedemo, or better yet, boot into an old Linux (Red Hat 6.2, Slackware 8, etc.) and compile a kernel.

Also: have you found this board won't take a SIMM bigger than 16 MB?

Cache Timing 2-1-1-1
SRAM hit 0
DRAM waitstate 1 (0 = crashed instant, need faster PS/2 memory )
DRAM Page mode fast

256kb SRAM has 15ns timing.
It is stable with 5x86 at 160 Mhz
But Performance is a bit lower ( DOOM 59,99 FPS versus 60.33 FPS )
I am stastified.

I haven't check with bigger SIMM (i have few 64MB FPM SIMM) . but Manual tell me: max 16MB double side SIMM.
Unterstood, how did this hacking: Good, it help much !
Greetings
matt

What are you doing to improve these bioses ?
I have a 486 with a SiS 471 chipset and I think I know which jumper is supposed to turn write-back mode for CPU cache, but flipping it prevents the motherboard from starting up.

A question: did you have modified my uploaded BIOS Image ? If yes, you're damn fast 😀

I have compared my and your modified BIOS image.. only 42 byte is changed

Madao wrote on 2020-11-10, 04:55:

A question: did you have modified my uploaded BIOS Image ? If yes, you're damn fast 😀

I have compared my and your modified BIOS image.. only 42 byte is changed

My board had the same BIOS that I modified. The only difference is to set up some of the chipset registers the way that the M912 BIOS does.

dionb wrote on 2019-10-20, 23:14:

Anything with good documentation for starters 😉

I really like FIC motherboards for build quality and documentation. Very sad my 486-PVT is dead. That would fit your requirements perfectly. It has a Via chipset.

That aside, I like my Biostar MB-1433UIV with UMC8498F chipset, although its jumpers are baaaad - but that's no different on most 486 boards. Only a few Asus ones seemed to have been made with easy config in mind.

Hello!
I'm trying to get an IBM 5x86 running on a Biostar MB-1433 UIV (sorta, it's a weird "Markvision Flexboard" branded variation of the board, look here: How to identify this 486 motherboard techmedia UMC um8498F) but according to this Cyrix compatibility list https://web.archive.org/web/19961221092122/ht … 86/5x-mblst.htm I need BIOS version "AMI; 7/24/95" from Rev.7 of the MB-1433 UIV
Do you happen to have that version/BIOS? If so, would you be able to dump it?

Thanks in advance!

Those both look like really good VLB motherboards.
To find out which is faster. Set the motherboard CPU type up for 486dx-33
And pop in the Pentium overdrive CPU.

Benchmark both motherboards.

I use Philscomputer lab benchmark CD.
Just download, extract and burn to CDR.
Put in your 486 and run benchmarks.
For best results its best to copy the CD to your 486 hard drive and then run the tests.
https://www.philscomputerlab.com/dos-benchmark-pack.html