help needed with finicky 386 board

[luckybob]

That's a picture of the board. I got it off ebay for like $20 a couple months ago, and it has sit in a bag since. Mostly because of other projects. Tonight I found myself "getting around to it" and firing it up. It seems to run fine but has developed a few quirks. The most major of which is some sort of "gate a20" error i get loading dos. The dos boot disk I use basically loads a xms manager (dos 6.22 himem.sys), a cd driver, and cute mouse. Pretty basic. I think it would be best if I list what I did and in the order I did it. If for nothing else that retrace my steps.

drop in 4x 4mb simms, test with vlb video, boots
spend 2 hours finding jumper settings online LINK: http://stason.org/TULARC/pc/motherboards/U/UNIDENTIFIED-486-PANDA386V.html
install 256kb cache
benchmark*
remove processor and install amd386/40 (change appropriate jumpers)
10 minutes to find out vlb wont work with this processor and install tseng et4000, works
benchmark
revert to 486dlc and install cyrix 387 co-pro
benchmark
install 32mb ram
night goes to hell
revert to previous ram
standard troubleshooting
15 minutes of staring at the bios settings
realize its almost 4am and you might need help

* During my testing, I find out that the board automatically enables the L1 cache on the chip. However the default setting for the L2 is off. using the cyrix program to "turn the cache on" does nothing. IDE drive performance is WORSE than my "real" 386. in benchmarks it scores a 19 in speedsys and manages 21 fps in 3dbench. installing a cyrix math-co drops this to 9.9 and 18 respectfully. cache benchmarks show plateaus in the expected locations. 1kb,256kb

For the meat and potatoes of my evening, it seems when I installed 32mb of ram (8x4mb) it seems that everything hit the fan. I got no errors during post, but once himem was called I get this error; "Unable to control A20 Line". speedsys wont do its memory benchmarks however everything else seems fine. after several hours on teh internets, I'm no closer to a solution than before. One website says that the a20 line is the keyboard controller, but that doesnt make sense to me. This is a memory issue, and the keyboard works just fine.

please forgive me if I'm hard to understand, its now almost 4:30am, but I have 2 real issues. The most pressing issue is the himem a20 issue. This board was just fine the first few hours of the night, I dont know what happened or where, but some insight is welcome. The 2nd issue is moot unless I fix #1, but I'd like to know why the performance goes down the shitter when I put a cyrix math-co in this thing. I would think a cyrix core 486 and corresponding cyrix math-co would at the very least do no harm...

thanks in advance, time for bed.

 

[njroadfan]

Going by the Daewoo logo and power connector, this board likely came out of a Leading Edge machine.

 

[njroadfan]

For the A20 gate issue, its likely related to the keyboard controller. First try another keyboard, sometimes that clears it up. Second, try another keyboard controller.

 

[lucasdaytona]

What is himem.sys version? It matchs with the DOS version you're using? What are the modules you're using with the 386? Have you respected the memory config as told in stason? This keyboard is fine(tested in another board) ? The battery is fine? Did you tested everything without the 387?

 

[Agent Orange]

Lucky:

I think you may have a HIMEM.SYS configuration problem. Try this:

DEVICE=C:\DOS\HIMEM.SYS /M:1

Also, try bypassing the CONFIG & AUTOEXEC on bootup and see if the A20 problem clears.

BTW love that chip! 486 instruction set on a 386.

 

[kithylin]

Have you tried running memtest86 via floppy on the 32mb of ram and make sure first that no bad modules exist?

 

[luckybob]

Have you tried running memtest86 via floppy on the 32mb of ram and make sure first that no bad modules exist?

The speedsys benchmark that I run does a intensive test of the cache and all ram. Not as extensive as memtest, but a test none the less. I will do this next.

What is himem.sys version? It matchs with the DOS version you're using? What are the modules you're using with the 386? Have you respected the memory config as told in stason? This keyboard is fine(tested in another board) ? The battery is fine? Did you tested everything without the 387?

Its the version that came with my dos 6.22 disks. The modules I have are 4mb x 9 but they are "fake" parity. I use what I have and there has been no issue until I tried to use all 8 slots. I haven't tested the keyboard on another system, I will now. There is no battery, and yes I did. First rule of troubleshooting is "remove everything, add things back one at a time"

Lucky:

I think you may have a HIMEM.SYS configuration problem. Try this:

DEVICE=C:\DOS\HIMEM.SYS /M:1

Also, try bypassing the CONFIG & AUTOEXEC on bootup and see if the A20 problem clears.

BTW love that chip! 486 instruction set on a 386.

My issue is, it was working fine until I added more ram. And it diddnt stop when I went back.

For the A20 gate issue, its likely related to the keyboard controller. First try another keyboard, sometimes that clears it up. Second, try another keyboard controller.

Do I just steal one off another board? I'm going to assume that the pin-outs are all going to be the same, right? I don't have another board with an identical chip. It seems everyone is different in some way.

 

[twolazy]

Whats funny is I was bidding on this board and you ended up winning it >.<

Quest for a 386dx board continues...

Anyways, that board looks really dirty, could be the photos/lighting though. I start by giving it a good cleaning, and spray the memory slots, keyboard controller socket, and keyboard port, and maybe the cpu/mathco sockets too, with some electrical cleaner and see if it clears up... You can use non chlorinated break cleaner, available cheap at most car stores, in a pinch if you dont have anything better available.

As for the poor mathco performance... Could be that the system expects to see cache memory or maybe a dirty/tarnished socket? Just a guess... Either way no cache = slow performance.

Didn't early leading edge machines require an XT keyboard *scratches noggin. I know the Model D did .... Can you read the keyboard controller model/part # please. Just a hunch. Also if its your standard typical clone keyboard controller, you can usually substitute for a different part number of the same brand as long as its around the same era and bios brand.

 

[SpidersWeb]

With the math situation, I suspect the 486DLC chip is just faster at running them than a 387.
Being a "D" as far as I'm aware in the 486 range meant it already has a FPU built in.

Edit: apparenrtly not the case with this CPU, math co-pro should be a 387DX-40, and it should be faster with it fitted :/

 

[twolazy]

Yep! >.< Used to own a machine with that very same CPU! Thats why I'm leaning towards a dirty socket or anemic psu on the mathco problem. Doubt its the missing cache but could be, anyone have the manual to check? Possibly bad tant caps too.

 

[Agent Orange]

Lucky:

One other suggestion: If you changed the board configuration try resetting the BIOS. I have one 486 board in particular that needs to be reset anytime you add or subtract hardware. Just a shot in the dark. Good luck with it.

 

[luckybob]

Just an update:

I haven't forgotten anything or anyone, it is just that my collectors edition of diablo 3 is here and like a magpie, I'm easily distracted. I also *just* acquired an ATI mach 64 with 2mb vram. So I WILL be finishing this project, come hell or high water. (thank you collector!)

 

[luckybob]

6/25 update:

I found out what was wrong. I had a jumper setting that was incorrect for the type of cpu I had. I set it back up for the cyrix 486 chip and it worked great. LAST NIGHT. I turn it on today and find that my ati mach 64 (vlb) no longer seems to work. Do I have bad luck or what? the vlb slots are still working, but the card will not post in any vlb board I have now.

either way, here is the board on the th99 database: http://stason.org/TULARC/pc/motherboards/U/UNIDENTIFIED-486-PANDA386V.html

 

[themikepeng]

Processor maybe, I had a 386/486 board with a wierd keyboard issue, it had the same tx 486 dlc 40 cpu! The keyboard blinked when it started up but I couldn't input anything.