Unstable (but lovely) 286 clone

[Paul-VT]

Hello Everyone,

I recently purchased a 286 clone from eBay. The steel horizontal case ans PSU are in great shape and for $100 I thought it would be a good start to a fun retro project. To my delight, the motherboard inside looked to be in good shape also. With a bit of vinegar, alcohol and elbow grease, I removed the slightly leaky battery and cleaned the motherboard.

It also came with a working MFM/Floppy controller and working ST-225 which tested good in my IBM 5150.

The motherboard is a 286-12Mhz clone with the VLSI 82C100 chipset and 1MB RAM. I have found similar boards online but no exact match.

The board boots fine but there are some issues during use:

1. With a known good vga card (Trident 9000) on the bus, the board always boots but half the time the vga card's bios message which should be in color text shows in white text only and no colors are possible in any programs/games. The other half of the time it boots properly and color vga games work fine. No difference between cold boot and warm boot with reset switch or ctrl-alt-del. This is the main concern as it happens with no other cards on the bus.

2. The UMC IDE controller seems flaky, especially with a IDE to CF adapter. I have tried different CHS values in the bios and different CF cards but I consistently get random write errors ("General failure error writing to drive C"). Especially when installing DOS 6.22 from floppy. This controller/CF adapter combo tested good in my 486.

3. The 3-Com, parallel tasking, 3c509 works but using ftp to get files from my server causes lockups and I have to reboot. This networking card tested good in my 486.

I have tried different CPUs (12 and 16 MHz), and RAM. The board is not overclocked. I have de-soldered the keyboard connector and cleaned underneath. I have checked for broken traces with my meter. I have removed and re-seated all socketed chips. I have tried different power supplies. No luck.

Unfortunately, my 3 year old Siglent 'scope now refuses to boot (tried the 'math' button trick already but no) so I can't use it to check for clock or voltage instability right now.

Could it be a bad BIOS or flaky old BIOS chips? A sneaky bad old tantalum capacitor? Unstable clocks and voltages?

Any thoughts would be much appreciated.

Thank you.

-Paul in Vermont

 

[acgs]

The board boots fine but there are some issues during use:

1. With a known good vga card (Trident 9000) on the bus, the board always boots but half the time the vga card's bios message which should be in color text shows in white text only and no colors are possible in any programs/games. The other half of the time it boots properly and color vga games work fine. No difference between cold boot and warm boot with reset switch or ctrl-alt-del. This is the main concern as it happens with no other cards on the bus.

There isn't enough info in your post to be sure about things, but I'm going to take a stab at this one and guess there isn't a problem with the motherboard or the VGA card.

If I'm right, the problem is with your monitor and the VGA/motherboard combo. I'm guessing that your monitor has a power saving mode that is activated when there is no signal coming from your VGA card. The possibility is that your motherboard isn't always initializing the VGA card fast enough, and on those times there is no signal to your monitor long enough for it to go into power saving mode. When the VGA card is initialized and the system tries to determine what kind of monitor it's attached to, the monitor is still "asleep" and doesn't respond before the system times out. The system then defaults to the "fail-safe" setting of monochrome monitor.

I've experienced this same problem even on 486's with certain power saving monitors. If you can, try hooking it up to an older monitor without a power saving mode and see if the problem goes away.

ONE FINAL QUESTION: When you say it doesn't happen with any other cards on the bus. Do you mean if you try other EGA, VGA, etc. cards or literally any other type of card on the bus?

 

[maxtherabbit]

It doesn't have anything to do with the power saving, the problem is the difference between how modern VGA displays communicate what modes they support to the host PC (DDC) and how early 90s ones did it (voltage level sense pins). The reason it seems erratic in result is the video card is expecting a fixed voltage level on the pin but the monitor is trying to use it so send serial data.

There will often be a jumper on the video card to override this behaviour and force colour mode. Sometimes it's a trace to cut. Post high res pics of your video card and I'll help you find it

That is a biostar motherboard, you can tell from the logo on the ROMs. I have the exact same board so I can try to help you with any additional issues. Regarding the other problems, I'd suggest you try using a real mechanical HDD or a DOM instead of a CF card and see if your woes go away.

https://theretroweb.com/motherboards/s/biostar-mb-1212v

 

[acgs]

It doesn't have anything to do with the power saving, the problem is the difference between how modern VGA displays communicate what modes they support to the host PC (DDC) and how early 90s ones did it (voltage level sense pins). The reason it seems erratic in result is the video card is expecting a fixed voltage level on the pin but the monitor is trying to use it so send serial data.

I see I didn't make myself clear enough. I apologize for not giving enough information about my stab-in-the-dark regarding the problem.

I assumed the vintage 286 and VGA card were being hooked up to a vintage CRT monitor from the early '90s or earlier. If the monitor is newer, then maxtherabbit is probably right. But, when using a power saving monitor of that vintage and a TRIDENT 9900 series VGA card, I often experienced the exact same "erratic" behavior of dropping to monochrome-only mode that the OP was describing, while I never experienced the same problem by hooking up a non-power-saving monitor to the same VGA card and motherboard. The problem also appeared using that card in a 486.

Studying the effect over several weeks, I noticed that I could always get color mode by turning the monitor on after the computer at just the right moment. I also could get color mode by rebooting the computer twice: the second reboot during the POST of the first.

 

[Paul-VT]

Thanks for the replies. I thought it was the Biostar Motherboard MB-1212v but there are minor differences. At POST the board claims to be a MBVLSI-168

I will try with a few different monitors. The MFM controller and 20MB hard disk seem to work okay but I'm wary of using such a gem for the long term, and it is really LOUD!

 

[maxtherabbit]

Thanks for the replies. I thought it was the Biostar Motherboard MB-1212v but there are minor differences. At POST the board claims to be a MBVLSI-168

Maybe it's this one then:
https://theretroweb.com/motherboards/s/biostar-mbvlsi-168-b

Mine is branded Shuttle, but also identifies itself as MBVLSI-168 in the POST string, pretty sure they all just used a copy of the same AMI BIOS.

Regardless of which exact variant it is, all those VLSI 100 series boards are basically identical reference designs.

 

[Svenska]

Hi,

1. With a known good vga card (Trident 9000) on the bus, the board always boots but half the time the vga card's bios message which should be in color text shows in white text only and no colors are possible in any programs/games. The other half of the time it boots properly and color vga games work fine. No difference between cold boot and warm boot with reset switch or ctrl-alt-del. This is the main concern as it happens with no other cards on the bus.

This is a communication issue between the VGA card and the monitor, as maxtherabbit has explained: Your card expects a fixed voltage on the ID pins to detect the monitor type, while the monitor uses DDC (basically I2C) to communicate its type serially.

There may be a jumper / DIP setting on the card to override the detection. Or you can use a modified VGA cable (or modify the card itself). A third option is to run a small program to force the card into "color" mode.

2. The UMC IDE controller seems flaky, especially with a IDE to CF adapter. I have tried different CHS values in the bios and different CF cards but I consistently get random write errors ("General failure error writing to drive C"). Especially when installing DOS 6.22 from floppy. This controller/CF adapter combo tested good in my 486.

I had all kinds of random problems with my 286 clone when using a CF card. Not only write errors, but also data corruption when writing.

However, the bigger problem was data corruption when reading: Memory corruption. Applications would randomly fail for no reason. I suspect that either the executables were loaded incorrectly (DOS applications generally do not check themselves) or that disk activity caused memory to be corrupted. This caused lots of head-scratching until I physically removed the CF card and the system started to behave.

None of these problems appear when using the XTIDE BIOS, which I added as a network card EPROM. No data corruption and no weird crashes since.

3. The 3-Com, parallel tasking, 3c509 works but using ftp to get files from my server causes lockups and I have to reboot. This networking card tested good in my 486.

The CF card issue on my 286 clone was most obvious when doing network traffic. Keep in mind that network traffic to modern machines will cause lots of bus activity in a short time, increasing the probability of collisions destroying data. If that data is executable code, you're done. On a 286, the ISA bus is the only system bus, shared between memory and IDE (later systems tend to isolate the memory from the rest of the system).

Could it be a bad BIOS or flaky old BIOS chips? A sneaky bad old tantalum capacitor? Unstable clocks and voltages?

I'd suspect the CF card. Try the XTIDE BIOS.

Although I had a run with a strange Cx486 system: It would run mostly fine, but performance was garbage and every once in a while it would hang. Landmark showed performance values all over the place. Turns out the main crystal had failed and the oscillator circuitry was doing its best to turn random circuit noise into a clock signal. It worked surprisingly well...

 

[Anonymous Coward]

On this motherboard, are the ISA slots running off the same clock as the CPU?
If you run in de-turbo mode (I assume 6 or 8MHz), does the data corruption problem go away?

 

[VERAULT]

Lovely huh?

 

[maxtherabbit]

something about the AMIBIOS on these systems just doesn't tolerate CF cards, it may be related to the lack of multisector transfers by the cards but that's just a guess

I second the motion to try the XUB

 

[Svenska]

True, I should have noted that my 286 clone (with the massive CF card-caused data corruption) also uses an AMIBIOS.

 

[RFE]

Okay, enough talk about the monitor/VGA card communication failing But this was also my first thought when reading the initial post before all the other answers. The card goes into black/white mode for thinking to detect a monochrome screen.

For the CF card issue, I would not say it is the AMIBIOS (maybe certain revisions but I have a Headland HT12 board with a 286-16 which runs pretty well with AMIBIOS and 256MB CF Card - I always use industrial cards, my small 256MB cards for very old systems are from Xmore)

A pity your scope does not work because my feeling is that the clock might be the elephant in the room. On the early 286, the ISA bus was kind of a local bus at CPU speed, but 12MHz is a little too much. There must be something like a clock divider or at least waitstates.

The power lines look a little corroded around the keyboard connector, seems like a not too bad battery damage that did likely not harm anything - but you never know, maybe worth looking if all lines are okay.

 

[maxtherabbit]

That board runs the ISA bus at a normal 8MHz

 

[Svenska]

For the CF card issue, I would not say it is the AMIBIOS (maybe certain revisions but I have a Headland HT12 board with a 286-16 which runs pretty well with AMIBIOS and 256MB CF Card - I always use industrial cards, my small 256MB cards for very old systems are from Xmore)

I had a single CF card which worked reliably in my system (using multiple I/O ISA cards). All my "industrial" cards failed really bad, and some consumer cards were good enough to make debugging a huge pain. On those, I didn't notice the corruption until I started checksumming the files on both the 286 and the originating system. As far as I am concerned, it is the combination of marginal timing on the ISA bus paired with code which does not cope with it.

My recommendation would be to checksum your data, just to know for sure.

That board runs the ISA bus at a normal 8MHz

That is quite good actually. Many boards used CPUCLK/2 instead, so 6 MHz on the OP's system. I wonder if Paul succeeded in fixing his issues.

 

[Paul-VT]

I've not been able to work on the board for a few days. However, I did try three different monitors with only the VGA card plugged into the bus. All monitors behaved the same way. During about half of the reboots (cold, reset, ctrl-alt-del) I would see the VGA card's bios mesage in color, the other half in monochrome. The monitors were a mix of LCD, and two different CRTs from the late 90's. I tried to be sure the monitors were not in power saving mode when I booted the computer by making sure the power indicator was green, not amber. So no luck there.

My 'scope is now working again thanks to a helpful tech at Siglent.

Back to the basement on my day off. More soon.

-Paul

 

[Paul-VT]

Success!

Probing the board with my scope didn't find any obvious problems. I decided instead to try a different vga card. So I bought a Cirrus Logic CL-GD5429 on ebay and guess what? It works perfectly. Boots up in color mode every time and works flawlessly in games so far.

I also found a IDE controller and an 80MB Maxtor 7080AT drive that also work perfectly. The drive has that late '80s clicky sound I remember from the 286 machines I used in college in the late 80s. I will save the Seagate ST-225 and its MFM conroller as a spare for my IBM 5150.

The problem I had with the FTP client stalling/crashing during transfer was fixed by tweaking the ftp settings in the network.cfg file. Transfers still pause for maybe 20 seconds periodically when transferring large amounts of files but complete normally eventually.

I'll probably keep playing around with a CF card to see if I can get it working with the IDE controller and post my results if I can get it working reliably.

Thanks again.
-Paul in VT

 

[Svenska]

In other words, you have taken none of my recommendations and done something completely different instead.
Congratulations anyway; a working system is always nice to use.

 

[maxtherabbit]

Simply posting a photo of the original video card would likely have been enough for me to tell you how to stop the monochrome problem...

 

[acgs]

Congrats to getting everything to work!

 

[Paul-VT]

These are the trident TVGA 9000 cards I tried. Both showed the same mono/color problem at post and after booting on several different monitors. That's why I got the Cirrus Logic VGA card.

What identifies these cards as incompatible with my motherboard? And how to remedy? If you want one, send me a PM.

Topic for another day: The second card came with faulty 286 motherboard (weird keyboard errors and won't post at full CPU speed). But the board did come fully populated with 36 411000 ram chips which all work

I've been trying to piece together a nice 286 for a few years now and happy with what I have now. Thanks again everyone.