Thanks for the options and suggestions.

Unfortunately the board remains stubborn and recalcitrant. Maybe there's some secret agent stuff in there or some hidden coded message!

The board is cosmetically perfect and would pass for NOS. It's nostalgic, coming from a time when I got into PCs. But perhaps it's time to move on to the next project

I think one of those old BIOS chips have a failed/corrupted section in one of them. Have a P2B bios chip that has a bad spot and would lock during POST, found it with a TL866 when trying to read the contents to make sure it was same as what was supposed to be programmed in it when it failed a flash upgrade.....Just a thought.....

Horun wrote on 2021-06-11, 01:16:

I think one of those old BIOS chips have a failed/corrupted section in one of them. Have a P2B bios chip that has a bad spot and would lock during POST, found it with a TL866 when trying to read the contents to make sure it was same as what was supposed to be programmed in it when it failed a flash upgrade.....Just a thought.....

Yes, but that's why I suggested testing in PCem, and it sounds like there it worked fine.

jakethompson1 wrote on 2021-06-11, 01:25:

Horun wrote on 2021-06-11, 01:16:

I think one of those old BIOS chips have a failed/corrupted section in one of them. Have a P2B bios chip that has a bad spot and would lock during POST, found it with a TL866 when trying to read the contents to make sure it was same as what was supposed to be programmed in it when it failed a flash upgrade.....Just a thought.....

Yes, but that's why I suggested testing in PCem, and it sounds like there it worked fine.

OK I miss read that part, when he said "I have a TL866II+ and dumped the BIOS chip. PCEM shows the same screen." I thought he meant it showed no F1 like on his screen.
Well was a thought 😀

Sorry about that if I mis-worded anything.

Yes. My ripped BIOS from the original board worked fine in PCEM. I found the next revision BIOS, G22-2 in a pile of files. Plugged it into PCEM. And it worked fine there too. I was able to get into the screens and make changes, in PCEM.

I used 2 different 27C256 EPROMS, burned them, HI and LO, plugged them into the bastardboard and it still failed in the same manner we've been discussing.

Meantime, from PCEM:

Ok.... Glanced through the whole thread again. Have you tried it with the turbo button jumpered or unjumpered - whichever is different than it has been? What about the other front panel connectors such as reset?

Is it mounted in a case? If so, try it outside the case.

Have you tried with a different video card?

If none of that works, my bet is either the CPU or cache.

Forget about the not being able to get into the BIOS. That is just because it is resetting before that option is even available.

A brief tip when using a 'prommer and old old old chips:

I had to clean the sides of the IC's pins before I got a consistent read. This is because regular motherboard sockets wipe two sides of the pins. Fine. That's normal. But the TL866 and all other ZIF-sockets like it make contact on the sides that aren't ever touched by a standard DIP socket. The contacting surfaces are offset by 90 degrees.

There is no wiping action to remove the thin layer of oxidation that can build up on those narrow edges.

The simple solution is to slide the chip back and forth in the ZIF socket while it's halfway closed. This gives you the same beneficial wiping and cleaning action as removing and re-inserting in a standard DIP socket. Or manually clean the IC if you don't want wear and tear on your ZIF socket.

Try that if you have parts that aren't working/testing/programming in any 'prommer. I also had the issue with some 74xx logic from the 1970's. They'd work in the original Apple II motherboard. But would fail ID and testing when in my 'prommer. Until I cleaned the other sides. All 4 sides of a square/rectangular pin.

Ideally all vintage ICs should be given a light cleaning when the opportunity presents itself. As far as abrasives go, I try to use an eraser, gently, or paper. If needed maybe a relay contact cleaner pad. You get bonus points for using Deoxit D5 or other appropriate contact cleaner. Boss Bonus if you use Deoxit Gold. Or Gold Guard.

Alright. I'm keeping the bastardboard off to the side of my workbench, but hooked up a power supply, monitor, keyboard, videocard and miscellaneous front panel stuff like Turbo, Reset, KeyLock, TurboLED, PowerLED, and Speaker. I have a 3.5" drive and a Multi I/O board standing by, but not connected. Sticking with the 16MB RAM configuration. This way I can continue working on other things and come back to this as ideas present themselves.

I've been testing the board bare on my workbench the whole time. No case within miles.

I have tried 2 videocards.
STB Evolution Cirrus Logic based CL-GD5422 1MB.
STB Horizon Cirrus Logic based CL-GD5429 1MB.
The first card has an 8/16 bit BIOS jumper, and 2 BIOS chips interleaved for performance I presume.
Both cards have a WaitState and IRQ jumper.
Changing those options doesn't seem to affect anything.
I could try a 3rd card but I don't believe the video subsystem is at fault here. Pretty standard stuff.

The Turbo Button doesn't seem to have an effect. But by looking at the setup screens, the system likely needs to be set to "SLOW" and then the Turbo Button would toggle it between SLOW/FAST. And it likely wouldn't actually change any clock speed. I think (somehow) that it inserts wait states to simulate slower operation. That's a guess. Only did the earlier machines like the ones with the CPU operating at bus speed actually change speeds..? A guess too.

The Reset Button appears to work correctly. It resets when you jam on it.

I can try another CPU later maybe. I have an SX2/50 laying around someplace. It's of course minus the FPU but that isn't an issue here. It's still clock-doubled @ 25/50. Hmpff! Other than parts-binning faulty vs working FPU parts of the chip, was there any other reason to make an SX2? Sounds like useless busywork by marketing. Pointless granularity in a rapidly changing product lineup!

The S-Spec of the existing CPU is SX626.

Not much is left. I have yet to swap the CPU. But I think it's alright.

I have high confidence the following stuff is in good working order:
CPU
SIMMs
Power Supply
Videocard
BIOS EPROMs
82C206 IPC w/CMOS memory, RTC, IRQ & DMA controllers
8242 Keyboard controller w/Phoenix firmware

Remaining items to check:
Micronics Chipset MIC461A & MIC462
2nd level cache (socketed)
ISA slots
Glue and miscellaneous logic (some socketed)
POST card test codes
Try to determine the function of all the <RESERVED> DIP switches (no documentation)
X-Ray the PCB for cracked traces (no, not really)

Not sure how I'd go about testing a chipset and miscellaneous glue logic without datasheets. Only thing would be to observe signal levels and address/data activities to see if they're healthy and working. Though I suspect they're just aggregations of AT PC logic - some of which can be tested at the connection points of well-documented standard components like the ISA slots, CPU, or DRAM. The 82C206 is adequately documented too, with timing diagrams. So there's that.

Not sure what to do with the cache. It either works or it doesn't.

Searched high and low for chipset datasheets. Would be a boon to have. Love to know how the glue logic connects the TAG RAM and chipset and L2 cache with the rest of the board. A curiosity if anything.

I've searched hi and low for datasheets for the micronics chipsets and have never found any. Looks like you have tried a lot. I have a similar issue on a board with a missing keyboard controller. I use it for testing suspect ISA cards. May be something to check out. See if it boots without it and you get the same error. Often the battery has killed a trace that's part of it.

Also (as a last ditched effort) does it work if you connect a battery? I have a Phoenix BIOS on an IBM PC-330 that plops its self if the battery isn't connected.

Keatah wrote on 2021-06-11, 08:00:

I can try another CPU later maybe. I have an SX2/50 laying around someplace. It's of course minus the FPU but that isn't an issue here. It's still clock-doubled @ 25/50. Hmpff! Other than parts-binning faulty vs working FPU parts of the chip, was there any other reason to make an SX2? Sounds like useless busywork by marketing. Pointless granularity in a rapidly changing product lineup!

The S-Spec of the existing CPU is SX626.

Hum, this board seems older than DX2s, the only other pic I find of one online has a DX in it. Do you have a DX50, or maybe a DX40 to try in it?

You've mentioned that the board has a 50 MHz crystal.
I thought that for 386 CPUs, the crystal is double the clock speed, and for 486 CPUs, the crystal is equal to the clock speed, but with this being a DX2 CPU, it would actually be half the clock speed.
So I think the crystal should be 25 MHz. Unless there is some kind of divider on the board?
Could the DX2-50 be badly overclocked and actually running at 100 MHz?

It's more like a 90:10 rule, 90% of each class are like that but 10% of them do the other thing.

No go with the 8242 keyboard controller out of circuit. No video even. I feel confident the chip is working. I don't believe there's any damaged traces in that area of the board. Unless it's highly intermittent.. Or under a soldered component.

Having a battery connected (or not) doesn't seem to make a difference.

I happened to have a DX50 that I use for show-n-tell. No difference. Maybe the memory counted almost imperceptibly slower with the slower core because it's a non-clock-doubled CPU running at 25MHz now.

I connected my 'scope at various points and the bus activity is clearly 25MHz. No radically massive overclocking today. It's firmly 25MHz external, 50MHz internal.

This board design may be older than a DX2/50 chip. But there is an OEM label in the corner saying DX2 alright. Yes. The board has a copyright date of '87-'92. And I believe DX2 processors became available in 1992.

Things were changing pretty quickly back in those days. A product's lifecycle might be less than a year. More like 6-months when talking speed grades of CPUs. Just look through some catalogs. What was available in the spring was deeply discounted or phased out by winter.

Ok. I think the L2 cache might be a problem. Working on it.

For Real! Thanks to all! The world now has one more slightly smarter hobbyist, and another functional vintage motherboard.

Excellent news!!! I'm glad you were able to get this system working again. What was the problem after all?

The configuration of the Level-2 cache. The board's jumpers and switches were set for cache size of 256K, but the installed cache was actually 64K. They looked to be the original factory-installed chips, just that somehow the board was mis-configured.

8 chips of 8K x 8 SRAM vs 8 chips of 32K x 8 SRAM.

It was suggested the cache could be at fault. So I pulled the chips and put them in my TL866 'prommer which can test SRAM. After 3 of them showed bad I got to wondering. Got the datasheets for them and a size discrepancy was immediately apparent. These 8 chips were 8K x 8 meant for a 64K cache. I retested them as 8K chips and they were fine. Reconfigured the DIP switches and jumpers and it powered up as usual. But this time it gave me the F1 and F2 options. F2 took me right into the BIOS setup screen.

In effect I suppose the improperly configured cache made the first 64K work incorrectly. This was also mentioned earlier the discussion - BIOS was never able to proceed far enough.

Back in the day it would have been a big deal. Maybe a $50 or $100 kind of big deal. Today my scrap box has all kinds of stuff. Including extra SRAM parts.

Excellent news, congrats. Are you able to edit the post title? What would you say to add some short note in the end of title that issue was solved and root cause was L2 cache, including also Mainboard model numer. This will help others in the future with similar problem.

I can't edit the title. But I could add a line to the first post. I will look up the mainboard numbers tomorrow. There might be two numbers, one for the DX and another for the DX2 version. But they would behave the same.

These motherboards were typically sold only as OEM parts to computer assemblers, like Gateway or whatever else company. Therefore they were most likely outfitted with their CPU and cache at the the factory. Boards like these don't support many CPU upgrade options, if any.