On 3/6/21 3:10 PM, Sean Conner via cctalk wrote: I'll add that, at least in the PC AT world, the switch to real mode is accomplished by writing a value into a reserved cell in CMOS (configuration memory--I wish they'd lose that 4-letter appellation--what, in a modern PC *isn't* CMOS?). Upon executing the reset code, the BIOS checks for the "reason for shutdown". If it was a switch to real mode, then all of the various hardware tests are bypassed, the register file is restored and execution continues. What this means that if your CMOS (ugh!) memory isn't functioning, the switch to real mode won't work. --Chuck

On 3/6/2021 3:10 PM, Sean Conner via cctalk wrote: The 286 can exit protected mode with the LOADALL instruction. Microsoft's extended memory driver pissed off the world (Intel) when someone noticed it was addressing +1mb memory with no hint of the reset.? The rest of the world used a driver and memory buffer algorithm which optimized accessing > 1mb because of the horrible latency involved in what you are describing. Another client and some friends I had were caught with their s**t stinking because they'd used it in their bios (Micro5 systems). I suspect there's some sort of failure involved like that, but not sure how it could live to print the message in the case of going into protected mode but not getting back out. As far as a defective 80286 I can't imagine it passing a lot of the bios test code at all if it was internally damaged. Any POST card (port 80) handy?? Maybe some hints there? thanks Jim

I think that is unlikely because the board comes with 1M onboard and I believe the system is designed to work with just that memory. I can try it though if I can find compatible 30-pin simms (I do have some somewhere). Curious why you think it might require access to memory above 1M though? I am currently working to the hypothesis that it needs the non-volatile memory to work to remember that the reset was part of the POST. The NVR is provided by the RTC element of the 82C206 chip. I may have to get the logic analyser out to see if that is happening or not, but at the moment I still have an intermittent problem getting the board to start at all, I think due to my poor work on the repair wires for the damaged tracks. Regards Rob

follow I should have said that I have found a few bad tracks and I have fixed them by adding wires. Previously it would not even POST, but it does now. The CPU is physically distant from the battery damage. I am trying to understand if this particular test could fail due to external factors or not so that I can then investigate if there are other tracks I need to fix. Incidentally, my repair wires are done very badly, are there any tips on how to do this well? I have ordered some wire wrap wire because I believe that is what I should be using, but I haven't got the wire yet. Thanks Rob