I built up my active ISA terminator board, and the Xi 8088 ran all night without memory test failure or lockup. This may be a permanent fix.

I need to dig out my ccopy of Champions of Krynn. 😉

nice!!

More details? Memory testing has been better but general stability is still TERRIBLE.

https://www.electronicspoint.com/threads/isa- … uestion.113863/

When you run your champions of krynn test, are you just leaving it at particular stage of the game, or are you doing something else? I had mine running in "demo mode" (where it shows various cut scenes, goes through a battle, etc) and after about four hours it exited with a "runtime error 200", which is a turbo pascal division by zero error. I'm not sure if that is a corruption issue or just a bug in the game for leaving it in demo mode for an extended time.

As for more details on the terminator board, I have a write up at http://www.smbaker.com/diagnosing-xi-8088-stability-issues

I let it loop the demo scene. I have never gotten a runtime error 200, it just "freezes", usually at "A battle begins" with the dragon head.

Also, the XT memory expansion doesn't work either. I don't believe external memory is an option with the xi8088.

keenerb wrote:

Also, the XT memory expansion doesn't work either. I don't believe external memory is an option with the xi8088.

Given that my buffered RAM board is quite similar to the Lo-Tech 1 MB board, I wonder if they suffer from the same problem with the Xi 8088? I've sent the lo-tech guy a message to see when the boards come back into stock; I'd be curious to build one.

Scott

James has told me he's switching from selling boards to licensing his designs to other builders. I wouldn't expect to find any more in stock at lo-tech.co.uk.

keenerb wrote:

James has told me he's switching from selling boards to licensing his designs to other builders. I wouldn't expect to find any more in stock at lo-tech.co.uk.

Well, that's too bad. They looked useful and fun to built.

I tried putting my scope on the Xi 8088 with and without my active terminator board. I'll be damned if I can find a difference significant enough to justify it working fine with the terminator board, and being an epic failure without. Perhaps it's just and occasional weird glitch on one of the bus pins, that's being damped by the terminator board. Maybe if I had a logic analyzer, I could program it to trigger on the glitch. But, I don't know how much more time I'm going to invest in this. If it's working reliably, even if I can't explain it, that's good enough for me.

The Xi8088 does have a lot of 10K pullups... I do wonder if stability would be improved if a smaller value resistor was used.

Well as long as it works well. I'll let you run it for a while and see if it's TRULY better and not just a lucky run, and order a few myself. They look simpel enough.

Champions of Kyrnn appears locked up at the moment... What I have is definitely an improvement, but it's not "right" yet.

Scott

That has been my experience, memory error rates don't particularly indicate stability in othe areas.

Executed 99 "full" memory tests, had errors on 37 of them btw.

I've had 100% success rate on 75 "full" memory tests, in 8 Mhz Turbo mode.

smbaker wrote:

I've had 100% success rate on 75 "full" memory tests, in 8 Mhz Turbo mode.

It's entirely possible the CoK demo runs out of memory or suffers from an integer overflow, I can't say for certian it's a good test. I have also used Ultima 3 as a long-term benchmark, it's run for four days without crashing before. Typically it suffers from garbled sprites as well if there's a pending crash.

I'll loop CoK demo in dosbox on one of my machines at home and see if it eventually locks up as well.

Maybe I'll start with Ultima 4. That's the first one I recall playing.

I agree it could just be something weird with Krynn. In particular I'm using a version I downloaded from the web, rather than my original disks, which are packed away in a box somewhere.

It's been running Ultina 4 for over 24 hours without any issue, but I don't know that means a whole lot. The Krynn demo had a lot of disk access. Ultima 4 is just sitting there with a few jester sprites wandering around the throne room.

I'm temped to write my own burn-in test script, something that exercised memory, disk, and video.

Scott

Running something like Ultima in a loop is probably only testing certain memory blocks.

Maybe something like copying a largish file and comparing md5sum over and over and over again would be a much better test. It would test both read/write and memory.

That's what I was thinking, something that involved reading files and filling main memory with their contents, then a hash check to see if the memory contents are correct.

I might also see what it takes to cause some DMA transfers. Simply using the disk may involve DMA -- I haven't looked into xt-ide to see if it uses DMA to read from the compactflash or not.

I don't think the compact-flash XT-IDE from James PEarce uses dma at all.

keenerb wrote:

I don't think the compact-flash XT-IDE from James PEarce uses dma at all.

It doesn't, only the CPLD version does.

Well there goes that easy plan.

I'd still like to work some DMA into the burn-in test. Maybe I can dig up some memory-to-memory DMA code.