RAM-testing - odd errors

[Iznougoud]

Recently, I've experienced infrequent crashes on my A2000. Black of frozen screen, blinking power-LED or just a spontaneous reboot.

This has caused me to check the memory, and the results were so odd, that I had to use my unexpanded A1200 as a reference. This is what I got:

A2000: 1Mb ChipRAM, 128Mb FastRAM. Blizzard 2040. ROM3.1 AOS3.9.
A1200: 2Mb ChipRAM. 68020/14. ROM 3.0 AOS3.1.

MemTest A2000: No errors when Workbench is loaded. Errors reported when Workbench isn't loaded(ChipRAM & FastRAM, which is strange, since I have the same amount of ChipRAM available regardless of whether Workbench is loaded or not).
MemTest A1200: Reports errors.

CheckMem A2000: No errors.
CheckMem A1200: No errors.

Memory Diagnostics (MD) A2000: Now it's getting exciting. Reports errors when Workbench has been loaded. Reports errors outside Workbench when a soft reset has been done. Reports NO errors outside Workbench when a cold reset has been performed(!).
Memory Diagnostics (MD) A1200: No errors.

Now, I find these results inconclusive. To put it mildly. One would imagine that the OS3.9 ROM-image might mess things up, but given the variety of results that doesn't seem likely either.

Does anyone have a clue?

[Hodgkinson]

FYI: The Advanced Amiga Analyser software has a RAM tester built-in, see http://l8r.net/install/misc.html

[Zac67]

Sounds more like a software/driver issue. Testing RAM is most meaningful when no software or as little as possible is loaded.

[Iznougoud]

That would be a reasonable assumption. But given the results from testing this far,  I doubt it's a driver issue. On the contrary. A driver issue should cause the errors to occur when Workbench is loaded, rather than the other way around.

@Hodginson; that link proved dead. But from what I managed to find out about it, you're referring to a piece of hardware that would be nice to have, but probably harder to come by than a completely new A2000-setup  :-D

[Hodgkinson]

Dud link? The page is loading fine at this end :-)
Direct link to the file:
http://l8r.net/install/misc/aaanalyzer2.DMS

BTW, You could always just build your own, as I did :-D . Fortunately the program shows all the wiring links on the diagnostic screens, so all you need is a few D-connectors, a couple of hoods, some wire, some resistors and a few LED's for the power rails. And an old audio cable to cut up if you want to connect to the audio pins in the serial port ;-)

http://www.amiga.org/gallery/index.php?n=2481

(Not quite sure why they need a huge armour-plated box for a couple of LEDs, mind you...)

[Iznougoud]

That link worked. I'll have a look at it, although my wiring- and soldering skills arent' what I'd like them to be.

Anyway. Given the contradictions in my testing results. I've decided to focus on MemTest. And as far as I can tell, it reports errors in every environment BUT WB3.9 (I've tested no WB and WB3.1 - it's a reasonable assumption that I would get the same results in WB2.01 and WB1.3 as well). Does anyone even have a theory of what this could be, or could it be that there's a bug in MemTest?

[Hodgkinson]

;-) Oh, sorry, you don't need the plugs and adaptors for the RAM test portion of the AAA software to work. The RAM test will work quite happily without all the additional hardware :-D

EDIT: I think the disk is self-booting with its own system.

[Iznougoud]

Yes. I noticed. For some reason, it does seem to test FastRAM only though (finds the ChipRAM, but never get passed testing FastRAM). Ah well. I suppose I'll just have to disable the A2040-board and run the test again. Not that big a deal.

This far, it's run 22 passes on the FastRAM with no errors. So that's a good sign. I hope. Doesn't, however, solve the mystery of the random crashes.

Edit: My bad. From the looks of it AAA does check bot Fast- and ChipRAM. It just doesn't allow you to choose either.

[rkauer]

@Iznougoud:

 Did you check the power supply and ventilation on the 040?

 Random lockups are common with bad power supplies (due ripple) and when the 040 goes too hot.

 Even if you don't have a oscilloscope, you can check the ripple on the +5V line using a humble digital multimeter (simply put it on the smallest AC scale and read the LCD). Acceptable "noise" is in the 50mV tops!

[Iznougoud]

Or even more simple, I might switch power-supply. The chance of two power supplies going bad at the same time would be - if not astronomical so at least slim. Maybe I even should consider rebuilding an ATX-PSU..

I am however a bit confused by the output given by MemTest..

[rkauer]

Iznougoud wrote:

--zip--

I am however a bit confused by the output given by MemTest..

 Ripple can cause the symptoms you are experienced. Alas anything sourced from the +5V line can suffer. Sensitive devices first (memory, CPU, SCSI and IDE controllers...).

[Iznougoud]

Yes. That much I understand. What bothers me, is that the symptoms doesn't vary - MemTest reports errors outside AOS3.9, but can't find any inside AOS3.9. Always. Not sometimes. Other RAM-testers doesn't find any errors at all. Regardless. If this was to originate from the PSU, I would expect the errors to occur on a more random basis. Regardless of RAM-tester used.