A3000T - memory issue?

[Tumbleweed]

Hi all,

After getting my A3000T running again its been working fine for a few months and then I started getting random "recoverable alerts" just after WB screen appears and when I open certain apps - iBrowse. Was an annoyance and I thought it might be software related. I did a clean install of OS3.2 and it seemed to be ok until a couple of days ago when the 3000T started suffering recoverable alerts and this morning critical failures (screen shot below) before booting WB and got stuck in an reboot cycle. 

The guru I think is telling me a problem with memory - chip mem?

I have run a couple of different memtests (memcheck and ziptest from aminet) and all show ok but maybe these are giving me false positives as the machine has booted fine prior to these being run.

My plan is to remove the Zip ram but want to avoid pulling all of the chips as they are fragile and I want to avoid bent pins or worse broken pins. Before doing so is there anything else I can check in software to identify a problem with the memory?

Weed

[Castellen]

is there anything else I can check in software to identify a problem with the memory?

Use Amiga Test Kit (boot from floppy disk) and use the 'Test All Memory' function.
https://github.com/keirf/Amiga-Stuff

Let it run overnight and see if there are memory problems reported the next day.

Also try applying gentle pressure to the ZIP memory (both fast memory and chip memory) while the test is in progress and see if that triggers a memory error or a crash of Amiga Test Kit.  If this happens, there's usually a contact problem with one of the ZIP sockets.  This can be tricky to track down exactly where it is, so remove all ZIPs from sockets, apply isopropyl alcohol to the sockets, then carefully re-insert the ZIPs while the alcohol still wet.  Give it half an hour or so for the alcohol to evapourate before powering on.  Don't mix up the chip and fast memory ZIPs as they will be different sizes.

Just to state the obvious, always double check the ZIP polarities before inserting them, check the pins are all aligned, and check pins are *between* the socket contact pairs before pressing them in place.  As opposed to employing a mutant gorilla with a sledge hammer as many people seem to do when fitting ZIP memory.....

[Pat the Cat]

The location shown (08003778) is neither in the chip RAM (00000-1FFFFF) or the Zip RAM (07000000-07FFFFFF).

08000000 and up is the RAM on the CPU card.

https://amigacoding.com/index.php?title=Amiga_memory_map

Zip RAM is  physically fairly robust, it's really fragile to static electricity (use an earthing strap connected to your wrist and and earth point).

I've blown up multiple Zips with just one fleeting touch on a pin while not earthed.

One way to apply pulling press without crushing the chip is a sheet or rubber or masking tape and short nose pliers.

Thumb and forefinger and rubber gloves if you got a lot of finger pressure.

But I think it's the simm RAM on the Cyberstorm Mark 3.

[Tumbleweed]

Thanks both.

I've a CSMKII with 128MB maybe its a dodgy SIMM or one has come loose. Ill check and reseat them later today. The RAM test I ran overnight with no issues. I'll leave the machine off for a few hours and then re-run overnight again. The issues seemed to be when I booted up from cold. Once booted it seems to be much more stable.

Best.

Weed

[Tumbleweed]

Left it off for a few hrs. Switch on - same issue red guru as before, followed by a reboot loop. Switched off. Removed the 5 1/4 drive cage and re-seated the SIMMs on the CSMKII. Switch on and booted off floppy into Amiga Test Kit. Removed floppy rebooted fine off the SCSI2SD.

Will run more tests - maybe an ill-seated SIMM?

Weed

[Tumbleweed]

All looked ok until I refitted the drive cage. I got a recoverable alert when starting iBrowse. Switched off. Removed drive cage and then removed the ram from the CSMKII. Switch on - no boot - black screen.

I then removed all boards - RTG, Network cards and the CSMKII. Switch on. No boot - black screen.

Removed 1MB of chip ram - the other 1 MB is soldered to the board. Switch on - boot and get the 3.2 boot screen.

If its a problem with the 1MB chip ram - that may be down to user error - I suspect I bent a pin when reinstalling it after the board was fixed. There is one ZIP with crumpled leg (its shorter than the others). Ill try straightening the leg, but if I need to replace the ZIPs - can you still get them easily?

Weed

[Tumbleweed]

Straightened the bent leg and refitted the ZIPs. Switch on - black screen. Removed all the chip mem ZIPs again switched on - black screen. I can here the drives clicking - floppy and the SCSI CD Rom is active - its almost as if its stuck in a reset loop. Monitor is flickering coinciding with the polling of the drives.

Caps lock key does not go off after more than a dozen pushes.

What's really odd is that after connecting the cables for the drive LEDs and the key lock, switching on, then off and then removing the cables for the LEDs its booting to the OS3.2 kickstart. Inserted floppy with Amiga Analyzer and can now run the memory tests. CHIP Ram test - 30+ rounds all ok.

With the me tests looking OK. I reinstalled the MKII - no memory installed on it. Switch on blackscreen. Soft reset a couple of times - no rainbow colours to show the CS as active. Then after switching off and a soft reset the CS rainbows appeared. Booted into Amiga Analyzer and now I can run the tests again. Could this be a PSU issue?

Installed the SIMMS in the CSMKII and now its in a continual reboot cycle.

Weed

[Pat the Cat]

Could be Psu, could also be reset circuitry having issue on a cold start.

That Caps Lock light sticking on is odd, usually controlled by the keyboard controller inside the keyboard.

I think you might have more than one problem here. I would try powering on without the keyboard to see how far it gets.

EDIT: Using DeOxit on CPU and RAM connectors might help. The Cyberstorm (040 version) hasn't got much in the way of power smoothing capacitors, very much depends on everything else to behave.

[Tumbleweed]

Pat thanks for that. Th reset circuitry is not something I had thought about. The MB has been recapped and repaired. Was working fine after I rebuilt it until the last couple of weeks when I started to get the recoverable alerts and then the full software failure guru.

The initial symptoms were that it wouldnt boot straightaway - had to do a soft reset. Then I found I had to do more frequent soft resets to get it to boot from cold. The next thing I started to get was the red guru, and now its struggling to even boot - gets into this reset loop which it doesnt get out of.

The CS should be ok as I ran it in my A3000D and its works without any issues.

Edit: Just after posting I wiggled the CSMKII. Switched on and it booted straight into Amiga Anlayzer and everything looks ok. If I switch off and then switch on though Im not confident it will boot.

How can I test the reset circuitry? The issue seems to be when boot from cold, but once booted its mostly ok (maybe the odd recoverable alert though).

Weed

[Tumbleweed]

Could be Psu, could also be reset circuitry having issue on a cold start.

That Caps Lock light sticking on is odd, usually controlled by the keyboard controller inside the keyboard.

I think you might have more than one problem here. I would try powering on without the keyboard to see how far it gets.

EDIT: Using DeOxit on CPU and RAM connectors might help. The Cyberstorm (040 version) hasn't got much in the way of power smoothing capacitors, very much depends on everything else to behave.

I used this contact cleaner to clean the board before I rebuilt it. https://www.toolstation.com/wd-40-specialist-contact-cleaner/p87848 I also soaked the ZIPs in Isopropyl Alcohol before I fitted them. No harm in doing again though. Ill try the DeOxit this time.

EDIT: I ran AmigaMARK - Memory Benchmark and got this:

[Pat the Cat]

How can I test the reset circuitry? The issue seems to be when boot from cold, but once booted its mostly ok (maybe the odd recoverable alert though).

Weed

This is not straightforward on an a3000T. Ideally means removing the motherboard. And tracing the signal with a logic probe. Working from a schematic.

On powerup, the motherboard reset signals waits for a shortwhile, then applies a momentary low pulse into the Rst line. This gets fed around the board to different components.

Chips with power on them start in an "undefined" state, Rst clears them to an initial zero point. The keyboard circuitry for reset feeds ino the same system.

I think you have more than one problem here, so probably a good idea to just reset on power up, then look for the memory fault.

The think with cleaning connectors is - you have to clean both of them, and oxidation builds up on BOTH sides.

Board cleaner does not remove oxidation from pin contacts and solder points. You have 30 years of copper "rust".

OK, this may not fix your memory problem, but it will help to keep a machine reliable.

[Pat the Cat]

EDIT: I ran AmigaMARK - Memory Benchmark and got this:

OK, that definitely points the finger at a chip RAM problem.

If you can get it to do that consistently, at least the first 512kb is OK.

It could be one or more of the 74245's looking after the address lines rather than a defective ZIP.

A little odd that that AmiTestKit didn't find a problem.

Another thought... is the CPU card suppoted by risers or is it just plugged into the slot?  Might be a big part of your issues.

[Tumbleweed]

Thanks Pat. The CSMKII is supported with risers. The board is held in place with two screws.

Checking out the reset circuit is going to be a challenge as I dont have a logic probe, the schematics or knowledge on how to do it. The only chip mem on the board is the ZIPs soldered in place as I didnt refit the ones I removed from the sockets. Think my soldering skills are sufficient to remove / replace these though.

Its very odd though. When the machine starts it runs fine for hrs without any signs of memory problems. I can run the memory diags for hrs without any errors being reported. Checkmem, CheckmemHD, Ziptest, Amiga Analyzer all show no errors.

Edit: Booted fine first thing this morning. Went to change screen mode and got a recoverable alert.

from the error code looks to be:

0x0100000F   Recoverable   Memory header not allocated. Usually an invalid address passed to FreeMem()

So by the looks of it a memory issue or a memory controller issue - could that be a problem with Agnus?.

Thanks for your help.

Weed

[Pat the Cat]

Could be but unlikely - rarely do you get an intermittent fault.

Agnus does control the RAMOE (Ram Output Enable) which feeds to all the Chip RAM components through. Goes to U203, a 16L8D, and comes out on pin 16.

However, if it wasn't working at all, you should get a green screen rather than a recoverable alert,

Likewise the data pins, you have a display, therefore all the data pins are OK.

Which is why I expect it's the 74245s that select the right chip combination from the address. It seems the low end of chip RAM is OK, but there can be issues with the upper part of chip RAM.

https://www.amigawiki.org/dnl/schematics/A3000T.pdf

Has the Psu ever been serviced, recapped or replaced? That could be the root cause.

[Tumbleweed]

Thanks PAT. Ive not serviced the PSU. its where I'll start looking.

I ran it all day today - running the memory diag from ATK v1.18. It generated an error screen shot attached. This doesnt mean an awful lot to me other than it looks to be an error on the 32 bit bus.

Any ideas as to how I should interpret it?

EDIT: And after a short time - chip ram = 0 Other mem = 4,160,555,704 (see screenshot of WB). I soft reboot and now its back into a reset loop after power off / on.

Best.

Weed