Quote:
As genixia pointed out, from

Checking for extra DRAM:
c0c00000: wrote ffffffff, read ef1ba9f0

it seems it is not properly recognizing the 12MB-16MB block. I'm thinking there are a few possibilities:

- The installed memory was bad to begin with. Hugo, who sent Stu the chips to install, mentioned somewhere that they had a high fail rate. I'm not sure if they were tested pre-install.

- Something might have gone wrong with the install. It was the first (only?) Mk2 upgrade that Stu had done. He posted a pic here, and more details are scattered about that lengthy thread.

- I don't remember reading about any other Mk2 memory upgrade successes, so maybe there is something else that still needs to be worked out for Mk2s specifically in Hijack. I know it's a long shot, but it's what I'm hoping for given the alternative of being out my player for a week or two with the shipping turnaround to Stu.

If anyone has any other ideas, I'd appreciate it...


No, you've pretty much covered the bases.

At first glance, the error seems to indicate a problem with the RAS line, although as Mark has noted, the photo looks correct. If the RAS line came loose at the CPU then I'd expect to see this kind of error. Basically neither chip appears to be answering the CPU. It would be unusual for that to be caused by other soldering issues as it would require identical soldering defects on each chip.

The high memory failure rate does concern me. ISTR Hugo quoting something like 1 in 4 failure rate. That would give a 1/16 chance of a Mark1 upgrade having both chips be bad, and a 1/2 chance that one of them is bad. (And for a Mk1 upgrade, a 1/1 chance that one is bad). I hope that Hugo was exaggerating! Of course, it is unlikely that a bad chip would catastrophically fail (which would be required for this failure mode), so it probably isn't this.

(As an aside, this causes concern for Mk1/Mk2 upgrades in general.)

And until we see the first successful Mk2 upgrade reported, we cannot totally rule out software either. However, it does look like it's doing the right thing at this stage. If it was detecting the RAM fine and then falling over, or failing at some nonsensical address then it would be suspect, but c0c00000 is exactly right for the other failure modes...


Edited by genixia (14/09/2004 01:23)
_________________________
Mk2a 60GB Blue. Serial 030102962 sig.mp3: File Format not Valid.