Uh Oh

My troubles started the other night after success with the CSound pvoc opcode. I installed Audacity on my Linux AOpen XCcube (UK79G-1394 motherboard. 2 Gig’s of DRAM) with the hopes of piecing together some new material. I plugged in my $7 microphone and tried to record. Nothing.

What followed, followed the time intensive anti-pattern I’m calling “Chain of Despair.” It goes like this: In order to trouble shoot problem A, you need to run an experiment using B but, B doesn’t quite work and you have to figure out why. Guess what, you’ll need C to figure that out and, you guessed it, C is just a few ‘to do’ items away from being in your tool box. This is often further complicated when the technician notices other items not functioning perfectly and starts trying to solve those other problems instead of focusing on the original issue.

Its always darkest just before it goes pitch black

For me in my most recent case of Chain of Despair, the system at one point wouldn’t boot. (I wish I could remember the error). This had happened months earlier and I eventually RMA’d one of the 2 1Gb RAM sticks. Along with the booting trouble, I had some defective areas in the display (see below.) These would show in both graphics and text modes to varying degrees of severity. But when that one stick was in the mail (and not in the machine), I had no boot problems and none of these display defects. When the stick was replaced with a new one from the manufacturer, the boot problem went away (until now) but the display problem was still there.)

So when the machine wouldn’t boot the other day, I knew what to do. I yanked one of the sticks back out and began a flurry of booting in to Knoppix, swapping RAM sticks and, running memtest.

Here’s a picture of Knoppix failing to boot that also shows the funky display pattern I’ve been talking about.Knoppix not booting while display is defective

See those three rows of sixteen ‘sprites’ at the top of the screen. Those are not reflections. They would appear when ever both 1 Gig sticks are installed in the system. Did I mention the video on this thing is a shared memory architecture? I tried playing with those BIOS settings too but, I couldn’t fix these problems.

While memtest was running I came across this from the memtest86 home page: “Don’t buy cheap memory and expect it to work with an Athlon!”

Oh well. I’m not sure if we are supposed to consider Corsair Value Select Memory as “cheap”

The conclusion from trying every stick/socket combination was that, the memory is probably fine. When I only put one stick in, I get zero memtest errors. Doesn’t matter which stick or which socket. But as soon as I put in both sticks, memtest generates over 22,000 errors.

So, for whatever reason, I decided to update the bios. Maybe it was because diddling with the RAM timing settings only made things worse. Maybe because I read something on the internet about improved memory compatibility. At any rate, this was no minor undertaking since, it is Linux and the bios-flasher utilites are mostly packaged in Windows friendly packages and launchers or, require a floppy drive. The machine in question has no floppies. So it took me quite some time to burn a bootable El Torrito CD that I could use to run a DOS based BIOS loader. Well that’s what really did it. None of several attempts suceeded. After one such failure, I gave the DOS that was running the three finger salute to reboot in to SHIFT-F8-CONFIG.SYS-AUTOEXEC-A-STEP-AT-A-TIME mode and not load HIMEM.SYS (the loader had complained about that) but, the monitor just stayed black.

At this point, I’m happy to say, I didn’t lose it.

So after quickly analyzing my assets and liabilities, I’ve spent the last few hours searching for replacement hardware. I’m assuming that those 2 Gig of DDR memory, my Athlon, and the hard disk are all still good and its only the BIOSthat’s wasted. So the plan is going to be to scrap the “barebones” part of the machine and replace it with another kit that will accept the Athlon and the DDR. The AOpen XCcube EZ18 cost me about a buck and a quarter 14 months ago, far less than the cost of the memory and the processor. I want to have firewire IO that I have come to enjoy with this box (since I earlier figured out how to capture DV from my Panasonic PV-GS500 to this Fedora machine). Another requirement for the replacement is a nForce chipset. In a perfect world, it would be an ASUS, small form factor that would fit in the original case. Then I considered buying a new case separately but, just finding such an ASUS board, in any form factor, proved to be just too much to ask. So, I settled on a Jetway MiniQ 760S barebones.

Now, I just have to wait for it to get here, put Humpty back together and hopefully have a happy ending here in a week or so.

Fortunately, almost all my data is backed up to a Maxtor Personal Storage 3200 external USB drive. But, none of the Windows machines around here can read it (yet) because I formatted the Maxtor ext2. There’s probably a smarter strategy but, just formatting the 200gig thing in a Linux friendly way seemed easier than trying to get Linux to write to NTFS. If I really need any of that data in a hurry, the plan has alway been to install the fs-driver on the win2k or XP machine