ECC errors

[DeadSirius]

I bought a pair of 2 GB FB-DIMMs for my Mac Pro, and tested them immediately. I ran Rember on it all night for two nights, and Apple Hardware Test on it twice. Neither app reported any problems. It's been almost two weeks, and they seem to work great. No apparant hiccups, though I haven't noticed much speed difference, other than less drive accessing.

Anyway, tonight I glanced into System Profiler, and the status on one of them was "ECC errors". So I'm wondering, how bad is this? Should I return it? (I didn't realize System Profiler had a status indicator until tonight, so I have no idea how long it's been like that.)

 

[Monyx]

I found the same thing with 1GB DIMM in DIMM Riser B, DIMM 2 slot. I re-ran a newly purchased Memtest 4.14 specifically for MacPros...5 interations over night still reported no problems, not sure what I should do either as returning DIMMS is a hassle from Australia --> US.

 

[orangezorki]

I think if you click on it in ASP, it will show you how many errors have cropped up. Of course, you could say that as it is correcting the errors OK, everything is fine, but I'd send them back if I were you.

David

 

[DeadSirius]

I rebooted and did a 90-minute, render-intensive Final Cut sequence to Compressor, and Profiler showed over 9800 errors. I read that someone else re-seated their RAM and fixed their problem, so I'll try that first. But otherwise, it looks like I'll have to return it.

 

[Monyx]

I contacted the author or MemTest, he said Memtest is not aware of ECC RAM so just reports ok if what it reads from RAM matches what it writes. So, the Apple profiler report is worrying and I have arranged to return my RAM. After a reset last night I noticed my 2x1GB DIMMS are now reported as 2x512MB DIMMS, so I've lost 1GB of RAM - strange. 2GB are back after a shut-down.

Despite my ECC errors numbering ~212800, machine is stable no crashes, my compressor renders on short projects are okay - but certainly not acceptable for RAM a few weeks old.

EDIT: copy in the reply from MemTest author,

The whole point of ECC memory is to correct errors so they don't cause problems. Memtest is not aware of ECC memory being present - it only knows whether the data returned from memory matches the data written to memory. As long as ECC memory is working correctly, memtest will not report the errors because it didn't see any. If you look in your system.log file, I believe you'll find the ECC warnings generated during the memtest run. MacOS keeps track of these errors itself.

From a functionality point of view, no memtest errors means your system is fine. The fact that Apple profiler is reporting ECC errors is worrisome from a reliability viewpoint though. There were a large number of errors corrected but they were corrected so the system doesn't fail in any way. However, I'd question any DIMMs that reported ECC errors right out of the box. I've always looked at ECC memory as a fail-safe well down the road as the system ages. Memory chips become less reliable over long periods of time - ECC memory gives you peace of mind that when individual bits start failing, your system will continue to run. Essentially, you get an early warning of a potential future failure. I certainly wouldn't accept brand new DIMMs exhibiting ECC errors and I'd definitely wonder about the quality of those DIMMs. As it stands, you're already getting a warning of a potential future failure so now's the time to act.