ECC or not? AMD with ECC vs intel without, homserver, consumer parts, zfs.

Nah, seems like it won't fit in my old case, which is a pity, as the Fractal Design case I have is pretty good noise-wise, and I could live well without hot-swap disks where I have to pull out the front panel anyway.

Still, the fan is not LOUD on low RPMS, but my God, when it spins up, it's a vacuum cleaner! Felt like sitting beside an AC in high mode. Luckily, it only does that if the side panel is removed (for some reason). Hopefully it won't start to spin up even when I use it and it gets some load. I'm considering swapping out the CPU cooler anyhow, if I can fit a Noctua NH-U12 or similar to it somehow, Then I'll probably swap out the case cooler as well, or even remove it, have the CPU cooler blow out of the case, and have a separate fan on the HDDs.

Spent hours today getting the bloody thing updated, BIOS, iDRAC and whatnot, seeing as it hadnæt been done in ages, and so the built in update system wasn't working, and required a couple of pre-update-updates. Certainly not an ordinary desktop, this!
 
von_Gaden said:
With all my respect to reliability of ECC I'd note that if the error rate of non-ECC RAM is correct we should disbelieve and discard any data provided by devices with non-ECC RAM...
One thing to note in the quoted references is the age of the articles. For example, djb says "a typical computer with 256MB".

I remember many claims that we were reaching a limit to memory capacity (due to shrinking silicon feature size). The one that generated the largest hooplah was the often-repeated "any decrease in feature size will not work because the ceramic IC packages we use will emit so many alpha particles that the memory will be useless".

I've used just about every memory technology there is, from Williams tube (but that system was an antique at the time), mercury delay line (likewise), drum, core, Twistor, static and dynamic semiconductor RAM.

All of the server-class systems I specify have ECC memory, and log correctable / uncorrectable errors. As far as I know*, with over 300GB of ECC memory scattered among a dozen or so systems, I've only had one correctable error in the last year.

* Famous quote: "As far as we know, our computer has never had an undetected error."
 
Damn it. Seems like it's the X3430 processor inside, which does NOT have the AES instruction set. That was really quite a disappointment... How much toll would GELI take on this 2,4Ghz Xeon?
Seems like I can't find a Xeon with AES-instructions for the 1156 socket at all. Don't wanna give it up yet, might have to test it and see how it goes.
 
Use ECC RAM, you then get some cryptic messages from time to time. (if you're unlucky)

Code:
[CMD="#"]cat MCA.txt[/CMD]
MCA: Bank 4, Status 0x9422400121080813
MCA: Global Cap 0x0000000000000105, Status 0x0000000000000000
MCA: Vendor "AuthenticAMD", ID 0x10ff0, APIC ID 0
MCA: CPU 0 COR BUSLG Source RD Memory
MCA: Address 0x5f311e70
[CMD="#"]mcelog < MCA.txt[/CMD]
HARDWARE ERROR. This is *NOT* a software problem!
Please contact your hardware vendor
MCE 0
CPU 0 4 northbridge TSC 108d22803cdc 
ADDR 97f350 
  Northbridge RAM Chipkill ECC error
  Chipkill ECC syndrome = 2144
       bit32 = err cpu0
       bit46 = corrected ecc error
  bus error 'local node origin, request didn't time out
             generic read mem transaction
             memory access, level generic'
STATUS 9422400121080813 MCGSTATUS 0
MCGCAP 105 APICID 0 SOCKETID 0 
CPUID Vendor AMD Family 15 Model 31

Still waiting for the first laptop with ECC RAM to come.
 
Back
Top