Memory problem

[Ericloewe]

IMC? To me, that means Instrument Meteorological Conditions, but I'm pretty sure that's not what you meant by it.

Integrated Memory Controller.

The plot thickens. I removed the DIMM from D1 this morning, and the system wouldn't boot at all.

Weird...

 

[danb35]

Well, the motherboard manual has a chart on page 2-12 of the order in which the DIMM sockets should be populated, though it says that's "for optimal performance". To follow its recommendation and keep D1 vacant, I couldn't have more than 4 DIMMs in the system. For a board with 16 DIMM sockets, that sucks. Might be time for some experimentation...

 

[danb35]

So, if I want to keep the same (or similar) amount of RAM I have now (which honestly is a ridiculous amount of overkill for my needs), and it's actually the case that the board won't work with more than four DIMMS and D1 vacant (which needs testing), I have three options that I can see:

• See if I can find and repair a bent pin in the D1 DIMM socket
• Buy a new $500 motherboard (eBay doesn't show any used), and swap my current CPUs and RAM into it
  
• Spend about $1k on 4 x 32 GB DIMMs

I guess a fourth option would be a different dual Socket 2011 board, swap over the CPUs and RAM, and plug my 9211 into it.

 

[Ericloewe]

See if I can find and repair a bent pin in the D1 DIMM socket

I don't think I've ever seen that happen, I meant the CPU socket. Which is probably an even greater pain in the ass to repair, should it be the case. But it would also affect the other DIMM slot on that channel.

Honestly, I can't think of any realistic problem (assuming the board used to work well) that would cause this to just one of the DIMMs on the D channel. Even the unrealistic ones (traces being flaky to one slot but not the other one) would end up affecting the other slot, due to impedance mismatches.

 

[danb35]

Well, I let this go for a while, because it wasn't happening that often, none of the options to address it sounded very attractive, and I didn't want to have the system up and down while testing. And the options still don't sound too attractive, but I'm tired of the system intermittently falling over on me. So, replacement RAM is on the way, in the form of 4 x 32 GB sticks. Most of the old RAM will be reused, replacing 12 x 4 GB DIMMs in my Proxmox host machine. Crossing my fingers...

 

[danb35]

All right, memory arrived, and installed this morning. Running Memtest on it now. It would ideally run for a week, but it won't; assuming it reveals no errors, I'll probably bring the system back up tonight. Wonder if there will be any effect on system power consumption.

 

[danb35]

Memtest ran with no errors reported (by memtest itself, or in the IPMI logs). Booted back into FreeNAS and it seems to be running fine--crossing fingers that this did the job.

 

[danb35]

Well, a week post-op. IPMI event log is still clean. Average power consumption for this past week is 16W less than for the previous week, though I can't really say that's a result of the RAM. Still looking good.

 

[danb35]

Six weeks later, the event log is still clean, so I guess I can pretty well consider this solved--good thing, too; I'd be quite frustrated to have spent $700 on RAM and not have it fix the problem. Power consumption remains a bit (about 5%) lower as well, which is a nice little bonus.