Here's a strange one, maybe someone here has experience with something like it...
I just bought a "gently used" T7610 to use as a proper workstation after having a good experience using a cheap refurbished T3500 as a home server for the past year. After getting a fresh install of 64-bit Win10 up and running on it, I noticed it was only reporting 8GB of RAM (it should have had 16GB).
A quick trip into the BIOS showed 4 active channels, but the DIMMs in slots 3&4 were being reported as "
0 GB ECC RDIMM" (as opposed to
*** DIMM slot empty *** or some other more obvious error)... and swapping the DIMMs around between slots 1&2 and 3&4 gave the same results, so it didn't seem to be an issue with the RAM itself.
On a lark I tried a dual-channel setup, populating only slots 1&2 for each CPU... which resulted in all 16GB being recognized.
I had planned to upgrade to 32GB, so I already had on hand an identical set of 16GB 4x4GB Micron 1866MHz RDIMMs, same part number as was already installed in the machine... and populating the RAM according to the recommended install chart for quad-channel (4GB DIMMs in slots 1,2,3,&4 on both CPUs, (the ones with the white release-levers)) results in a similar issue--DIMMs 3&4 on
both CPU1 and CPU2 are reported as "0 GB" in the BIOS, showing up as 16GB in both Windows and the built-in hardware test. Even weirder: going against the recommended install chart and just putting all of the RAM in CPU1's slots also shows 16GB, but with slots 1,2,5&6 reporting "4 GB" and slots 3,4,7&8 reporting "0 GB"
Random things I've tried:
- Updated to the latest BIOS (it arrived running the BIOS it shipped with in 2014)
- Turning off RMT
- Allowing Memory Map IO above 4GB
- Booting with just about every combination of available DIMMs in only slots 1&2 just to make sure they're all working
...but nothing seems to make a difference, or point me in the right direction for a solution.
Other possible related issues:
- It came set up in a dual-processor configuration with two E5-2630 v2's, and 16GB of PC3-14900R in a 4x4GB configuration. The service tag implies that the original config was single-socket, and someone added a second 2630v2 at some point, but the machine is otherwise configured to match its service tag.
- The machine seems to have taken a spill during shipping; the front bottom corner near the SAS cage slightly crumpled and bent out of true, and some rivets had popped out of the case frame and were rattling around inside. I had hoped that this was just cosmetic, but could it have caused cracked traces on the motherboard or shorted something..?
- I've had a few "WHEA uncorrectable error" BSODs, including one from the install environment while first trying to install Windows (!). Probably not a great sign on a fresh install with everything stock and no additional software running.
I've wasted about 5 hours on this already, and am about to give up and return the thing.
Any ideas before I throw in the towel? This is my first dual-socket LGA2011 machine, so I'm not 100% confident I'm not screwing something up myself here, but I can't for the life of me figure it out. Physical/electrical issue with motherboard or DIMM sockets? (seems weird that it'd be the same slots on both CPUs)... Memory controller issue in the Xeon(s)? Or just a time-suck mystery not worth solving?