Heres a great one for you guys..
I have TWO M520s configured identically that exhibit this problem. Approximately every 1-2 weeks, they will slow down to about 5% of their normal speed -- as measured with some cpu benchmarking tools.
I've killed the network connection to them to limit any outside interference.. and rebooted (warm boot). The problem still exists. I've even tried to warm boot and boot using PXE (to eliminate the OS completely from the problem). I run memtest86+ v5.0.1 and you can see the speed problem while doing a memory test.. horribly slow.
The only thing that will "fix" the issue is if I cold boot the server (power switch or via the drac works fine). Then the servers are good to go for another 1-2 weeks before randomly slowing down again.
The hardware is a M520, single E5-2420, dimms in DIMMA_2 & 3 (16gb 1333mhz dual-rank), H710 raid, and no mezz cards. BIOS is 2.1.4.
Things I've tried: switching from memory-mirroring to multi-bit ECC mode, upgrading RAID card (was H310) and fw, upgrading NIC firmware, upgrading iDRAC fw (now on 1.57.57), upgrading LC fw.
I don't believe I had this issue when I was on v1 of the BIOS. I believe it is a firmware bug involving CPU throttling.