Memory and motherboard compatible?

macmanes

Distinguished
Nov 15, 2009
8
0
18,510
I am building a workstation for scientific computing (mostly next-generation sequence analyses), and would like some advice/opinoins about RAM-MOBO compatibility. A lot of research has gone into the configuration (including reading the FAQ), but I have little experience in building computers, so hope that you guys can help me avoid problems. Thanks in advance!

-MOBO: ASUS Z8PE-D18(ASMB4-IKVM) Dual LGA 1366 Intel 5520 SSI EEB 3.61 Dual Intel Xeon 5500 series Server Motherboard .
■DDR3 800/1066/1333
■ECC or Non-ECC
■Reg DIMM or Unbuffered

-RAM: Patriot Signature 12GB (3 x 4GB) 240-Pin DDR3 SDRAM ECC Registered DDR3 1333 (PC3 10600) Server Memory Model PS312G13ER3K-E Will end up with 24GB per socket of RAM.
■Triple channel SDRAM DDR3 1333
■ECC, Registered
■2 Rank Double-sided module


Is this combo going to play nice together?

Other parts, just FYI..

-PS: ENERMAX REVOLUTION85+ ERV1050EWT 1050W ATX12V / EPS12V this has 2 +12V commecitons (8pin and 4+4pin) to power the 2CPU's. Is this enough power?

-CPU: Intel Xeon E5520 Nehalem 2.26GHz 4 x 256KB L2 Cache 8MB L3 Cache LGA 1366 80W Quad-Core Server Processor Will get 2 of these..
 

macmanes

Distinguished
Nov 15, 2009
8
0
18,510
Actually ZIP, I'll fill the dimms with 4GB ram- would go with the 8gb sticks, but those guys are still very expensive.

I am using the computer for analysis of next generation sequence data.. In short, we are talking about the analysis of 50 million (give or take) 100 base pair reads.. Do the math, but 50,000,000 * 100 letters each is a sh!t load of data, which is represented in a single deBrujin graph. Here is the link, in case your not sure what that is (http://en.wikipedia.org/wiki/De_Bruijn_graph)..

Unfortunately, most of these tasks are non parallizable, and so the entire data set is loaded into RAM.. Even with the amount of RAM I am currently able to stick in this machine, I won't come close to being able to load the complete dataset at once, and combining multiple runs is, problematic at best..

Also, FYI, I run larger analyses on a supercomputer at TACC (http://www.tacc.utexas.edu/resources/hpc/#constellation). Not the fastest out there, but this 63,000 node computer runs with 4GB per node, and has some nodes coming online soon with over 100GB RAM..

Lastly, as you may have seen, Amazon EC2 now has High-Memory Quadruple Extra Large Instance using 68.4 GB of memory (http://aws.amazon.com/ec2/#instance) for some relatively common RAM intensive applications..

Obscene RAM usage is not all that uncommon in scientific computing,,