Hello all!
I am potentially getting to build a replacement for a slightly older number crunching machine.
The stats for the build thus far is:
2 x E5-2699 V4 (22 cores each)
2 x NH-U12S coolers
Asus - Z10PE-D16 WS SSI EEB Dual-CPU LGA2011-3 Motherboard
2 x 16GB unbuffered RAM DDR4 2400
960 Pro M.2 1TB (I/O speeds matter, lots of reading data)
small 1050 GTX GPU (could go smaller, mostly for MATLAB graphics and Google Earth)
I'll worry about the PSU and other stuff later.
The issue I'm having is finding a Case that can handle the cooling effectively enough. The current machine has 2 x E5-2680 V2 in a Thor V2 with Evo coolers and has been fine (been running 3 years non stop no issues). I just want to be sure there is sufficient cooling in the case as well as not having to worry about the MB warping. I'm looking to spend less than $200 if possible for the case, and have a handful of leftover 120mm fans from a older build that would be fine (silence and appearance is a secondary concern over fitting and cooling).
As for the "why", we do computations that are heavily parallelized and the more processors we have, the faster it gets done (needs to call CPU though). It is uncertain if CUDA will be of any benefit given the nature of the calculations. The jobs previously would have taken years to complete on a single core, and just going to the 40 threads makes those jobs take weeks. We're looking to cut that down to days/hours with the 88 threads. There is effectively a blank check for the moment, but I'm not looking to do something nuts like the $40k E7 builds. I'd say that $8k is the absolute limit, but major game changers are welcome.
I built the previous machine and have completed several builds at this point that all have been successful, so I'm willing to do stuff that is a bit "unusual" if it means speeding these calculations (ie, Phi processing which I know nothing about and can't seem to find answers on anywhere).
A case suggestion would be appreciated, along with any newfangled tech to potentially get more bang for the buck. Thanks!
I am potentially getting to build a replacement for a slightly older number crunching machine.
The stats for the build thus far is:
2 x E5-2699 V4 (22 cores each)
2 x NH-U12S coolers
Asus - Z10PE-D16 WS SSI EEB Dual-CPU LGA2011-3 Motherboard
2 x 16GB unbuffered RAM DDR4 2400
960 Pro M.2 1TB (I/O speeds matter, lots of reading data)
small 1050 GTX GPU (could go smaller, mostly for MATLAB graphics and Google Earth)
I'll worry about the PSU and other stuff later.
The issue I'm having is finding a Case that can handle the cooling effectively enough. The current machine has 2 x E5-2680 V2 in a Thor V2 with Evo coolers and has been fine (been running 3 years non stop no issues). I just want to be sure there is sufficient cooling in the case as well as not having to worry about the MB warping. I'm looking to spend less than $200 if possible for the case, and have a handful of leftover 120mm fans from a older build that would be fine (silence and appearance is a secondary concern over fitting and cooling).
As for the "why", we do computations that are heavily parallelized and the more processors we have, the faster it gets done (needs to call CPU though). It is uncertain if CUDA will be of any benefit given the nature of the calculations. The jobs previously would have taken years to complete on a single core, and just going to the 40 threads makes those jobs take weeks. We're looking to cut that down to days/hours with the 88 threads. There is effectively a blank check for the moment, but I'm not looking to do something nuts like the $40k E7 builds. I'd say that $8k is the absolute limit, but major game changers are welcome.
I built the previous machine and have completed several builds at this point that all have been successful, so I'm willing to do stuff that is a bit "unusual" if it means speeding these calculations (ie, Phi processing which I know nothing about and can't seem to find answers on anywhere).
A case suggestion would be appreciated, along with any newfangled tech to potentially get more bang for the buck. Thanks!