How to choose the cpu for simulation

sunti · January 26, 2012, 21:45

Hello all,

I have to choose the cpu for simulation.
By the setting cluster with 128 core, can anyone suggest me for choosing the cpu Intel or AMD would be match this configuration.

Now I looking for Intel Xeon x5690 and AMD Opteron ......
but the AMD clock is less than Intel.
I'm not sure it have the effect on the speed of the calculation or not?

Thank for all suggestions,

Sunti

rmh26 · January 30, 2012, 12:43

128 cores /6 cores/x5690 = roughly 22 processes at ~ $1600 = $35,000 for processors alone. How much money you budgeting for this project?

I would consider buying E3-1270 (sandy bridge) instead.
http://ark.intel.com/products/52276

Can't run them in a dual-proc setup but a large savings and you will get the AVX which doubles the SIMD regsiter from 128 to 256kb. This means each core can do 4 SIMD DP ops per cycle where as the each x5690 can only do 2 SIMD DP ops per cycle.

The Opterons (Interlagos) are cheaper but sort of misleading. An 8 core Interlagos has 8 integer units but only 4 AVX SIMD units. However they do support FMA which means a = a +b*c can be done in one operation. This provides a great speed up for many linear algebra operations, dot products, matrix vector mults. If your writing you own software and have good compilers you might get better performance from the opterons by using the FMA.

Sandy Bridge will save you money on power and the reviews for the first generation bulldoze(Interlagos) haven't been so great.

my vote: new intel sandy bridge xeon's (look for models without built in graphics unless you want to try going the opencl route)

January 26, 2012, 21:45	How to choose the cpu for simulation	#1
sunti New Member sunti Join Date: Nov 2011 Posts: 6 Rep Power: 14	Hello all, I have to choose the cpu for simulation. By the setting cluster with 128 core, can anyone suggest me for choosing the cpu Intel or AMD would be match this configuration. Now I looking for Intel Xeon x5690 and AMD Opteron ...... but the AMD clock is less than Intel. I'm not sure it have the effect on the speed of the calculation or not? Thank for all suggestions, Sunti

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
stop when I run in parallel	Nolwenn	OpenFOAM	36	March 21, 2021 04:56
Superlinear speedup in OpenFOAM 13	msrinath80	OpenFOAM Running, Solving & CFD	18	March 3, 2015 05:36
OpenFOAM 13 Intel quadcore parallel results	msrinath80	OpenFOAM Running, Solving & CFD	13	February 5, 2008 05:26
OpenFOAM 13 AMD quadcore parallel results	msrinath80	OpenFOAM Running, Solving & CFD	1	November 10, 2007 23:23
Dual Core CPU	hjasak	OpenFOAM Running, Solving & CFD	5	July 22, 2006 03:57

January 30, 2012, 12:43		#2
rmh26 Member Join Date: Jul 2011 Posts: 59 Rep Power: 14	128 cores /6 cores/x5690 = roughly 22 processes at ~ $1600 = $35,000 for processors alone. How much money you budgeting for this project? I would consider buying E3-1270 (sandy bridge) instead. http://ark.intel.com/products/52276 Can't run them in a dual-proc setup but a large savings and you will get the AVX which doubles the SIMD regsiter from 128 to 256kb. This means each core can do 4 SIMD DP ops per cycle where as the each x5690 can only do 2 SIMD DP ops per cycle. The Opterons (Interlagos) are cheaper but sort of misleading. An 8 core Interlagos has 8 integer units but only 4 AVX SIMD units. However they do support FMA which means a = a +b*c can be done in one operation. This provides a great speed up for many linear algebra operations, dot products, matrix vector mults. If your writing you own software and have good compilers you might get better performance from the opterons by using the FMA. Sandy Bridge will save you money on power and the reviews for the first generation bulldoze(Interlagos) haven't been so great. my vote: new intel sandy bridge xeon's (look for models without built in graphics unless you want to try going the opencl route)