• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Single 64-bit processor vs. 2 32-bit processors: Speed comparison under linux

O

olodum

Junior Member

I am running computationally intense simulations in fortran under linux-based environment (details below). Can anyone tell me how important it is to use a 64-bit processor as opposed to 32-bit? I am mainly interested in shortening simulation time. Also, how does performance on Athlon 64's, Opteron's and Intel Xeon / Itanium compare? Moreover, within a given category, what is the gain from each upgraded clock speed?

Simulation Details:
Nonlinear Partial Differential Equation, propagated forward in time stochastically via Cranck-Nicholson, or similar numerical scheme. Accuracy is important as correlations are investigated. Program written in fortran 77, cross-compiled with random number generator in C.

 
Get/build a system with an AMD X2 3800+. You'll have 64-bit and two CPUs, and won't break the bank.

BTW, you'll need a full 64-bit Linux distribution to take advantage of the architecture. Gentoo is working nicely for me.

-SUO
 
You will only benefit from using a 64-bit cpu IF your OS is 64-bit AND your program has been written and compiled with a 64-bit compiler. I'm willing to bet that since you are doing some pretty heavy math operations, you will get an awesome performance boost if you can take advantage of the 64-bit capabilities.

For performance comparisons check out some of the articles on the main page here

http://www.anandtech.com/cpuchipsets/
 
Since it's fortran 77 and he probably has access to the source code it shouldn't be a problem to get native 64bit versions. I figure he's using gcc's fortran stuff.

I don't know if 64bit would have much to do with it.. I don't know much about fortran or anything like that much at all.

But when your comparing dual x86 cpus to single amd64 cpu it's very difficult to eye ball it. It depends on a lot of stuff.

Is the code able to be run in parrellel? Can it be multithreaded, or at least split up so that it can have different parts run at the same time?

(if it can be split up into different jobs it's a good canadate for some HPC clustering, btw. A classroom full of modern 2.8ghz dell computers operating overnight can do some crazy amount of number crunching)

If it can only be run serially then dual cpu has little advantage over a single cpu.

Also different arches have different capabilities, different things they are very good at. In specific benchmarks Xeons/Pentium 4's are going to be much faster, were in others Opterons are going to be much faster.

These are the latest results of some Fortran 77-based (I beleive) benchmarks, the Spec 2000. They record speed and thorughput in int and floating point calculations.
http://www.spec.org/cpu2000/results/res2005q3/

Here is the description of the floating point portion of the test..
http://www.spec.org/osg/cpu2000/CFP2000/

Keep in mind that these are mostly ideal results put out by manufacturers using very optimized compilers and such.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top