Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

I am trying to build a PC capable of running in the best case both GPGPU-sim and BARRA sim. And if impossible to match I am going to stick with GPGPU-sim.

Below is a list of software requirements and specifications for both of them:

Barra: Tested platforms:

Ubuntu 8.04 x86_64 | Ubuntu 8.10 i386 | Debian Lenny x86_64 |

CUDA 2.0, 2.1, 2.2 and 2.3 are supported.


GPGPU

SUSE 11.3 | Ubuntu 10.04 LTS (But not tested).

CUDA 2.3,3.1


Conclusions:

Both require completely different OSes and CUDA versions.


I am not so good at hardware assembly and I need to find the right hardware for this. Please help.

I need to know the optimum processor, Motherboard and GPU requirement. (and any other parts if specially needed)


In hardware I need a CUDA GPU from the list here: https://developer.nvidia.com/cuda-gpus

EDIT: The list of support GPUs from wikipedia I found here http://en.wikipedia.org/wiki/CUDA#Supported_GPUs . But I am not so sure its complete list.

share|improve this question

migrated from stackoverflow.com Nov 26 '12 at 18:12

This question came from our site for professional and enthusiast programmers.

    
I have run GPGPU-sim successfully on RHEL 5.5 and CUDA 4.0 (and no GPU). I realize that doesn't help you much, except my point is that the requirements you list might not be hard requirements, but in some cases merely suggestions of what it was tested with. Also, I found that GPGPU-sim required careful study and experimentation to get it to run reasonably well. Even then it did not tolerate all codes I threw at it. Just advising you to have patience if you want to use it. You will need patience as well because these simulators run much slower than an actual GPU, but that's to be expected. –  Robert Crovella Nov 26 '12 at 16:07
    
Also, it's not difficult to have multiple versions of CUDA loaded on a machine simultaneously and switch between them with some $PATH and $LD_LIBRARY_PATH manipulation, or switch by using symbolic links. For example, put cuda 2.3 in /usr/local/cuda23 and cuda 4.0 in /usr/local/cuda40 and use a symbolic link /usr/local/cuda -> /usr/local/cuda40 for example. –  Robert Crovella Nov 26 '12 at 16:55

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.