Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

After setting up a new compute server for my research group I need to evaluate the overall performance of this machine, including both Tesla cards. I found some information about a CUDA enabled version of Linpack and how it is used, but no download links for the software.

So the problem is, where to get it from?

share|improve this question

2 Answers 2

up vote 1 down vote accepted

The real CUDA-enabled HPL benchmark, which is used for the TOP500 list too. Is available direcly from NVidia after registration. It is only accessible for members of the "CUDA Registered Developer Program".

share|improve this answer

We really try to not show any "RTFM" or "LMGTFY" attitude here, but honestly - the first hit for "hpl cuda" on google is exactly what you're looking for.


share|improve this answer
"This version of the code only acclerates one section of the factorization for a single GPU." As I read the documents on the net, I think there is a more advanced version, which is more actively developed. Because this version from git hub crashes immediately. –  Grisu Nov 20 '12 at 9:37
"The scheduling routine gpuUpdatePlanCreate() in auxil/HPL_gpusupport.c contains two tuning constants tune0 and tune1, which control the split of work between the GPU and host CPU(s). These must be tuned for optimal performance with a given GPU and host CPU/BLAS combination. How this is done is left as an exercise for the reader." –  pauska Nov 20 '12 at 9:38

Your Answer


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

Not the answer you're looking for? Browse other questions tagged or ask your own question.