To TESLA or not to TESLA

[cement_head]

Hello,

I'm in the processing of spec'ing a new bioinformatics workstation for a DNA Genomics Core (academic users) and I have a question on the use of TESLA's for bioinformatics.

Most of our user base is non-technical and will not be coding/compiling software. They are end-users of produced software, be it FOSS or commercial (e.g. CLC Genomics Workbench, Newbler).

Is it worth buying a TESLA equipped machine for bioinformatics, or not? We are planning on installing BioLinux 6 (7 when it is ready). If I have interpreted the state-of-affairs right now, GPU enabled software for bioinformatics is not widespread - but is becoming available.

A second question is: Is it difficult to make software use a GPU/TESLA option? Does the CUDA development kit make this easy, or it actually quite labour/coding intensive at the moment?

Any thoughts would be welcome.

Regards,
Andor

[ymc]

For academic beginners, I would recommend trying a GeForce card first, for example, a GTX570, because a Tesla equivalent is about 3x to 5x the price. The development kit is easy to use as long as you know the basics of C and its related development tools in Linux.

When you find the GTX570 is too slow for double precision math or you think it is not stable enough or you need more VRAM, then move on to the Tesla. For now, it would make sense to wait for the Tesla K20 that comes out end of this year.

[GenoMax]

If you are buying this machine mainly for end-users then consider omitting high end GPU and using that money to get additional RAM/SSD's. You would get more out of the machine for normal use that way.

You can always buy a GPU later and stick it in the machine if you determine that you must have one (or more) down the road.

[arvid]

I'd agree with GenoMax, if you don't know for sure now that a major software you will be running can use the GPU, use the money for more RAM, cores, SSDs or storage instead.

[adaptivegenome]

We are using at GTX480 for development. It is not pricey but we can still develop and run CUDA applications.