Not all of us can afford a TESLA Personal Supercomputer; so I have two questions.
- As a developer of a GPGPU code, what is a reasonable fee to run my code on a GPU farm?
- Who can I contact that provides these services?
Amazon offers this for $2.10 per hour per instance.
I can’t see any GPU ressource on the link you provide.
My french company has a little hybrid compute center with tesla C1060 et can sell hours. (I you are interested, i can contact some people to better know how it works and prices)
But, anyway, i am not sure that paying for hours will worth it if you have a little budget.
If you’re not memory bound, you can buy a Geforce which is quite affordable.
Scroll down a bit and you’ll find “Cluster GPU Quadruple Extra Large with the following specifications: […] 2 x NVIDIA Tesla “Fermi†M2050 GPUs”
Yep, sorry, i was automacticly redirected on the french site and it wasn’t referenced in french…
Thats all, and especially Tera … I had no idea Amazon offered this service!
I’m representing a number of people who all wish to run GPU code and have a variety of memory needs. The codes vary from MD simulation, to large fluid dynamical problems.
Some of them are likely to be memory bound so the Tesla cores are a requirement. Obviously we all need to be running code at independent times and fast turn around of our results are critical, so access to the system should be easy and independent (e.g. multiple accounts). I still have to run some figures to determine if its advantageous to go out an price our own server, or whether a GPU farm would be more cost effective. If my group dont maintain sufficient utilisation of a private system, then it will be worth looking at the farm option.
FCS, I wonder if you could get me an indicitave price for your cluster?
At the moment I’m looking into the following purchase;
Poweredge-c6100 - Server rack
Poweredge-c410x - Expansion Slots, filled with Nvidia-Tesla-M2050