Friday, October 10, 2008

HPC in a box

Has your business been avoiding HPC because you lack the expert knowledge to install and maintain clusters? Have you been itching to run simulations and gain some competitive edge, but the world of job schedulers, mpi and other "standard" HPC methods was a new world of pain and confusion? A new solution may be on your horizon. this week Velocity Micro announced a new line of Visual Supercomputer workstations. With the power of a small supercomputer in a single box, preloaded with NVIDIA hardware and software, this is a turnkey solution without clustering, grids or other complex HPC elements. With pricing that looks to range from 1500$ to 6000$ and compute power that ranges up above 3 teraflops, this little boxes will suit many small to mid size business needs without breaking the bank. I am thinking of putting one on my Christmas list, so if you get one and have a review or comments, please speak up here.


Anonymous said...

Isn't that a GPU based solution? Would a company have to have vectorized code in order to take advantage of a Telsa-based system? Do you know what applications will be available to run on it? Applications pre-designed to run on the Tesla would be required in order for this to be a true turnkey solution.

Unknown said...

This is in fact a GPU based solution that uses the TESLA cards. However, NVIDIA has released CUDA which is a free compiler that will take any C Code and compile it to run on the GPU instead of a CPU. Any code that is uses highly parallel operations would potentially benefit from a GPU- based computing solution. You may also see such computer architectures referred to as GPGPU ( General Purpose ( computation on) GPUs). There is a good article on the topic and on NVIDIA's TESLA cards over at the Hardware Secrets site.
As the popularity and deployment of this hardware grows, I think it will become common to see major simulation and CAD/CAE software companies releasing version meant to run on the Tesla. GPU programming is growing by leaps and bounds- even the Folding at home freeware software has an ATI GPU version- and being able to program in a C code environment and then compile to Tesla makes it even easier. I still maintain that this product from Vector Micro is potentially a game changer for the industry and the way HPC is used in mid sized manufacturing shops. Expect to see other companies getting on board and offering similar hardware combinations in the near future.

Anonymous said...

Yes, this is correct. Application acceleration would require that software be modified or use libraries to take advantage of the hardware. I think the best case is when there are libraries that are ready to go.

What I have found is that applications that need a lot of truly random highly parallel access work well on the the GPU. An example would be image processing where you have a rather large kernel for filtering. When this type of application run on a CPU architecture it tends to suffer from cache misses.

I tend to think that if you want something to go really fast you have to invest money to make is so. I have no idea if the market will embrace changing source code for these new platforms unless they include standard linkable libraries.

I would also say that developing Tesla code is not as simple as C single threaded cpu code. And no its not just compile and go... its a bit more complicated.