Hi,
If you'd like to take advantage of the full power of X10, you will
probably need to have a look at how parallelism work. At a certain
point, you will be looking forward hardware which supports
parallelism. If you are newbie to this subject, please have a look at:
http://www.nvidia.com/object/what_is_cuda_new.html
http://x10-lang.org/X10+2.1+CUDA
Please see Note (c) in regards to supported CUDA cards.
OK. Let's start, then:
If you are backed only by yourself (like me!), you need to find cheap
hardware in order to get started with CUDA. So, let me give you some
hints and some ideas of how much you will spend.
If you have *any* GeForce Series 8 or newer card... you can start
playing with CUDA.
In the NVidia website it's mentioned that models 8400 and superior
support CUDA.
But you can try inferior models too and eventually it works: I have an
old GeForce 8300 and it supports CUDA :)
If you have a laptop, check it out. I have a Dell laptop with a
GeForce 8600M, which supports CUDA :)
If you decide for a buy, you may be interested on some information I
found:
=== IMPORTANT: Technology and figures as they are in the UK, December
of 2010 ===
Entry level, low cost, near end of life
-------------------------------------------------------
Some models are GeForce 8600 and 8800.
You can buy a brand new GF8600 for £25,00.
Some concerns are: relatively old technology and relative inferior
"compute capability".
Please see Note (a) in regards to compute capability.
Entry level, low cost, high performance
-------------------------------------------------------
I recommend GeForce GTX 460.
GTX460 is my "best buy" recommendation.
In particular, this model is very affordable. It performs better and
consumes less power than GTX 465, which numerically is 'superior' and
leads us to think it should be superior.
There are some other models which are clearly superior than GTX460 but
the cost is relatively too high, at the moment at least.
I found a Palit 2GB GeForce GTX 460 Sonic for £172 on eBay, brand new.
I bought it! :)
Professional, high performance
-------------------------------------------------------
This article does not cover anything like this.
Please research yourself. Look for Tesla on Nvidia website.
Some important aspects
-------------------------------------------------------
1. Verify if you have enough power supply in your system.
Please see Note (b) in regards to power supply requirements.
If you already have one graphics card and you will add another one,
you need to make sure your power supply copes with required power.
Certain models, like GeForce 8800 GTX are known to be 'power hungry'.
Beware.
2. Verify if your computer case is well ventilated, in particular if
you have 2 graphics cards (possibly in SLI configuration). Remember
that the energy effectively delivered by the power supply needs to be
dissipated somehow. Certain professional power supplies have a big,
well positioned fan which contributes to dissipate hot air, but you
still need to consider ventilation carefully.
NOTES:
-------------------------------------------------------
Note a. Compute capability:
http://en.wikipedia.org/wiki/Comparison_of_Nvidia_graphics_processing_units#Compute_capability_table
Note b. Power Consumption and Temperature
http://www.tomshardware.com/reviews/geforce-gtx-460-gf104-fermi,2684-13.html
Note c. Supported CUDA cards
http://www.nvidia.co.uk/object/cuda_gpus_uk.html
Have fun
Richard Gomes
http://tinyurl.com/frgomes