Actually got around to looking a CUDA on Nvidia cards at work yesterday.
Got the SDK and sample code up and running on the mac pro, next thing to do is find some maths problem to do on it which will be faster than on a CPU / Grid. well at least that it will be faster given a better GPU the NVIDIA GT 120 isn’t that fast considering there is 2 Xenon quad core chips 2.4ghz. on the mac then there at 6 macs connected to the Xgrid. so it will be a bit hard to do a direct comparison.
Also had a informal meeting with the lab manager from physics we discussed how that for ever increasing amount of remote courses at the university how he plans to do physics labs for the astronomy course next year.
Ended up sending him to have a look at the UTS remote labs http://remotelabs.eng.uts.edu.au/ and told him to keep us posted since it’s all of interest.
So any ideas how to show that CUDA and GPU computing is useful ? or how to conduct an astronomy lab when your students are in Queensland ?
CUDA with Open Inventor is probably the lowest point of entry I’ve seen. Though IIRC at the time the support was just starting to happen so not so complete.
Also:
http://www.youtube.com/watch?v=wviPeF1XTsk
Some compression algorithms should work nicely on the GPU, but actually writing the code would likely take a while.