I’ve learned recently that sometime in April nVidia is going to release an AGP-interfaced card with G84 graphics chip. It is going to be a VERY affordable solution with a price tag around 100 euro.
My question is as follows: Does this card support CUDA-based programming for general purpose computation?
To my mind this will greatly simplify the entrance for a new programmers as it represents a cheap testbed for initial designs. It appears that G84 will have 32 ALUs, 1/4 of the GTX flagman.
Here is another interesting question, IF and WHEN we’ll have a chance to see G8 family in MXM-interfaced cards?
The latter is not a pure curiosity. We now study the feasibility of porting of a certain computationally intensive pattern recognition application to GPU. It would be VERY beneficial to have a possibility to run it on a notebook.
Can you comment now? It’d be nice to be able to test CUDA development on an “inferior” platform. Are there any plans to make CUDA useful on older cards, or should we just find our own ways to access those GPU’s?