- Joined
- Dec 15, 2011
- Messages
- 13,640
- Likes received
- 5,494
- Thanks received
- 1,485
OK I get that, but if had access to a SPARC proc with what, 1024 effective vcpus - is that better? Or SGI's old NUBUS tech, where all memory was shared between CPU, IO, GPU the lot?
TBH, I don't know enough about SPARC hardware to comment. However, the thing w/ GPUs is that they're dirt cheap (several thousand cores for a few hundred quid; there are dedicated GP-GPU cards like the NVIDIA Tesla but, IIRC, they really aren't worth the money). In contrast, SPARC hardware is relatively expensive, right? Anyway, basically, what you're saying is right - it's just using a GPU for general purpose (GP) processing, hence GP-GPU, only it is a cost-effective way of achieving massive parallelism, with several thousand cores available on tap, and you can put a bank of GPUs together relatively easily.
I'll have some of that!!! That's more then NLadyUK!!