I will buy a “Tesla S1070” for my works.
In manual guide: Number of Multiprocessors of Tesla S1070 is 4x30.
I really don’t understand why number of multiprocessors is 4x30.
So does 4x30 Multiprocessors is the same as 120 Multiprocessors, is it right?
If anybody knows, please explain.
Thank you very much. :)
May be I should post in “General CUDA GPU Computing Discussion”
Tesla S1070 have four Tesla C1060 inside, each of them has 30 multiprocessors.
So, for CUDA application S1070 will look like four devices, not like one device, and each device will have 30 MP and 4 GB RAM.
Thank you, AndreiB.
So, it means that to get the all the best performance of S1070, I have to build a solution for multiDevice?