Does CUDA work on weak video cards in the same way as on more powerful ones?

Question

GPGPU

For example, if you write code for the 8400GS, and then run it on a 670 GTX with 1000 cores, will this increase speed, or will you have to rewrite the code? How many cores are on the 8400 GS, there is no such data on the site. What is the difference between versions 1.1 and 2.x of CUDA?

Answer 1

Depends on your application's core configuration. Naturally, with certain kernel configurations (size of grids, blocks, etc.), by increasing the number of stream processors, the number of thread pools that can be executed simultaneously will increase.
As for the difference in the CUDA version... There are heaps of nuances related to thread synchronization, work with global memory, calculation accuracy, etc.
From the available information, there are official guides. There is a table like this: