Overclock.net banner

1 - 3 of 3 Posts

·
Registered
Joined
·
4,592 Posts
Discussion Starter · #1 ·
This is an old presentation on the progression/development of Nvidia's GPU arch. (the latest arch. discussed is the GTX 280). Are the Stream Multiprocessors mentioned in the presentation the equivalent of CUDA cores now? Are CUDA cores simply the number of shader units? Or are they something completely different? Also why is there an integrated logarithmic function (i.e. in the SFU on page 44)? What use are logarithmic functions in graphics processing?

https://download.nvidia.com/developer/cuda/seminar/TDCI_Arch.pdf

I tried to find something more recent at Nvidia on their GPU arch. but couldn't find anything.
 

·
Registered
Joined
·
3,105 Posts
For newer cards, I suggest you search for the white papers for each architecture. They contain more information. A brief search pulled up the Ampere, Pascal, Turing and Volta white papers.
 

·
Registered
Joined
·
630 Posts
For newer cards, I suggest you search for the white papers for each architecture. They contain more information. A brief search pulled up the Ampere, Pascal, Turing and Volta white papers.
This documentation has high value to well trained brains with academic knowledge.
But this is also a great proof, that vast majority of people within OCN, they are clueless of NVIDIA goals and plans, at finding markets for their products other than gaming.
There is no need for someone to be able to fully understand mathematical terms, so to get good clues from these documents.
Because they also include application examples, that even untrained brains they can get clues of what is best for them to buy and or to ignore.
 
1 - 3 of 3 Posts
Top