Utilizing Nvidia's new GPU's with Tensor cores for backtesting

[chris500]

I've noticed that Nvidia's new gpu's have tensor cores.

Does anybody have any experience using/programming them for use when backtesting? Is it even possible?

The computing power seems immense (in the Teraflops).... much more powerful than CUDA cores.

 

[tendimsol]

Traditional:

https://devblogs.nvidia.com/hybridizer-csharp/

"lets you program GPUs and other accelerators from C# code or .NET Assembly."

ML related:

"Retrieves and returns additional properties (total memory, number of CUDA cores, etc.) for the specified GPU device."

"parallelization across multiple GPUs and servers."

https://docs.microsoft.com/en-us/cognitive-toolkit/Debugging-CNTK-source-code-in-Visual-Studio