Me2019H I knew when you show up you will say sign me in.
For newbies:
Cuda is a programing method use only by Nvidia.
Opencl is the standard, it is designed to run across multiple platforms. From Nvidia to none Nvidia GPUs, CPU, FPGA, Android Phones, Super Computers or even in your pull wagon (wheelbarrows ) you had when you where a child....
In general Opencl do as equal as Cuda. Now there are a few instructions that makes Cuda faster in certain application but in other Opencl will run faster.
What we are talking about?
GPGPU General-purpose computing on graphics processing units
Why?
Because in general most GPU have at most 4 cores. So at best you could split the CPU and make them run 4 time the same work load.
Lets Ignore that a PC has many other elements that affect speed.
And lets take cayoenrique's sample CPU Intel 4 Core @ 2.00GHz .
Computing Power = 4 x 2 Ghz =
8
Now lets compare with dvlajkovic GPU it has 16,384 cores @ 2.235 Ghz
Computing Power = 16,384 X 2.235 =
36,618.24
If some how I could use GPU to make simple calculations the I could have an improve!!
The improve from my CPU to dvlajkovic GPU is that in the same time I do 8 computations he do
36,618.24. I am so shame on my computer....
Code:
Any OpenCL based tool would be also welcome if it can surpass cudabiss.
I do not have it. But you guys can try to build one!!! Nor I can not provide assurances that your final build in Opencl will be faster than Cudabis. Only you can prove that to me.
Code:
I'd like just to point out that a very few people among us here are using linux, that's why we would really like a windows version, too -if possible.
I learn from my last Tutorial in Encryption. People did not got involve, even when I did post a thread how to run the lLinux instructions in your W10. So now all my sources have a makefile that can build either windows exe or Linux posix programs.
But you are required to install necessary drivers to compile Opencl.