So little explanation. To run a aplication with nVidia CUDA or ATI Stream you need to implement in the program very nonstandard application model of witch most of the programmers think that it is hart to do.
On the intel card we don't have any shader units or cuda cores or whatever, we have many of the x86 cpu so you can program in the normal way. So it is like having a graphic card only capable to do CUDA or Stream so it will accelerate some programs but if you want to play a game, heh sorry we can't do that. It is a Parallel processo as i said.
Well jump to the specs|
We have 32 x86 cores clocked at 1,2 ghz (in fermi 1,4 ghz for 480 cuda cores in GTX 480 or in ATI 850 mhz for 1600 stream processors)
also the processor is equipped with 64 L1 cache for each core and 8 mb L2 cache shared to all cores.
It has also a special 512 bit vector unit on each core witch enables 4x times more calculations than today's x86 cpu's
also something like more advanced hyper threating (it is called 4-way Hyper-Threating) so the card can do 128 threads simultaneously.
All of this is supported by 2 GB of GDDR5
It is a part of a bigger project called MIC (Many integrated core)
In plan is a unit with 50 x86 cores in 22 nm process.
One think i don't know when it will be relieved :D
No comments:
Post a Comment