Own goal from Intel: GPUs are until 14 times of more rapid CPUs say her engineers!
Time now, the epicentre of conflict between Intel and Nvidia is the output of processors of general use (CPU) and processors graphic (GPU). The own goal however that put Intel recently will remain in. history!
a team from researchers of Intel presented recently a work in International Symposium on Computer Architecture (ISCA) that proves that one roughly two years GPU, as Geforce GTX of 280 Nvidia she is not from ten until thousand times of more rapid eight months of processor of Intel, Core i7-960 but only 2,5 times (on average roughly) and until 14 times more rapid!
the researchers of Intel in deed admitted that they made better possible in order to they optimise the software that ran so much for the CPU what for the GPU, contrary to various tests of past that code of software were optimised for the GPU and no for the GPU.
the effort that they made optimise the code and for the two species chip proved, says Intel, that the shear in the output close drastically. Nvidia from the other, and well makes here that we say him, and it says that the last one did not optimise the code of programs that ran for the GPU, despite the tests ran out-of-the-box while it supports her statements, that is to say enough programmers and companies of growth of software have seen 100X increase in the output when they ran their applications in GPU instead of CPU. Moreover it stresses that it was used passed generation card graphic.
This of course is truth, however in this. to blaim is Intel but Nvidia that didn’t present the &? quot? Fermi&? quot? earlier. The terminal date for the deposit of work in the congress was and as we know Intel earlier had presented Core i7-960 while Nvidia had not presented Geforce GTX 480 and architectural Fermi in. functional piece of silicon.
Moreover, and according to the documents that they presented the engineers of Intel, were realised will optimise in the code that ran so much in the CPU what in the GPU. It can they did not also make. better that says the reason for the GPU, but no that they did not make also nothing as claims Nvidia. Other they show the elements in the work.
doesn’t
Undeniably, no one say that the &? quot? parallel computing&? quot? that it appears as advantage Nvidia, specifically in applications HPC (High Performance Computing) is not the better choice. Specifically for the purchase supercomputers and a exigent datacenters increase in the output, even at 2. 5X, she is particularly attractive, thus and differently, for the companies. Also, Intel with the particular work that it deposited, we cannot say that. aytopyrobolithike already alone her in the leg, but it is not also the most beautiful thing you admit that in certain cases your technology comes. second.
Nvidia, as always. excessive. In certain cases, yes, the &? quot? parallel computing&? quot? it is the ideal solution, but is not also. solution in everything. Specifically when her better GPU costs two and three times more (and we do not speak for a GPU of type Tesla thus) from a very good processor of Intel. The war in any case between the two companies does not say kopasei. We have in deed the impression that we found itself only in the beginning! More you can read.