Perhaps the microchip manufacturer should rethink its future or, at the very least, what its architects and executives say in public events because although it is done with good intention, the truth is that it can end up propagating the competition, in this case, Nvidia.
This latest company specializing in the manufacture of calculation microprocessors for graphics cards (GPUs) is about to disembark in the field of CPUs for mobile devices such as smartphones, MID’s and netbooks thanks to its Tegra family doing so the competition to Intel. The environment between both manufacturers is, therefore, more tense than usual, and this has been noticed in a certain dialectical battle between managers of both companies.
One of the fields of discussion is that of the power of GPUs for use in intensive calculations, in which there is a kind of “urban legend”: the power of a GPU multiplies by one hundred that of a CPU. Well, in an international symposium on computer architecture held in the French town of Saint-Malo, Intel presented a white paper that tried to break this myth, but the remedy has been worse than the disease, since in it the company California states that an NVIDIA GeForce GTX 280 GPU running a kernel (operating system kernel) is “only” 14 times faster than an Intel Core i7 960 CPU, one of its latest models.
I don’t know what Intel intended exactly, but to say that your rival’s product is “only” 14 times faster than your own doesn’t seem like a very compelling sales pitch … or, at least, I don’t they would have won me as a customer to buy them a supercomputer.
How could it be otherwise, Intel’s argument has served nVIDIA to boast about it since, in addition, the GeForce GTX 280 model is not precisely its latest generation, while the Core i7 960 is one of the last micros of the Californian.