Director server group Intel company- Diane Bryant, which, in our opinion, is prepared for the presidency of Intel company - answered a series of questions in an interview with analysts at IDC. One of the questions asked have offended the head of the server business unit. She wondered the benefits and the market share of GPUs NVIDIA Corporation in the supercomputing market.
Recall, this products in NVIDIA Tesla line. Tesla Accelerators are aimed at the implementation of non-graphical calculations GPU forces or, for short - GPGPU. We must pay tribute to NVIDIA, it is the first in any appreciable volume beginning a few years ago to supply the market of supercomputers GPGPU-boosters. Later, AMD joined, , and another appeared later Intel Knights X accelerators, which then became known as the Xeon Phi. So the market NVIDIA- accelerators really topped with a noticeable margin from the competition for several years.
The emergence and spread of processors Xeon Phi beginning reduce the demand for NVIDIA Tesla products.
In addition, to work with NVIDIA Tesla CUDA programmers must learn, etc., whereas Xeon Phi programming is almost the same as for the x86-compatible platforms. Finally, the new Xeon Phi - actually processors with the ability to boot the operating system, and Tesla require a complete platform with the same Intel Xeon processors.
New GPU GP100 in Pascal architecture on double-precision operations are able to demonstrate the performance of 5.3 teraflops, while the Xeon Phi Knights Landing issued in the same time a little more than 3 teraflops. And while the market for deep machine learning is small, Intel company on NVIDIA share of supercomputers on the market really can not worry.