According to the rating of the top 500 supercomputers, first of this year, the fastest machine was the Chinese Tianhe-1 (the "Milky Way"), who is in the world ranking of fifth place (563 teraflops).
But in November the top 500 supercomputers in the world system led Tianhe-1A, located in the State Supercomputer Center in Tyanzhine in China, the best performance which is 2.57 Pflop / s. One and a half times more than the most frisky of the old favorites — located in the U.S. Jaguar supercomputer.
The third place the system also took from China — Nebulae performance Pflop 1.27 / s.
The most high-spirited conditionally Russian supercomputer — deep in the ass, at 17th place — the highest performance system "Lomonosov" from «T-Platforms" is 350 Tflop / s.
But not even that bad. Creators Ranking noted that included in the latest edition of 42 supercomputer from China — the second largest number of systems after the United States. Russia there just is not visible.
But supercomputers are needed for a reason. Consider their fundamental tasks — from nuclear physics to genetics and pharmacology, from the aircraft to the flow of plasma stabilization methods. The number of supercomputers — is a measure of how active in the country are carried out practical research and development of truly advanced level.
As we see, in China with it all perfectly. And in Russia — it is terrible.
Worse than that — China already has its own microprocessors, fully competitive with global-level minions. I'm talking about the line of Loongson.
The latest novelty there — it Loongson 3, which different from those already available in the consumer PC hardware Loongson 2F broadcast x86 instructions and includes a large number of cores (4 to 16) able to process commands independently.
Quad-core version with an estimated 1-1.2 GHz clock speed and 2 64-bit nodes to perform floating-point operations in each nucleus should find application in a wide range of products — from desktop computers to set-top boxes (in the chip implemented additional annotations specially to optimize the multimedia playback.)
Eight-core version is likely to become the "heart" of the supercomputer. It will include four cores and four ordinary coprocessor GStera, created for the rich mathematical calculations. These elements are important only because they are better managed with mathematical operations test Linpack, using linear algebra for performance measurements fastest mainframes (based on the Linpack and compiled the top 500).
Microprocessor Report Analyst Tom Hafhill explicitly stated that even if the Chinese microprocessor does not show record-breaking performance, the occurrence of the Chinese competitive in comparison with the "Western" chips for supercomputers — it is only a matter of time.
And at that time doing in Russia? Where, for God's sake, Babayan and Elbrus (microprocessors for which, by the way, you still bought in China)?
Yes here percent this Elbrus-2000 — made in China factory TSMC. It even surpassed 500 MHz Intel Pentium III tests in SPEC. Cool as much horror. Especially in comparison with the Chinese, have long walked out for the level of performance gigahertz microprocessors with the lowest energy, so a clock running on batteries in communicators.
To work on Chinese microprocessor Loongson 2F already ported Debian GNU / Linux, gNewSense, Gentoo Linux, Red Flag Linux, NetBSD evbmips / gdium, OpenBSD OpenBSD / loongson. And, of course, ported Windows CE and Google Android. Work on the preparation of Slackware version Linux.
Computers on this Chinese microprocessor — from the desktop to the tablet a la iPad and smartphones to massive — are already available, with prices on the fun. For example, models a la iPad touch screens cost from $ 100.
What ported to the Elbrus-2000? Nothing. What devices on it are available for sale? Yes, no, as the cost of the microprocessor simply insane.
Worse than that — Loongson 3 already has a hardware broadcast x86 instructions, which allows it to run on Windows and in general anything designed for the Intel architecture x86.
Will there be a hardware translator Elbrus x86? How annoying would that sounds, it's not even existed in the long term.
Why is the Chinese "stupidity" is so successful, and the Russian "brain" in fact produces results that differently than ridiculous and not to name, anyway?
Maybe it's time to remove the arrogance and start to look more closely at the experience of Chinese colleagues?