United States blockade of hyper numerical technology export to China what s
United States embargo on the news of the super computer key technology exports to China, let the word super is also became the focus of media attention.
Homemade Super is “milky way, 2nd” global champion for have on several occasions, there are few but proud of netizens “supercomputers” a more profound understanding of this concept. This paper will introduce the super computer-related knowledge, and the sequel will explore United States blockade policy on China’s hyper-development impact.
What is hyper-calculated?
Supercomputers, also known high-performance computing, High Performance Computing in English, referred to as Super is/HPC. HPC to address industry, born out of the need for massive computing power technology, modern Ultra is essentially a large number of configuration host close to ordinary home computers/business workstations connected by a high-speed network the host cluster. It is run by a dedicated operating system, each host is a node; the computing power of a large number of nodes together, won far more than a single or several common PC computing power.
Technical characteristics of ultra is
Super is often related news see the following comment: “such a strong run xxx game card card? “Ordinary consumers” speed “concept is usually played some big games how many frame rate, detail what you can achieve grades. Funny thing is, ultra is to perform the tasks with the 3D games have many similarities.
In the computer world, “computed” can be divided into two integer and floating point calculation classes. 8X4 is an integer arithmetic, and 1.5912/0.4 is a floating point calculations. Our most common programs are mainly for integer calculations, complemented by a small amount of floating point part; but 3D gaming, image rendering, video effects of such work would be a lot of floating point code.
In science and engineering fields of common computing tasks are partial floating of an otherworldly: typical atmospheric modeling, material stress, protein molecular simulation, model analysis and so on. Of integer operations in these areas is very small, and our common PC programs are quite different.
Floating point arithmetic tasks feature is very easy to highly parallelized. Split is a simple calculation can give lots of calculating core to complete, unlike the integer calculations many times only to a few core. In addition, intensive floating point program can use the high floating point instructions-for example AVX, FMA to significantly enhance the speed. Procedures used in contrast to general consumers often only uses old SSE, SSE2, is not easy to gain speed in the new directive.
Because of these features, areas of high performance computing applications can easily benefit from the CPU’s floating point arithmetic capability improvements. Below is the theory of generations of Intel Xeon CPU’s floating point performance, can be seen to benefit from advanced instruction increases, every new product properties are promoted more than 1 time; us ordinary consumers it is hard to feel such a big advance, but benefited from scientific and engineering calculation programs. YSL Galaxy S5 Case
More floating-point performance, characteristics of massively parallel scientific application is suitable for running on the GPU. GPU can be seen as there are many small core of a CPU at the expense of integer performance and single-core performance in Exchange for the whole chip superior floating-point performance. CPU and GPU jointly perform a scientific calculation is called heterogeneous computing, heterogeneous computing began promoting in the industry in recent years.
In addition to performing operations the CPU, the GPU core, Super two key technique is memory and the Internet. High performance computing tasks usually require a lot of memory, and the increased space and calculation of the number of cores increases; Moreover, the memory also has a high bandwidth to quickly pass the data to the processor core. Internet is key to super system thousands of hosts built, hosts the more network bandwidth, latency and stability requirements are higher.
Super technique development YSL Galaxy S5 Case
From the foregoing we can know, Super requires floating point performance very strong, very high memory bandwidth, and capacity of computing chip, as well as high-performance large-scale Internet. High performance chip technology monopolized by four giant: Intel, IBM, AMD and Nvidia. Among them, IBM and Intel CPU on the one hand, on the other hand developed a similar super multi-core GPU chips. Such as the new generation of Intel Xeon processor Phi is dozens of special enhanced floating-point performance of the Atom core integrated together, accompanied by 3D packaging memory bandwidth up to hundreds of GB/s manufactured by. A new performance of Xeon Phi 2 trillion floating point operations per second, equivalent to 3 times times the fastest Xeon CPU. Nvidia and AMD production calculated on the basis of their GPU cards Tesla and FirePro, and its floating point performance also reached the level of 2–3 trillion times per second.
Due to manufacturing technology and cost constraints, Tesla and Xeon chips like Phi cannot do a lot of memory capacity, cannot execute code that requires large memory space, which limits the penetration of such chips. But because the chip’s performance than CPU has a clear advantage, and several companies are trying to solve the problem of memory space, they will be the mainstream of the industry in the future. Intel hopes its Xeon Phi can unify in the future; Nvidia collaboration with IBM, will launch a Power CPU+Tesla computer card heterogeneous solutions AMD pushed their HSA heterogeneous scenarios, hope Opteron CPU+FirePro computer card combinations can take place. Now Intel’s advantage is large, can get a certain share of the Nvidia/IBM Union.
Internet technology is not monopoly like computing chip. At present, many companies are provided with a Web solution in this regard in China also have a world class level. Ultra is the most advanced network bandwidth up to 100Gbps can make tens of thousands of compute nodes communicate both stability and low latency.
Super is of national significance
As noted above, service above technique is mainly industrial and scientific research. Powerful computing power to businesses, schools and defence research has a great deal of help, even indispensable. When United States is announcing the end of nuclear testing, its clout is hyper-numerical simulation test technology is mature enough. Due to the heavy investment, the payback period is long, States of a project implemented by the Government or the subsidy. Today, Super technique and the level of resources has become one of the important indicators of a country’s technological strength.
Chinese Super technology developed rapidly in recent years, many times world champion before and after dawn, Tianhe ultra-performance chart is. United States Announces Super technique embargo against China is no doubt felt the threat of China in this field. United States embargo would have any effect on China? Please look at the specific analysis of an article under.
Tesla Model s-P85D
P85D Tesla MODEL s is Tesla sells cars in the fastest and most expensive, in terms of appearance and the remaining few models did not much change, the main change is to drive and battery capacity. MODEL s-P85D, 691 peak horsepower BHP, four wheel drive, 0–96 km/h in 3.2 seconds.
View details of the voting >>
Originally published at carbonfibercase.wordpress.com on December 31, 2015.