Little Intel Xeon Phi Now Everyone Can Get

Intel Xeon Phi - a unique processor, like no one else reveals all the advantages of parallel execution of tasks. Created using Intel's Many Integrated Core (MIC) technology, it provides you with dozens of powerful computing cores and a decent chunk of integrated high-speed memory. I think that many programmers, both beginners and experienced ones, would like to “run” their code on such a processor in order to find its bottlenecks, evaluate the impact of concurrency on performance, and so on. It stops one: the cost of the youngest model Xeon Phi is $ 2500, and this is only the processor itself. It is unlikely that many will venture to acquire such a system for personal needs, and such a need, as already mentioned, happens.

Now the life of enthusiasts is becoming a little easier. Education CentreColfax Research, with financial support from Intel, has launched a remote access program for a server cluster based on Intel Xeon Phi. Details of the program - under the cut, but first briefly about Intel Xeon Phi itself - a long time ago we did not write on this topic.

Intel Xeon Phi - A Short Excursion

The company was engaged in the architecture of Intel Many Integrated Core (Intel MIC) for a very long time, starting from the beginning of the 2000s, perhaps even longer than it would have liked, since the visible results did not appear immediately. Initially, it was supposed to release a GPGPU product, but then the plans changed, and as a result, a computer-to-computer coprocessor was created with its own operating system, computing resources, and a system for interacting with the host platform.

The first generation of Xeon Phi coprocessors was released in early 2013. Their specific feature (reflected in the name) was that, being with processors, the Xeon Phi x100 seriesThey were intended only to speed up calculations and could not be used to launch the main operating system of the entire host. However, their task - the acceleration that can be parallelized - they coped very well, as much has been written, including Habre .


First-Generation

Intel Xeon Phi Intel Xeon Phi x100 series provided the user with about 60 x86 cores with 512-bit vector modules operating at about 1 GHz and providing double-precision computing speeds of more than 1 TFLOPS. As you can learn even from the above link, this provided a multiple increase in the speed of programs, so Xeon Phi has deservedly registered in many research and scientific computer systems.

The new generation of Xeon Phi,x200 , was announced in the middle of last year, we already wrote about their characteristics and nomenclature . The new Phi have three distinctive features compared to the previous ones. Firstly, they have up to 4 threads per core (these are the only Intel processors with such characteristics). Secondly, they have their own onboard volume of ultra-fast MCDRAM. And thirdly, the new generation got rid of the prefix co - now these are full-fledged processors with the ability to boot the OS. Thus, nothing - except for monetary aspects, probably - does not prevent you from assembling a computer based on Xeon Phi and getting your own 60 with a tail of cores, since there are also motherboards for these processors .


Second Generation Intel Xeon Phi in Computing Module Design for Server

Parallel Programming Week

Now we proceed directly to the initiative. The Colfax Research educational center, known for its articles and training materials on programming in general and parallel in particular, with the participation and assistance of Intel, installed a server cluster on its platform based on the Intel Xeon Phi 7250 platform . Their resource is proposed to be used by everyone.

To participate in the program , just register . After the approval of your application, as free capacity becomes available, you will be given SSH access to the cluster for a period of one week (7 consecutive days). The server has everything necessary for the programmer to work, taking into account the specifics of Intel Xeon Phi:

• Intel Parallel Studio XE Cluster Edition, including:
  
  
  • Intel C, C ++, and Fortran Compilers
  • Intel MPI Library
  • Intel OpenMP Library
  • Intel Threading Building Blocks (TBB) Library
  • Intel Math Kernel Library (MKL)
  • Intel Data Analytics Acceleration Library (DAAL)
  • etc
• Intel Python 2.7 and 3.5
• Intel Caffe
• Intel TensorFlow

Each user is allocated 50 GB of disk space. Do not forget to take with you everything important. Creative success to you!