NVIDIA Tesla GPUs in the new HP ProLiant SL390s G7 Servers

HP's new SL series is actively gaining momentum — the idea of ​​combining power and cooling in one basket and installing half and full width servers in it in various combinations has proven to be quite successful. This, in general, is not surprising: for a relatively small amount of money, you can build a high-density system that, in some cases, loses to traditional blade solutions, but in return provides excellent performance in terms of the volume of rack-mount capacity.



And with the release of half-width new servers under the SL390s G7 index, the already rosy picture has become even more colorful. Now, in addition to the faithful Xeon's, the lucky SL390s G7 owner has at its disposal the latest NVIDIA Tesla third-generation graphics computers. If desired, in the cool server you can arrange the Crysis 2 championship at the maximum settings * , but it is better to do otherwise: “sharpen” the software tools you use to work with Tesla and enjoy the increased performance.



Who may need it first? If you do not take into account the recently popular Bitcoin mining as a worthy occupation for small and medium-sized businesses, the SL390s G7 can best show themselves where you need to make serious parallelizable calculations. For example, calculate trends based on financial statistics. Or to predict the behavior of the earth's crust in the long term, or to simulate the aerodynamics of a racing car, taking into account the turbulence of air from the vehicle in front. And, you can calculate the properties of the division of proteins, or, for example, render video for a new Avatar. Many, many lessons can be found for the Tesla GPU installed in the SL390s G7. And even if today's familiar tools still do not know how to use the full power of graphic computers, tomorrow they will be able to, and the happy owners of the SL390s G7 will be able to meet new technologies fully armed. Moreover, drivers and patches for mainstream operating systems are already implemented, as well as special tools for managing computing modules from NVIDIA.

')

The applications of NVIDIA Tesla and SL390s G7, respectively, are not based only on marketing constructs. There are quite a few practical implementations. Tesla graphic computers are used, for example, in calculations for the computational dynamics of liquids at the National Center for Atmospheric Research in the United States. Or for the provision of geographic information services in BAE Systems (formerly British Aerospace). Or even for such a seemingly distant IT field as the fashion industry. The Israeli company Optitex models the structure and behavior of the fabric in virtual garments shown by virtual models. To do this, use a special plug-in with Tesla support in the program of three-dimensional modeling.



A few words about the SL390s G7. Unlike previous models, they are available in several different heights and are the first servers in the series, equipped with NVIDIA Tesla graphic computing modules. By releasing these products, HP became the first major vendor to support the technology of increasing computing power through the use of GPU resources in a massive and affordable solution. Previously, this promising technology was used only in specialized models, most often inaccessible to small and medium-sized businesses. "







Servers SL390s G7 are available in three different versions, differing in height. The single-height version does not contain Tesla graphics cards and is designed to organize standard high-density solutions. But the double and quad height servers already contain three and eight NVIDIA Tesla adapters, respectively. With their help, it becomes possible to combine up to 256 Tesla compute modules in one standard rack size.



The well-known Integrated Lights Out (iLO) technology of the third version is responsible for management and control. In addition to it, the package includes a standard set of software tools, including Smart Start, Rapid Deployment Pack and PXE. Another factor of convenience in the work are the I / O ports located on the front panel, which facilitates the switching of individual servers.



The rest of the “stuffing” of the SL390s G7 also does not lag behind the set bar: up to two Intel Xeon 5600 series processors can be installed in each server, 12 DDR3 slots are allocated for the RAM, and 2 Gigabit Ethernet ports and one 10 Gigabit Ethernet are responsible for the communication ( SFP +). Optionally, you can install the module Infiniband QDR, if suddenly in the server economy of the company there is a corresponding switch.



Under the curtain, it is imperative to separately note that the range of tasks for the SL390s G7 is not limited to the listed areas only - these servers should be useful wherever there is a need for high-density, scalable and easily manageable computing systems. That is, in almost any serious business. The main thing - rather to implement support for computing using the GPU.