Home > Services > GPU Computing
The solution to many popular problems such as computer modeling, video processing, 3D visualization, pattern recognition, computational biology and chemistry, seismic analysis, time series forecasting, financial analysis, etc., requires a significant amount of CPU time to perform the computations.
This is especially true for computer simulation tasks, where the quality demands are constantly increasing and detalization of 3D models describing the behavior of real objects, which as a result, leads to the increase of the computational costs.
Equally important is the parallelization of computational algorithms for real-time systems, where the active time of software is the most important criterion. Generally speaking, the user of any software product is interested in processing a greater amount of data – more records, pixels, or use physical scale in less time.
Nowadays, not only PCs are equipped with multi-core processors, numerous tablets and smartphones are already supplied with several cores. The GPGPU (general-purpose graphics processing units) technology, i.e. the use of graphics accelerators equipped with hundreds of cores for performing computations in a variety of areas, but not only computer graphics, is gaining more and more popularity. Supercomputers that were previously available only to large corporations are now widely used by small companies for various applied tasks.
All of these modern trends evidence the relevance of the use of the parallel computing technology in the development and improvement of software products of any company.
For many years Simmakers Ltd has been developing solutions for the parallelization of computations using MPI, OpenMP, CUDA and OpenCL technologies. We offer the development of computational modules, as well as turnkey software products for solving a wide variety of tasks using parallel technologies on any architecture—from supercomputers to smartphones. Simmakers partners with NVIDIA in the area of graphics accelerators usage for solving computational problems, which allows us to be among the first in the implementation of new technologies developed by NVIDIA.
Thus, using the parallel computing technology Simmakers Ltd has developed a software package for computer modeling of thermal processes. The numerical solution of equations describing the processes of heat transfer, filtration and diffusion on a grid consisting of 17,828,087 units, is computed in it for 6 minutes on a PC with a graphics accelerator that supports CUDA technology. Earlier, such computing speeds were available only by using supercomputers.
In the area of GPU computing, Simmakers successfully competes with global leaders such as Dassault Systèmes and ANSYS. For example, the transfer of computations from a quad-core CPU onto a GPU with 442 cores in computer simulation software of these vendors, SIMULIA Abaqus and ANSYS show a slight reduction in the computational time — a maximum by 3.8 times.
At the same time, performed by Simmakers Ltd, the parallelization of the numerical solution of equations used in these software packages, on the graphics processors made it possible to reduce the computational time by more than 10 times.
Our company also has extensive experience in the development of parallel algorithms for the visualization of 3D objects in real time (ray tracing algorithms, computing the volumes, sorting the objects, collision detection, etc.), processing tomographic images, pattern recognition, financial analysis and many other fields.
There are many successful examples of the parallel computing technology usage for the improvement of performance of software products, which resulted in the increase of the number of their users. According to NVIDIA, on the financial market Numerix LLC announced implementation of the parallel computing technology in its new application for counterparty risk assessment and achieved 18x acceleration. More than 400 financial institutions are using Numerix software.
Almost all the basic applications for video processing are already using the CUDA technology for accelerating computations, including the products by Elemental Technologies, MotionDSP and LoiLo. CUDA now also accelerates AMBER – software for molecular dynamics simulation used by more than 60.000 researchers in academia and pharmaceutical companies worldwide to shorten the period of medical drug development.
The constant increase in the use of graphics accelerators by Fortune 500 companies, such as Schlumberger and Chevron in the energy sector, as well as BNP Paribas in the banking services sector serves as an evidence of the wide application of the GPGPU technology.
With long-year experience in the development and application of the parallel computing technology, Simmakers Ltd offers GPU computing, as well as parallelization implementation services to companies from a wide range of industries. This will increase the computational performance of your software products and will enhance their competitiveness on the software market.
Although the increase in the computing capacity of your software product will cost you a certain investment, expert-level implementation of the parallel computing technology will bring your product to a new level and to significantly increase the return on investment in its development.
As a result of parallelization of numerical algorithms your software for computer simulation, visualization, financial analysis or other problems will require less time for processing more information in comparison with competing products. As a result, users become aware that they can leverage software product for performing more computations, analyzing more situations of different nature, are able to handle more data and thus get a solution to their problem with a better quality. Accordingly, more and more users are switching to your product, and your sales volume will grow.