GPU Computing

GPU Computing Creates New Paradigm

The origin of high-performance computing (HPC) was based on clustering multiple CPUs, and more recently, clustering scores of multi-core CPUs to achieve higher levels of computational power. The GPU (Graphics Processing Unit), and the recent advent of GPU computing, have created a quantum shift in computing architecture by introducing a hybrid model whereby GPU I/O cards now work in conjunction with CPUs.

Utilizing the latest Intel® multi-core CPUs and PCI Express 2.0/3.0 technology, Trenton servers are ideal for building GPU computing solutions based on NVIDIA® Tesla™ GPU 20-series processors and the CUDA™ parallel computing architecture. The TCS4501 4U rackmount computer supports up to four Tesla cards in a rugged enclosure.

Servers Designed for GPU Computing

While CPUs are excellent sequential processors adept at serial operations, GPUs were designed from the start to excel at processing similar data that can be split into many pieces and processed in parallel. This is due to the fact that GPUs contain hundreds of parallel cores which are capable of running thousands of parallel threads.

Maximizing GPU performance, however, begins with the underlying hardware architecture.  At Trenton that means single board computers & PCI Express backplanes which are designed from the start with data throughput in mind.

Strategic NVIDIA Partnership

Programmers take advantage of this architecture by directing the most performance-critical sections of their programs to run on multiple GPU cores. By offloading the CPU, 10x or greater performance increases are common, and this factor is sure to increase as future generations of GPU computing technology are developed.

As a member of the NVIDIA Tesla Preferred Partner (TPP) program Trenton is now creating GPU computing solutions which support NVIDIA’s Tesla GPU 20-series computing products and CUDA™ GPU architecture.

Many uses of GPU Computing

GPU Computing Applications

Trenton’s knowledge of high-speed bus technology, BIOS configuration, system power requirements, shock & vibration, and thermal characteristics is incorporated into every computer system, single board computer, backplane and embedded motherboard.