Advanced Computing in the Age of AI | Friday, March 29, 2024

AMD, HP Join Forces to Accelerate Server Graphics 

Chipmaker AMD said this week its latest server graphic processor has been designed into the latest generation of Hewlett-Packard's Proliant servers.

Separately, HP said it Helion cloud platform would support a range of Oracle database applications.

AMD and HP said the combination of the FirePro server GPU and the HP ProLiant DL380 Gen9 server targets a range of high-end computing workloads used in academic and government research projects along with "deep" neural networks as well as oil and gas research.

AMD said it server GPU, which is said to deliver a "double-precision" floating-point capability, leverages the software ecosystem built around OpenCL 2.0-based computing.

The GPU server chip is based on AMD's Graphics Core Next architecture that targets high-end compute workloads, the company said. Along with accelerated double precision floating-point computation, AMD said the new graphics processor tops the 2.0 TFLOPS double precision barrier.

The processor is also based on the AMD Stream technology that allows FirePro graphics cards to be used for high-end computing workflows by leveraging massively parallel processing. Hence, the parallel processing architecture is said to be useful for application beyond graphics.

As with most of today's high-end processor, AMD stressed lower power consumption along with performance. The FirePro with 16 gigabytes of GDDR5 memory comes with a maximum power consumption of 235 watts.

AMD also stressed that the graphics processor design seeks to leverage open standards like OpenMP as well as OpenCL designed to boost the performance of GPU computing and multiple graphics processor in a server. OpenMP is an API for high-level parallelism in C, C++ and other languages.

HP said the integration of the AMD processor in its ProLiant DL380 servers would boost the performance of graphics-intensive applications. It expects to target government, university and energy industry customers with the souped-up server.

AMD also claims its GPU server outperforms rival Nvidia's Tesla K40 single GPU server card by delivering peak double-precision computing horsepower of 2.53 TFLOPS. Peak single-precision floating-point performance totals 5.07 TFLOPS, AMD claimed.

In targeting the energy sector, AMD noted that the oil and gas industry along with engineering and computer science sectors have made substantial investments in OpenMP to scale workloads. The chipmaker said it also collaborated with Pathscale Inc., a developer of X86-based compilers.

Along with HP ProLiant servers, AMD said its server GPU also will run on HP's ProLiant WS460c graphics blade server.

Separately this week, HP rolled out an applications suite for Oracle databases on its Helion managed cloud platform. The move is designed to help ease organizations' transition to the cloud by providing services on the Helion platform that support Oracle applications. The service includes application provisioning and support for a range of Oracle applications, middleware and databases, HP said.

HP said its Helion platform for Oracle business applications is available immediately worldwide.

EnterpriseAI