GIGABYTE Introduces Accelerated Computing Servers with Significant Memory Bandwidth Gains Using NVIDIA HGX™ H200 Platform

August 15, 2024 - Giga Computing, a subsidiary of GIGABYTE and an industry leader in generative AI servers and advanced cooling technologies, today added two new 8-GPU baseboard servers to the GIGABYTE G593 series that support the NVIDIA HGX™ H200, a GPU memory platform ideal for large AI datasets, as well as scientific simulations and other memory-intensive workloads.

G593 Series for Scale-up Computing in AI & HPC
With dedicated real estate for cooling GPUs, the G593 series achieves stable, demanding performance in its compact 5U chassis with high airflow for incredible compute density.

Maintaining the same power requirements as the air-cooled NVIDIA HGX™ H100-based systems, the NVIDIA H200 Tensor Core GPU optimally pairs with the road-tested GIGABYTE G593 series server that is purpose-built for an 8-GPU baseboard. To alleviate the memory bandwidth constraints on AI, including AI inference, the NVIDIA H200 GPU offers a sizable increase in memory capacity and bandwidth compared to the NVIDIA H100 Tensor Core GPU. The H200 GPU has up to 141GB of HBM3e memory and 4.8TB/s of memory bandwidth, translating to a 1.7X increase in memory capacity and 1.4X increase in throughput.

New GIGABYTE GPU Servers
G593-SD1-AAX3: Supporting 5th Gen Intel^® Xeon^® Scalable processors
G593-ZD1-AAX3: Supporting AMD EPYC™ 9004 Series processors with up to 96 CPU cores

Besides the CPU platform, the new GIGABYTE servers share the same overall layout and specifications that are refined for all customer applications in accelerated computing. They will soon be posted in the NVIDIA Qualified System Catalog alongside other qualified GIGABYTE G593 series servers, including four G593 series servers that have passed functional and performance tests and are NVIDIA-Certified Systems™.

The G593 series is built for a high degree of networking and storage. Supporting up to eight U.2 Gen5 NVMe SSDs, the series can process data quickly using NVIDIA GPUDirect^® Storage, one of the NVIDIA Magnum IO™ technologies, for a direct path from GPU memory to storage for improved bandwidth and lower latency. For connecting to other servers in the cluster or for data transfer outside the data center, twelve expansion slots are available for NVIDIA InfiniBand or Ethernet networking via a NIC or DPU. To deliver abundant power and continuous operation, there is a 4+2 3000W redundant power supply configuration.

Integrating into GIGAPOD for a Rack-scale Solution
The importance of a well-designed computing cluster cannot be overstated, especially as AI advances are ramping up and businesses and academia are eager for new insights. Last year, GIGABYTE launched its rack-scale solution, GIGAPOD, which has found great success in deployments using NVIDIA HGX™ systems. So far, these systems have all used the G593 series server to integrate NVIDIA HGX™ H100. With the addition of support for the NVIDIA H200 GPU, GIGAPOD is set to help enable even greater performance and efficiency for accelerated computing workflows.

Launching today, these new GIGABYTE G593 series servers will scale up production in 2H 2024.