GIGABYTE G593-SD0-LC 8-GPU liquid-cooled AI server. Supports 8× NVIDIA H200/H100/B200 SXM GPUs with direct-to-chip liquid cooling, dual 5th Gen Intel Xeon, 24× DDR5 DIMM slots. 5U rack-mount AI training server with NVSwitch fabric and 900 GB/s GPU-to-GPU interconnect.
5U Rack-Mount
8× SXM (H200/H100/B200)
Dual 5th Gen Intel Xeon
24× DDR5 DIMM (12TB max)
Direct-to-Chip Liquid Cooling
900 GB/s GPU Interconnect
4× 3000W Titanium PSU
8× 400GbE / IB NDR400
The G593-SD0-LC is GIGABYTE's flagship liquid-cooled AI training platform, purpose-built for large-scale LLM training clusters. Direct-to-chip liquid cooling eliminates the need for 15K RPM fans, reducing data center noise from 85 dBA to under 50 dBA while handling 10 kW+ of GPU thermal load per node. Eight H200 SXM GPUs interconnected via NVSwitch deliver 900 GB/s GPU-to-GPU bandwidth, enabling near-linear scaling for trillion-parameter model training. A typical 128-node cluster (1,024 GPUs) achieves >95% scaling efficiency on Llama-3 400B training. QS Compute provides full rack-integration services including CDU sizing, coolant chemistry specification, and on-site commissioning.
Need G593-SD0-LC? Enterprise liquid-cooled AI servers.
Configured, tested, deployed. Contact us for pricing and availability with full rack integration.
Request Quote