- 96-core ThreadRipper was paired with two Nvidia H100 NVL GPU with 188GB reminiscence
- Final month, Comino introduced it was taking preorders for Nvidia’s subsequent technology GPU
- Comino unveiled an AMD workstation with six RTX 4090 GPU in early 2024
Comino, recognized for its liquid-cooled programs, just lately launched the Grando H100 Server, a high-performance answer tailor-made for AI and HPC workloads.
This new mannequin options an AMD Threadripper PRO 7995WX processor, boasting 96 cores and 192 threads. Primarily based on AMD’s Zen 4 structure, the CPU helps as much as 2TB of DDR5 reminiscence, delivering beneficiant bandwidth for large-scale information processing and computational duties.
The processor, on this configuration, is paired with twin Nvidia H100 NVL GPUs, geared up with 94GB of reminiscence every, for a whopping 188GB of GPU reminiscence. The system is constructed across the ASUS SAGE WRX90 motherboard, which offers ample PCIe Gen5 lanes for GPUs and high-speed storage.
A superb alternative for enterprises and professionals
To take care of optimum thermal efficiency, the Grando H100 incorporates Comino’s superior liquid cooling system. This encompasses a centralized water block and quick-disconnect fittings, simplifying upkeep and minimizing the chance of leaks.
This environment friendly cooling design not solely ensures stability underneath heavy workloads but additionally operates with decreased noise. For enterprise reliability, the server contains 4 1600W redundant energy provides.
StorageReview, which tested Comino’s Grando Server and a workstation with six RTX 4090 GPUs earlier this 12 months, describing them as “prime examples of GPU-driven efficiency programs completed proper,” benchmarked the new server and got here away with blended outcomes.
As you’d anticipate, the Grando H100 excelled in AI-driven computational workflows and CPU-intensive duties. In checks like Blender’s multi-core rendering, 7-Zip compression, and Y-Cruncher, it simply outperformed competing programs. Nonetheless, the system’s efficiency in conventional rendering benchmarks, akin to Luxmark and OctaneBench, noticed the server lagging behind programs geared up with GPUs that are optimized for these duties.
Summing up its findings, StorageReview famous, “General, the Comino Grando H100 is a wonderful alternative for enterprises and professionals prioritizing AI optimization, computational duties, and reliability in demanding environments. Its distinctive design and cooling improvements provide flexibility and efficiency for AI-driven workloads. Nonetheless, different configurations just like the RTX 4090-equipped Grando Server or RTX 6000 Ada-powered programs could also be extra appropriate for customers centered on conventional GPU rendering.”
We’re nonetheless eagerly awaiting a evaluate of the Grando Server equipped with eight Nvidia RTX 5090 GPUs, which Comino introduced beforehand and is taking pre-orders for, however sadly there may be nonetheless no signal of any benchmarks for that individual powerhouse.
