At the SIGGRAPH conference this week, Nvidia announced its RTX Pro 6000 Blackwell Server Edition GPU is making its way to slimmer 2U rack mount servers from several of its global system partners through its Enterprise AI factory-validated design initiative. Previously, customers who wanted to deploy these cards in their own clouds needed to make room for 4U servers with higher cooling and power demands. Now, these cards are coming to one of the most popular rack-mount form factors used by on-premise data centers.
Cisco, Dell, HPE, Lenovo, and Supermicro will offer 2U, x86-powered servers featuring two RTX Pro 6000 Blackwell GPUs in various configurations. For just one example, Dell announced a new 2U server at SIGGRAPH: the Dell PowerEdge R7725 2U, sporting two RTX Pro 6000 GPUs built with Nvidia's AI Data platform.
These new 2U servers join existing 4U, 6U, and 8U RTX Pro servers, designed to house Nvidia's RTX 6000 Blackwell GPU in its various forms. These servers provide all the functionality needed to accelerate AI workflows with Nvidia hardware, including the integration of Nvidia's BlueField-3 DPUs and ConnectX-8 SuperNICs with built-in PCIe Gen 6 switches.
Nvidia claims 2U RTX Pro servers will provide up to 45x better performance and 18x higher energy efficiency compared to CPU-only 2U systems, due to transitioning from CPU-only AI processing to GPU-accelerated AI processing. The company expects that performance will lead to extensive consolidation for customers looking to condense their server fleets into denser, higher-performance systems.
The RTX 6000 Pro Server Edition is Nvidia's latest datacenter GPU, and is essentially the fully enabled version of Nvidia's desktop RTX 5090. The RTX 6000 Pro features 24,064 CUDA cores, 600W TBP, and a 512-bit memory interface paired with 96GB of 28Gbps GDDR7 memory.
For servers, the RTX 6000 Pro Server Edition is a much more affordable alternative to Nvidia's fire-breathing B200 and B300 Blackwell GPUs. Contrary to those GPUs, the RTX 6000 Pro Server Edition is air-cooled, which significantly reduces server costs to support AI GPUs. The B200 and B300 consume enough power to virtually require liquid cooling, which is expensive and complex to deploy.
RTX 6000 Pro Server Edition GPU support with 2U rack mounts provides enterprise and datacenter customers with the most compact methods of deploying AI GPUs, perfect for onsite edge computing deployments. 2U mainstream RTX Pro servers will allegedly be available starting later this year.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.
11 hours ago
5
English (US) ·