RTX 6000 Pro Explained: Bringing Blackwell to your Business
Allan Kaye, CEO and co-founder at Vespertec
By now, digital revolutions are NVIDIA’s day-to-day. Each new chip announcement promises to upend high-performance computing, and then does just that. So it was for Grace-Hopper architecture, and so it will be for Vera Rubin and later Feynman architecture announced at this year’s GTC in San Francisco once those are rolled out in earnest.
Right now, we’re living through the Blackwell era. From complex, data-heavy research to use in financial services, NVIDIA’s Blackwell architecture has accelerated the rollout of AI and HPC applications across industries.
However, as with every new technology in the fast-moving world of HPC, there are different challenges. Some will not wish to commit to an entire 8-way, NVLink HGX, or superchip-based system off the bat. Others are simply eager to supplement existing systems.
This is where the RTX 6000 Pro Blackwell Server Edition comes in. This PCIe card-based system is compact, powerful, and fits into many servers that were initially designed to support NVIDIA’s H200 cards, whilst also maintaining the same 600W power profile. In short, it’s the perfect way for teams to test-run a single Blackwell GPU without having to overhaul existing environments, whatever the specific use case might be.
Each enterprise will have unique needs, and testing hardware out yourself is often just not an option when delays can cost teams valuable time.
Fortunately, the Vesper team’s been shipping the RTX 6000 Blackwell since day one. We’ve had the chance to gain hands-on experience with the system, and I thought it’s about time we shared what we’ve learned so far – a kind of ‘cheat sheet’ for anyone seeking to deploy 4 and 8-way solutions for this GPU:
Option 1:
Lenovo’s SR675 V3 is a 3U 4-GPU server with hybrid closed-loop liquid cooling. One customer told us it’s among the most power-efficient Blackwell-class systems they’ve seen.
—
On top of these, we also recently identified three new Supermicro systems: the 4U SYS-422GL-NR, the 3U SYS-322GA-NR, and the 2U SYS-212GB-NR, which each suit slightly different deployment scales and NVMe configurations.
This is by no means an exhaustive list. We’re testing and shipping new configurations all the time, but the above selection should give a handy starting point for businesses looking to get started with the RTX-6000 Blackwell Pro.
Knowing the info above is half the battle – but only half. Each of these systems is already validated for TX 6000 Blackwell deployment, but every use case is different. Thermals, space constraints, and the precise workloads are all equally vital considerations, and any one of those can drastically affect which setup suits your business needs.
If you’d like to take those next steps and find out exactly how the RTX 6000 Blackwell Pro can help level up your HPC architecture without the need for an overhaul, or test the system out in Vesper’s AI Lab environment, drop me a line at info@vespertec.com or give us a ring at +44 161 947 4321.
