Nvidia has unveiled plans to carry RTX Professional 6000 Blackwell Server Version GPUs to plain enterprise servers later this 12 months. This transfer will make Blackwell accessible to a broader set of enterprises for AI and different conventional workloads.
On the annual SIGGRAPH convention in Vancouver, Canada, this week, executives mentioned Cisco, Dell Applied sciences, HPE, Lenovo, Supermicro and others will make the GPU out there of their hottest 2U enterprise server techniques.
Beneath the Hood
The 2U Nvidia RTX Professional Servers are x86-based, air-cooled, PCIe-based {hardware} designed to slot in normal enterprise information middle racks and meet enterprise IT necessities, mentioned Rev Lebaredian, Nvidia’s vice chairman of omniverse and simulation applied sciences.
“Enterprise information facilities have been constructed for low-power, predictable workloads, counting on air-cooled servers in power-constrained racks. AI has shattered this mannequin. It calls for a brand new stage of computational depth,” he mentioned. “Nvidia RTX Professional Server makes the transition from general-purpose clusters to an AI manufacturing unit seamless with no full information middle overhaul.”
Nvidia first announced the newest RTX Professional servers, powered by the RTX Professional 6000 Blackwell Server Version GPU, on the Computex present in Taiwan in Might. Whereas server distributors initially provided 4U techniques, as we speak’s announcement extends the lineup to 2U type elements.
‘Blackwell for the Lots’
Analyst Matt Kimball of Moor Insights & Technique mentioned the brand new Blackwell GPU-powered Nvidia RTX Professional Servers are a strong choice for conventional enterprises that need to do AI coaching, tuning or inferencing on-premises, in addition to digital desktop infrastructure, information analytics or different accelerated workloads.
Most individuals affiliate Nvidia with high-end GPUs which are mixed by the hundreds to coach massive basis fashions, he mentioned. However this announcement focuses on enabling conventional enterprises to undertake AI in a sensible vogue.
“That is just like the Blackwell for the lots,” Kimball mentioned. “This RTX is a right-sized GPU for the enterprise buyer, and by enterprise, I don’t imply the most important clients. I imply the 5,000-employee, 1,000-server store that should deploy agentic AI to make their manufacturing and provide chain way more environment friendly.”
Kimball mentioned Nvidia’s push to offer enterprises with the {hardware} to undertake AI and different accelerated workloads is a brilliant technique. “It’s actually a slick play as a result of it’s getting out in entrance of that rush for the industrial enterprise,” he mentioned. “Nvidia goes on to ‘Principal Road’, if you’ll.”
The technique additionally provides Nvidia a aggressive benefit as AMD doesn’t at the moment have a competitor to the RTX Professional 6000 Blackwell Server Version GPU, Kimball mentioned. AMD at the moment competes towards Nvidia on the excessive finish of the GPU market.
Nvidia RTX Professional Server Efficiency
As enterprises purchase hundreds of thousands of servers yearly for enterprise workloads, they will refresh their CPU-only techniques with these mainstream GPU-powered, 2U servers, Nvidia executives mentioned.
The brand new 2U Nvidia RTX Professional Servers ship as much as 45 occasions higher efficiency for conventional CPU-based workloads like information analytics, simulation, video processing and graphics rendering, leading to 18 occasions greater power effectivity and decrease price of possession when in comparison with CPU-only 2U techniques, the corporate mentioned.
“This implies enterprises can consolidate tons of of CPU techniques to only a handful of RTX Professional Servers,” Lebaredian mentioned.
The brand new RTX Professional Servers, which assist Nvidia AI Enterprise and Omniverse software program and can be out there starting this fall, additionally present a robust, scalable platform for AI workloads, together with agentic and bodily AI use circumstances, he mentioned.
In comparison with enterprise servers operating Nvidia’s earlier era L40S GPUs, the brand new techniques velocity LLM inferencing by six occasions and carry out real-time rendering and artificial information era by as much as 4 occasions sooner, Lebaredian mentioned.
Nvidia’s Further SIGGRAPH Bulletins
On the SIGGRAPH convention this week, Nvidia additionally introduced:
-
New additions to the Nemotron mannequin household, which allow AI brokers to purpose higher and make smarter choices.
-
New Omniverse SDKs and libraries at the moment are out there for constructing and deploying industrial AI and robotics simulation functions.
-
New Cosmos World Basis Fashions, which allow sooner artificial world era, and bodily AI reasoning for information curation, annotation, robotics, and visible brokers.
-
New Nvidia Blackwell desktop GPUs for workstations.
