AI Increases Data Center Rack Demands
GPU-based AI servers are deeper and heavier than CPU-based models, requiring more power and space than standard racks can provide.
October 29, 2024
Channel partners selling into data center applications have undoubtedly noticed a growing trend — and by growing, we mean literally. As customers increasingly deploy larger GPU servers to accommodate the rapid expansion of generative AI (GenAI), the escalating computing demands are transforming the landscape of today's enterprise, co-lo and hyperscale data centers.
Why Your Racks May Soon Fall Short
Because GPU-based AI servers are deeper and heavier than CPU-based models ― on average, 44 inches versus 30 inches ― they require more power and space than standard racks can provide. And it’s not just the depth that has increased; GPU-based machines are exceeding 100 pounds or more per 1U of rack space, compared to their 20- to 40-pound CPU counterparts. As a result, AI-inspired servers are tipping the scales beyond a standard rack’s weight capacity long before the enclosure runs out of space.
If you haven’t yet encountered customers with equipment racks bursting at the seams, chances are you will soon. A recent survey of technology leaders at 1,000 companies around the world found that 96% of respondents plan to expand their AI infrastructure in 2024, with 40% intending to buy more GPU machines.
The Rack Revolution: Eaton’s Answer to AI
In response to these changing requirements, Eaton has introduced a new line of Heavy-Duty SmartRack (SRH) floor-standing enclosures. With extra-deep frames capable of accommodating up to 54 inches of usable depth and the ability to hold up to 5,000 pounds, this new rack has been specially designed to support the new generation of AI servers.
The only enclosure on the market with the extreme depth and weight capacity needed for high-density GPU-based servers, the Heavy-Duty SmartRack’s reinforced structure features thicker steel, more welds and extra supports. Most competing models, on the other hand, simply attempt to patch over existing designs ― a Band-Aid approach that offers inadequate support for current AI servers and even less suitability for future applications. Available from stock in 18 different models ― in 42U, 45U or 48U sizes ― Eaton’s new enclosures enable your customers to future-proof their investment and accommodate ever-changing specifications of IT equipment. Custom configurations can also be quickly assembled.
Brains and Brawn: SmartRack’s Total Package
And if you think the Heavy-Duty SmartRack is all brawn and no brains, think again. Your data center customers will appreciate the array of additional thoughtful benefits afforded by the enclosure, including ample space to house the high-density cabling, cable managers and multiple rack PDUs needed to meet the increased power requirements of GPUs. Even more, the rack is designed to enhance airflow and provide space for next-gen liquid cooling solutions, since GPU servers run hotter than CPUs.
The value of the Heavy-Duty SmartRack is enhanced by its white powder coat finish, which reflects more than 80% of available light in the data center, compared to less than 5% for black racks, according to NYMAS Group. This superior light reflection creates a better environment for improved contrast and visibility, helping technicians find equipment quickly and prevent connection errors.
Interim Solutions for AI Growing Pains
To help you provide customers who are experiencing AI growth pains with an interim solution, Eaton recently tested and recertified the standard SR enclosure, boosting its weight capacity from 3,000 to 3,500 pounds. While this recertification may help you buy them some time, most customers supporting AI will need to migrate to the more robust rack in the near future.
Craig Watkins is Director of Product Management, Eaton.
This guest blog is part of a Channel Futures sponsorship.
You May Also Like