According to Nvidia CEO Jensen Huang, the amount of computation necessary to run artificial intelligence (AI) is 1,000 times higher than the computing power needed to run non-AI software.
While traditional datacentre racks generate 20 to 40 kilowatts per rack, Ryan Hotchkin, senior director of datacentre and business management operations for SHI, says Nvidia is doubling that every year.
“With the GB200s and NVL72s we’re looking at 120 kilowatts per rack. More powerful graphic processing units [GPUs] equates to more demand for power distribution units [PDUs], and that has a knock-on effect on the electrical infrastructure,” he adds.
View the full article here.