Report: New Nvidia AI Chips Struggle with Overheating in Servers
The latest Blackwell AI chips from Nvidia are experiencing challenges with overheating servers, compounding the earlier delays they have encountered.
The Information reported on Sunday that Nvidia’s latest Blackwell AI chips, despite experiencing prior delays, are now grappling with server issues, specifically overheating problems. This development has raised concerns among customers regarding the timely setup of new data centers.
Sources familiar with the matter reported that the Blackwell graphics processing units experience overheating when linked in the tailor-made server racks crafted by the company.
Nvidia employees, along with customers and suppliers familiar with the situation, stated that the AI chip manufacturer has requested multiple redesigns of the racks to address overheating issues. The suppliers involved were not disclosed in the report.
Nvidia failed to provide a timely response to a comment request made outside of normal business hours.
In March, Nvidia introduced the Blackwell chips, originally planned for shipping in the second quarter. However, delays have arisen, posing potential impacts on clients like Meta Platforms, Google from Alphabet, and Microsoft.
Nvidia’s latest Blackwell chip combines two silicon squares, each the size of its previous chip, into a single unit that boasts a 30-fold increase in speed for tasks such as generating responses from chatbots.
