Key takeaways
- Nvidia’s Blackwell chip is overheating while linked to servers, which is a follow-up to a previous production delay.
- After multiple unsuccessful attempts to date, the company is now collaborating with vendors to develop a solution for the server design.
- Consumers are worried that it will affect their data center plans.
The latest Blackwell processors from Nvidia are having new issues with the servers they run on, leading to overheating and other issues.
When linked to server racks that can hold up to 72 chips, the sophisticated GPUs are overheating, as The Information first reported, citing people familiar with the problem.
After multiple attempts to identify the root of the issue, the market leader has put pressure on its suppliers, who are responsible for supplying the racks, to provide alternative solutions. Customers are worried that they won’t be able to set up data centers in time for the current deadlines because of the emergency scenario.
Nvidia CEO Jensen Huang told reporters during a trip to Denmark last month that a design problem that had previously delayed the company’s manufacture of the B200 chips has now been rectified.
He acknowledged that a “design flaw caused the yield to be low,” blaming Nvidia “100% of the fault.” However, a remedy was found with the help of manufacturing partner TSMC.
Nvidia Seems to Downplay How Serious the Situation Is
In a recent statement given to Reuters, Nvidia appears to minimize the seriousness of the setback with regard to the overheating issue.
The business stated that “the engineering iterations are normal and expected” and that it still collaborates with top cloud service providers as a crucial component of its operations.
Nvidia’s cutting-edge Blackwell chip combines two silicon squares the size of their previous product into a single unit that can complete tasks like chatbot responses up to 30 times faster.
At the forefront of AI development, they are a highly sought-after product.
