Nvidia has overheating issues with Blackwell AI chips

Nvidia has overheating issues with Blackwell AI chips

By admin, Kasım 19, 2024

Nvidia has overheating issues with Blackwell AI chips





Nvidia has overheating issues with Blackwell AI chips




See Full Size


The world’s leading artificial intelligence and high-performance chip manufacturer Nvidianew generation Blackwell experienced in processors overheating He is going through a difficult period due to his problems. According to a report published by The Information, these issues are not only leading to design changes and production delays, but are also raising concerns from major customers such as Google, Meta, and Microsoft.

Overheating and loss of performance

Blackwell GPUs face serious heating issues, especially in high-capacity server racks with 72 processors. of these servers up to 120 kW on a single shelf It is expected to consume as much energy as The heating problem both limits processor performance and creates a risk of damage to components. It is stated that Nvidia had to review the design of its server racks many times to solve these problems.

Nvidia’s suppliers to prevent overheating many design changes instruction It was reported that he gave While engineering revisions were made to improve the cooling systems, this process also caused shipping dates to be postponed. While Nvidia says such changes are normal for large-scale launches in the tech world, customers are concerned about disruptions to production and distribution schedules.




Nvidia has overheating issues with Blackwell AI chips




See Full Size


Problems with Blackwell processors are not limited to overheating. Previously, Nvidia had to postpone production due to efficiency problems encountered in the production of Blackwell GPUs. TSMC’s CoWoS-L packaging technology used in these processors requires extremely precise engineering to ensure the connection between the two chips. However, structural problems caused by thermal expansion differences caused system failures. Nvidia announced that it solved the problem by making various adjustments in the processor design to solve these problems.

Artificial intelligence work may be affected

Nvidia’s delayed shipments Google, Meta ve Microsoft It may affect the plans of large technology companies such as. These companies have advanced technologies such as large language models. artificial intelligence uses Nvidia’s GPUs to train its systems. The delay of Blackwell processors may also disrupt the development and implementation processes of these systems.

In response to the delays and overheating issues, an Nvidia spokesman reminded Reuters of collaborations with cloud providers and described design changes as part of the normal development process. As a result, the latest revision of Blackwell GPUs could only enter mass production at the end of October; This means that Nvidia can ship these processors from the end of January. The original plans were targeting the second quarter of 2024.

























What do you think?

Bir yanıt yazın

E-posta adresiniz yayınlanmayacak. Gerekli alanlar * ile işaretlenmişlerdir