“100% Nvidia’s fault”: Jensen Huang admits Blackwell AI chips had a concerning design flaw
CEO admits AI chip design flaw was ‘100% Nvidia’s fault’
When you purchase through links on our site, we may earn an affiliate commission.Here’s how it works.
NvidiaCEO Jensen Huang has confirmed a design flaw in its top-endBlackwellAI chips which had affected production was an entirely internal problem, which has now been fixed.
“We had a design flaw in Blackwell,” Huang said at an event in Copenhagen,Reutersreported. “It was functional, but the design flaw caused the yield to be low. It was 100% Nvidia’s fault.”
First identified in August 2024, the delay to Blackwell B100/B200 processors had raised eyebrows around the world, but Huang reassured that it was Nvidia’s own doing that caused the issue.
Blackwell delays
Blackwell chips have been in high demand since Nvidia unveiled the platform earlier in 2024, with Huang describing it as,“the world’s most powerful chip,“offering previously unheard-of levels of AI computing power.
Set to begin shipping in the latter part of 2024, Blackwell binds together two GPU dies, which are connected by 10 TB/second chip-to-chip link into a single, unified GPU. This uses using TSMC’s CoWoS-L packaging technology, which relies on an RDL interposer equipped with local silicon interconnect (LSI) bridges that need to be located specifically to allow fast data transfer - the misalignment of which resulted in the issue.
Initial media reports had claimed the issue had caused friction with manufacturing partner TSMC, but Huang dismissed the claims as “fake news”.
“In order to make a Blackwell computer work, seven different types of chips were designed from scratch and had to be ramped into production at the same time,” he said.
Are you a pro? Subscribe to our newsletter
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
“What TSMC did, was to help us recover from that yield difficulty and resume the manufacturing of Blackwell at an incredible pace.”
Blackwell is set to be up to 30x faster than its Grace Hopper predecessor when it comes to AI inference tasks, whilst also reducing cost and energy consumption by up to 25x.
More from TechRadar Pro
Mike Moore is Deputy Editor at TechRadar Pro. He has worked as a B2B and B2C tech journalist for nearly a decade, including at one of the UK’s leading national newspapers and fellow Future title ITProPortal, and when he’s not keeping track of all the latest enterprise and workplace trends, can most likely be found watching, following or taking part in some kind of sport.
This new malware utilizes a rare programming language to evade traditional detection methods
Google puts Nvidia on high alert as it showcases Trillium, its rival AI chip, while promising to bring H200 Tensor Core GPUs within days
Arcane season 2 confirms the hit series isn’t just one of the best Netflix shows ever made – it’s an animated legend that’ll stand the test of time
