Nvidia video cards will allow TSMC to dramatically accelerate the creation of new generation chips

The world’s largest contract semiconductor manufacturer TSMC will begin to actively use the Nvidia cuLitho computational lithography platform in production – it will help speed up product release and push the physical limits for advanced next-generation chips, Nvidia said.

Image source: tsmc.com

Computational lithography is a critical step in chip manufacturing, necessary when moving electronic circuit designs to silicon. This process requires complex computations: electromagnetic physics, photochemistry, computational geometry, iterative optimization and distributed computing. Semiconductor manufacturing uses data centers to perform these calculations, but this stage still remains a bottleneck in bringing new processes and computer architectures to market.

Computational lithography is the most resource-intensive workload in the entire semiconductor design and manufacturing process. A typical set of masks to print a chip can take 30 million or more hours of CPU time. But 350 systems based on Nvidia H100 accelerators can replace 40,000 CPU systems—helping reduce production time, cost, space and power consumption. Nvidia has previously said that its system can handle jobs that previously took two weeks in one night. Nvidia’s cuLitho library is responsible for deploying accelerated computing in the field of computational lithography. In TSMC’s case, it will help accelerate the development of advanced semiconductor components.

Nvidia has also developed methods for applying generative artificial intelligence to the cuLitho platform, which provides an additional 2x acceleration in system performance. Generative AI helps in creating a near-perfect inverse mask that accounts for light diffraction in computational lithography; accelerated computing and AI are also responsible for optical proximity correction (OPC). Together, they make it possible to more accurately model physics and implement mathematical methods that were previously considered extremely resource-intensive. As a result, the time required to create each mask at the factory is radically reduced, which means the development cycle time for a new technological unit is also reduced.

Calculations that were previously impractical are also possible. Inverse lithography techniques have been described in the scientific literature for two decades, but their precise implementation on a full-chip scale has been largely ruled out because the computations involved are too time-consuming. Nvidia cuLitho helps solve this problem – advanced factories will be able to use it to create powerful next-generation chips.

admin

Share
Published by
admin

Recent Posts

Amazon launched a Mass Effect series, and BioWare added armor to Dragon Age: The Veilguard in honor of N7 Day

BioWare warned that due to the release of Dragon Age: The Veilguard there would be…

51 mins ago

Arcane became the most expensive animated series in history by a wide margin.

On the eve of the start of the second season of the animated series Arcane…

2 hours ago

Sony has priced the PlayStation 5 Pro 30th Anniversary in the style of the first PlayStation at $1000

Sony today began sales of the PlayStation 5 Pro game console, and also announced the…

2 hours ago

The first portable set-top box with Ryzen AI 300 goes on sale for $1099–1699 – you will be given a medal for its purchase

One-Netbook has announced the start of accepting pre-orders for the OneXFly F1 Pro portable gaming…

2 hours ago

Nvidia will limit monthly gaming time in the GeForce Now cloud service – you will have to pay for additional hours

Nvidia has announced that starting next year it will change its subscription policy for its…

3 hours ago

Old school steampunk role-playing game New Arc Line will be released soon, but not completely – release date and early access announcement

Fulqrum Publishing (formerly 1C Publishing) and developer Dreamate have announced the release date for the…

4 hours ago