TSMC, CadenceKLA, Siemens and Synopsys are advancing semiconductor manufacturing by adopting the NVIDIA CUDA-X and NVIDIA Blackwell platforms.
NVIDIA Blackwell GPUs, NVIDIA Grace CPUshigh-speed NVIDIA NVLink materials and switches, and domain-specific NVIDIA CUDA-X libraries like NVIDIA cuDSS and NVIDIA cuLitho are bettering computational lithography and gadget simulation for superior chip manufacturing.
“Our collaboration with NVIDIA represents a major development in semiconductor course of simulation,” stated Jeff Wu, fellow and director for the expertise computer-aided design division at TSMC. “The computational acceleration from CUDA-X libraries and NVIDIA Grace Blackwell will expedite course of improvement by simulating complicated manufacturing processes and gadget behaviors at decrease price.”
NVIDIA cuLitho and Blackwell pace up lithography by as much as 25x. GPU acceleration permits main lithography suppliers and semiconductor producers equivalent to TSMC to foretell and proper lithography points earlier than manufacturing at an unprecedented pace.
Earlier this month, digital design automation (EDA) software program and companies supplier Cadence introduced its Millennium M2000 platform, constructed solely on NVIDIA Blackwell for the EDA market. The M2000 is a scalable turnkey answer for deploying NVIDIA Grace Blackwell and CUDA-X libraries with a totally accelerated portfolio of Cadence design instruments.
Cadence can be one of many first to undertake NVIDIA NVLink Fusion, enabling customized silicon scale-up to fulfill the necessities of demanding workloads for mannequin coaching and agentic AI inference. By adopting NVLink Fusion, Cadence permits hyperscalers to optimize and validate throughout your entire design spectrum.
This month, Cadence introduced the Millennium M2000 AI Supercomputer to rework silicon, system and drug design. Primarily based on the NVIDIA Blackwell platform, choices embrace the NVIDIA GB200 NVL72 system for tackling huge system-on-a-chip, 3D-IC, and subsystem implementation and signoff utilizing Cadence Cerebrus AI Studio and Cadence multiphysics system evaluation instruments, in addition to the brand new NVIDIA RTX PRO 6000 Blackwell Server Version GPU for smaller chip designs and simulations.
“Our collaboration with NVIDIA has at all times been about pushing the boundaries of what’s attainable in each digital design automation and system design and evaluation,” stated Michael Jackson, company vp and normal supervisor of the system design and evaluation group at Cadence. “The Millennium M2000 platform, constructed solely on NVIDIA Blackwell, isn’t nearly quicker simulation — it’s about redefining the infrastructure for AI-driven innovation, enabling what was beforehand not possible.”
Siemens is harnessing the parallel processing energy of the NVIDIA CUDA-X libraries and the groundbreaking efficiency of the Grace Blackwell platform to considerably speed up its Calibre platform.
This integration permits unprecedented pace and accuracy in essential semiconductor manufacturing steps, together with optical proximity correction with nanometer precision, complete bodily verification, sturdy design for manufacturability evaluation, thorough reliability verification and seamless integration and automation throughout the design-to-manufacturing movement.
“Leveraging NVIDIA CUDA-X and Grace Blackwell in our Calibre platform permits quicker, extra environment friendly optical proximity correction with out sacrificing accuracy for superior semiconductor nodes,” stated Mike Ellow, CEO of Siemens EDA. “That is particularly vital as chip complexity continues to develop.”
Moreover, Synopsys, a number one EDA software program and companies supplier, is utilizing NVIDIA CUDA-X libraries and Blackwell for its EDA instruments, together with Synopsys PrimeSim, Proteus, S-Litho, Sentaurus Machine and QuantumATK. By integrating with CUDA-X libraries, Synopsys achieved new benchmark outcomes for Sentaurus Machine, QuantumATK, and S-Litho on the NVIDIA B200, demonstrating a 12x, 15x and 20x scale-up, respectively, versus comparable CPU infrastructure.
As well as, Synopsys just lately introduced at NVIDIA GTC that they venture Synopsys PrimeSim to run 30x quicker and Synopsys Proteus to run 20x quicker on NVIDIA Blackwell platforms.
“Synopsys has a protracted historical past of collaborating with NVIDIA on accelerating our EDA options to maximise the capabilities of engineering groups. Constructing on our industry-first strategy, Synopsys is leveraging NVIDIA’s Blackwell structure throughout our TCAD, computational lithography and atomistic simulation merchandise to unlock unprecedented efficiency positive aspects,” stated Sanjay Bali, senior vp of technique and product administration at Synopsys. “By integrating NVIDIA’s CUDA-X libraries and Blackwell structure into our industry-leading simulation solvers, we’ve achieved transformative speedups and redefined how EDA is enabling semiconductor manufacturing innovation.”
Semiconductor course of management gear producer KLA and NVIDIA have labored collectively for over a decade to advance KLA’s physics-based AI with optimized high-performance computing options that faucet into GPUs and the CUDA ecosystem.
The worth of course of management in semiconductor manufacturing is rising on account of AI-driven traits, equivalent to extra complicated designs, accelerated product cycles, larger worth wafer volumes and rising superior packaging demand. KLA’s industry-leading inspection and metrology programs seize and course of pictures by operating complicated AI algorithms to search out probably the most essential semiconductor defects at lightning-fast speeds.
KLA is trying ahead to evaluating the NVIDIA RTX PRO 6000 Blackwell Server Version with CUDA-X libraries for sure markets to additional speed up inference workloads powering the semiconductor chip manufacturing course of.
By embedding NVIDIA Blackwell into EDA, manufacturing and course of management, NVIDIA helps the semiconductor {industry} ship the subsequent era of high-performance chips quicker.
Study extra in regards to the newest AI developments at NVIDIA GTC Taipeioperating Could 21-22 at COMPUTEX.
