AMD is coming Instinct MI300 APUs with next-gen CDNA 3 GPU & Zen 4 CPUs cores will power the El Capitan Supercomputer.
El Capitan supercomputer with AMD’s Instinct MI300 APUs (Zen 4 CPU & CDNA 3 GPU) for up to 2 exaflops of double precision computing power
According to HPC Wire, at the 79th HPC User Forum at Oak Ridge National Laboratory (ORNL), Terri Quinn revealed the deputy Associate Director of High-Performance Computing at Lawrence Livermore National Laboratory (LLNL) that the El Capitan Supercomputer, which will be deployed in late 2023, will use AMD’s next-gen Instinct MI300 APUs.
The El Capitan Supercomputer will use several nodes, each equipped with different Instinct MI300 APU accelerators that have a socketed design and can be accommodated by AMD’s SP5 (LGA 6096) socket. The system is said to offer 10x better performance than the Sierra, which rocks IBM’s Power9 CPUs and NVIDIA’s Volta GPUs and has been operational since 2018. In theoretical FP64 (double precision) computing power, the system is expected to deliver a mind-boggling two Exaflops. of horsepower and will do so under 40MW.
“It’s the first time we’ve said this publicly,” said Quinn, associate director for HPC at LLNL. “I finished these words [AMD’s] investors and that’s what it says: it’s a 3D chiplet design with AMD CDNA3 GPUs, Zen 4 CPUs, cache and HBM chiplets.”
“I can’t give you all the specs, but [El Capitan] is at least 10x greater than Sierra’s performance on average earnings,” said Quinn. “Theoretical peak is two double precision exaflops, [and we’ll] keep it under 40 megawatts – same reason as Oak Ridge, operating costs.
Designed by HPE, El Capitan features Slingshot-11 interconnects to link any HPE Cray XE rack together. The AMD Instinct MI300 will pave the way for future Exascale-class APUs. At AMD Financial Day 2022, the company confirmed that the MI300 will be a multi-chip and multi-IP Instinct accelerator that will not only be equipped with the next-generation CDNA 3 GPU cores, but also equipped with the next-generation Zen. 4 CPU cores.
To enable more than 2 exaflops of double-precision processing power, the U.S. Department of Energy, Lawrence Livermore National Laboratory and HPE teamed up with AMD to design El Capitan, expected to be the world’s fastest supercomputer, with delivery expected early. 2023. El Capitan Capitan will leverage next-generation products that include improvements to Frontier’s custom processor design.
- Next-generation AMD EPYC processors, codenamed “Genoa”, will feature the “Zen 4” processor core to support next-generation memory and I/O subsystems for AI and HPC workloads
- AMD Instinct next-generation GPUs, based on a new computer-optimized architecture for HPC and AI workloads, use next-generation high-bandwidth memory for optimal deep learning performance
This design will excel in AI and machine learning data analytics to create models that are faster, more accurate and able to quantify the uncertainty of their predictions.
AMD will use the 5nm process node for its Instinct MI300 ‘CDNA 3’ GPUs. The chip will be equipped with the next-generation Infinity Cache and features the 4th-generation Infinity architecture that enables support for the CXL 3.0 ecosystem. The Instinct MI300 accelerator will rock a unified memory APU architecture and new math formats, achieving 5x performance per watt over CDNA 2, which is massive. AMD also projects over 8x the AI performance over the CDNA 2-based Instinct MI250X accelerators. The CDNA 3 GPU’s UMAA connects the CPU and GPU to a unified HBM memory package, eliminating redundant memory copies and delivering low TCO.
AMD’s Instinct MI300 APU accelerators are expected to be available in late 2023, which coincides with the implementation of the aforementioned El Capitan supercomputer.
Thanks toFor the tip!