AMD’s upcoming Instinct MI300 APUs with next-generation CDNA 3 GPUs and Zen 4 processors will power the El Capitan supercomputer.
El Capitan supercomputer with AMD Instinct MI300 APU processors (Zen 4 processor and CDNA 3 graphics card) with up to 2 exaflops of double precision processing power
According to HPC Wire, at the 79th HPC User Forum at Oak Ridge National Laboratory (ORNL), Lawrence Livermore National Laboratory (LLNL) deputy director for High-Performance Computing, Terri Quinn, revealed that the El Capitan supercomputer that will be deployed at By the end of 2023, it will use AMD’s next-generation Instinct MI300 APU.
The El Capitan supercomputer will use several nodes, each of which is equipped with several Instinct MI300 APU accelerators, which have a socket design and can be operated via an AMD SP5 (LGA 6096) socket. The system is said to offer 10 times the performance of Sierra, which rules the IBM Power9 and NVIDIA Volta GPUs, and has been operating since 2018. In theoretical calculations of FP64 (double precision) the system is to provide a staggering value of two Exaflops of power and will do it below 40 MW.
“We made it public for the first time,” said Quinn, associate director of HPC at LLNL. “I cut these words out of [AMD’s] investors are documenting, and that’s what he says: this is a 3D chiplet design with AMD CDNA3 GPUs, Zen 4 processors, cache and HBM chipsets. ”
“I can’t give you all the specifications, but [El Capitan] is at least 10 times higher than Sierra’s performance in terms of average values, ”said Quinn. “The theoretical peak is two double-precision exaflops, [and we’ll] keep it below 40 megawatts – for the same reason as Oak Ridge’s operating cost.
through the HPC cable
The El Capitan will be designed by HPE and will feature Slingshot-11 interconnects to connect each HPE Cray XE rack together. The AMD Instinct MI300 will pave the way for the future of 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 not only includes next-generation CDNA 3 GPU cores, but also comes with a new-generation Zen 4 core processor.
To enable more than 2 double-precision exaflops of computing power, the U.S. Department of Energy, Lawrence Livermore National Laboratory and HPE have teamed up with AMD to design the El Capitan as the world’s fastest supercomputer with delivery scheduled for early 2023. Capitan will use next-generation products that incorporate enhancements due to the custom CPU design at Frontier.
- Next-generation AMD EPYC processors, codenamed “Genoa”, will feature a “Zen 4” processor core to support next-generation memory and I / O subsystems for AI and HPC workloads
- Next-generation AMD Instinct GPUs based on the new architecture optimized for computing for HPC and AI workloads will use next-generation high-bandwidth memory for optimal deep learning performance
This project will excel in AI data analysis and machine learning to create models that are faster, more accurate, and capable of quantifying the uncertainty of their predictions.
AMD will use a 5nm process node for its Instinct MI300 ‘CDNA 3’ GPUs. The chip will feature the next-generation Infinity cache and will feature the fourth-generation Infinity architecture to support the CXL 3.0 ecosystem. The Instinct MI300 accelerator will rock the APU’s unified memory architecture and new math formats, allowing a 5x increase in performance per watt over the CDNA 2, which is huge. AMD also predicts over 8x faster AI performance over the Instinct MI250X CDNA 2-based accelerators. The CDNA 3 UMAA GPU chip will combine the CPU and GPU with a unified HBM memory package, eliminating redundant memory copies while ensuring a low total cost of ownership .
The AMD Instinct MI300 APU accelerators are expected to be available in late 2023, the same time that the aforementioned El Capitan supercomputer appeared.
ThanksFor the tip!