AMD today received a new overcalculation order, combining HPE’s Cray to build El Capitan Supercomputing for the U.S. Department of Energy, with a $600 million budget, will use AMD’s next-generation CPU and Radeon accelerator cards, which will be available in 2023 and will be available 20 billion times. What exact specifications does El Capitan specify? With more than 3 years left, cpu/GPU details are unknown, and it’s certain that the third generation Ofafabric bus is the third generation, with four Radeon Instinct acceleration cards per EPYC CPU.
For the super-computing CPU, the more reliable argument is zen4 architecture of the next generation EPYC, according to previous speculation, Zen4 architecture will be a new upgrade, the platform will also be upgraded to DDR5, PCIe 5.0, these advanced technologies corresponding to the billion-dollar era, in terms of performance and energy efficiency to meet the requirements.
In the EPYC server, the product corresponding to the Zen4 architecture is Genoa Genoa, the Italian city, and now the EPYC code Rome, the two will also have Zen3 architecture of Milan Milan, the basic specifications are similar to the current Rome II, but also DDR4 memory, PCIe 4.0 bus, SP3 package interface, continue to maintain platform compatibility, it seems to focus on architecture optimization, performance improvement.
Now the biggest mystery is the Zen4 and Genoa processor architecture and technology, whether to continue to adhere to the maximum 64 core or upgrade to 128 core, is the use of 5nm or the recent news that continue to use 7nm sanding, these two improvements affect the A rice and the heart of the industry.
I believe everyone is looking forward to AMD on the 5nm process to build 128 core 256 thread processor, in high-performance computing continue to ride the dust.
Finally, forrest Norrod, amd’s general manager of data center and embedded business, said AMD will announce the Zen4/Genoa processor at an upcoming analyst meeting, which is Thursday, probably early tomorrow morning.