Intel’s supercomputer Aurora will finally be released in 2025 instead of the previously projected 2023 date. The Aurora exascale flagship project will debut at the June 2023 list of the top 500 fastest supercomputers globally. At the International Supercomputer Conference 23 in Hamburg, Intel showcased its Intel Xeon Datacenter CPU Max and the Datacenter GPU Max in separate demos to substantiate their existence and specifications.
The accelerator formerly named Ponte Vecchio, which will power Aurora, will be available from June, while the Data Center GPU Max 1100 is set to be released in August. The larger GPU relatives, the 1450 and 1550, will only be available via the Intel Developer Cloud.
Intel has announced further details regarding the Aurora supercomputer specifications. The system will have over 60,000 blades in total, with Aurora getting 10,624. The supercomputer will have 21,248 CPUs and 63,744 GPUs with 10.9 petabytes of DDR5 memory, 1.36 petabytes of HBM stack memory, and 8.16 petabytes of HBM memory in the GPUs.
Formerly known as Falcon Shores, Intel’s upcoming product will now be a “2025 product.” The model brings together x86 cores and an Xe GPU in one processor socket, offering more than five times the performance per watt, five times the compute density, and five times the memory capacity and bandwidth. According to Jeff McVeigh, Intel’s Corporate Vice President and General Manager of the Super Compute Group, criteria such as generative AI and potential in Ponte Vecchio are the reasons for the Falcon Shores postponement.
