While the Nvidia GeForce RTX 3090 may be the most powerful gaming graphics card you can buy today, chipmaker AMD has announced a new GPU accelerator that it claims to be the ultimate high-performance computing GPU. (HPC) fastest in the world. Scoring 11,54 teraflops in AMD internal testing, the Instinct MI100 is the first x86 server GPU to break the 10 teraflop performance barrier. The Santa Clara-based tech giant is launching the new chip for scientific research, highlighting its ability to handle heavy workloads in scientific computing.
General culture
In tests conducted by its performance labs, the AMD Instinct MI100 with 120 compute units delivered FP64 peak performance of 11,54 teraflops and FP23,1 peak performance 32 teraflops. This makes it several times faster than previous AMD accelerators. The new GPU accelerator is based on AMD CDNA architecture and is designed to work with AMD EPYC server processors. Designed to support the latest PCIe 4.0 specification, it features 32 GB of high-bandwidth HBM2 memory clocked at 1,2 GHz and offers 1,23 TB/s of memory bandwidth. AMD also notes that the new chip will allow developers to build high-performance applications using its ROCm 4.0 platform, which includes compilers, programming APIs, and libraries. He further adds that the company is working with several manufacturers, including Dell, Hewlett Packard Enterprise, Gigabyte and Supermicro, and that servers with the new chip are expected to be on the market by the end of the year.