AMD has revealed three new GPU server accelerators, promising a "dramatic increase" in performance, efficiency, and ease of implementation for deep learning and HPC solutions.
The new line of GPU server accelerators comprises of Radeon Instinct MI25, Radeon Instinct MI8, and Radeon Instinct MI6 and will be paired with AMD’s open ROCm 1.6 software platform.
The Radeon Instinct MI25 accelerator is based on the "Vega" GPU architecture and comes with 14nm FinFET process. AMD says it is "the world’s ultimate training accelerator for large-scale machine intelligence and deep learning datacenter applications."
The Radeon Instinct MI8 accelerator, based on Fiji architecture, is a small form factor HPC and inference accelerator with 8.2 TFLOPS of peak FP16|FP32 performance. AMD says this solution is "well suited" for machine learning inference and HPC applications.
The Radeon Instinct MI6 accelerator is based on the "Polaris" GPU architecture and is a passively cooled inference accelerator with 5.7 TFLOPS of peak FP16|FP32 performance.
These solutions are planned to be shipped to partners (Boxx, Colfax, Exxact Corporation, Gigabyte, Inventec and Supermicro, to name a few) in the third quarter of 2017.
Exact dates, as well as different prices for these solutions, have not yet been mentioned.