New AMD Radeon Instinct line-up with 7nm Vega GPU released

New AMD Radeon Instinct line-up with 7nm Vega GPU released

From the article: “the double-peak precision (FP64) of the MI60 is 7.4 TFLOPs, whereas the MI25 managed 768 gigaflops”, this is almost a 1000% increase in FP64 computation in respect to previous generation, hell…this is SERIOUS innovation, for comparison Nvidia RTX 2080Ti can reach only 420.2 GFlops in FP64, less than AMD’s previous generation MI25 !!
Nvidia is going to be thrown in the dustbin and completely out of the game from the HPC market and applications such as deep learning, science, financial services and other extremely high-performance areas.
Just imagine how fast these cards can do Ray Trace rendering, that is almost completely floating point based…

2 Likes

Application which need FP64 are very rare… the trend is the opposite : reduce precision for more parameters or iteration.