According to some interesting news about a pending
supercomputer update it seems as if there’s something
fast in the pipe of NVidia’s GPU releases. The University
of Indiana has delayed their second upgrade stage.

The reason for the delay is simple. It seems as if the upcoming GPU generation called Ampere/Hopper seems to boost the speed by up to 70% which means dramatical performance improvements to the current GPU generation.

The original Pase2 upgrade stage was planned with V100 Tesla GPU cards. The V100 is based on a fully-scaled Volta GPU as it is also seen in the Pascal architecture which is used for consumer GPUs. The Volta GPUs however are solely used for AI computing and thus have no graphical purpose. They’re built to achieve extremely high calculation speeds when it comes to AI calculation routines.

The successor is now taking this speed any further and going to boost the speed by up to 70% which means up to 19 TFLOPs in single precision mode (up to 21 TFLOPs in boost mode).

If this architecture also allows speculations for faster GPUs in the consumer section, is yet unknown. But NVidia already has announced their 3rd-gen RTX cards. so maybe a 3080 SUPER Ti might be based upon the new Ampere/Hopper architecture and allow for faster 3D graphics up to 8K (as 8K TV sets become more popular these days.)