Folding Forum

When V8 can made all GPUs work for same work unit at same time ?

That's very unlikely to happen. It'd require a *lot* of data transfer back and forth between the GPUs (threads need each other's data), which would slow everything down horribly.

They looked into it a number of years ago, testing showed too much PCI overhead transferring data between GPUs. It was determined that running one WU on one GPU was a better use of resources.

It's already hard to get WU that have enough atoms to fully use the biggest GPUs ... image the pain it would be with more than one at once ...