Titan RTX

Dual GPU llama.cpp tests reveal PCIe transfer differences between tensor and pipeline parallel

A Reddit user conducted tests on a dual-GPU setup using llama.cpp to measure PCIe transfer speeds under different parallelization strategies. The tests compared tensor parallel and pipeline parallel configurations, reve…