[raziel2p]

> Vortex is designed to support decoding data directly from S3 to GPU, skipping the CPU bottleneck entirely.

how is this significant? surely either the network or the GPU calculations is the bottleneck here?