LogTrim

The interesting part is that this is less "CUDA now runs on RISC-V" and more "Nvidia is defining what a CUDA-capable RISC-V server has to look like."

That could end up being a pretty influential de facto server profile for the architecture.

Likely much worse write endurance than DRAM, since it’s still flash underneath. The saving grace is that model weights are mostly read-heavy, so endurance may matter less than it sounds.

The peak-speed arms race stopped being interesting to me a while ago. I don't need 20+ Gbit/s theoretical Wi-Fi. I need the connection to stay boring when I'm at the edge of coverage or the neighbours are hammering the same spectrum.

The 95th-percentile latency and packet-loss targets here seem much more meaningful than another 2x headline speed increase.

My only scepticism is how much of this survives real consumer hardware. Wi-Fi 6 had things like OFDMA that sounded great on paper, but support/implementation was pretty uneven.