News for hackers. Stuff that matters.
Re:GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000... by Scene_Cast2 on Saturday October 24, 2026 @10:38AM (#50221680)
Really interesting.

On the other hand, this only applies to a limited set of workloads (training is often memory bandwidth bound for example).

I also wonder how this affects power consumption.

Re:GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000... by kingstnap on Saturday October 24, 2026 @11:10AM (#50222034)
Training is compute bound.

If its memory bandwidth bound you need to raise your batch sizes to leverage more parallelism.

Re:GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000... by Zagitta on Saturday October 24, 2026 @11:23AM (#50222165)
It's also useful for prefill in inference, VLLM sees up to 30% boost depending on the exact kernels and prefill size