What's the peak FP16 performance of NVIDIA Tesla P4?

الإجابة

Released in 2016, NVIDIA Tesla P4 is an Pascal-class accelerator with 8 GB of GDDR5, 192 GB/s of memory bandwidth, and FP16 TFLOPS of compute. FP32 peaks at 5.5 TFLOPS and the card draws up to 75W.

In practical terms: enough VRAM to load models into the ~8B-parameter range in FP16 (larger with quantisation), enough bandwidth to avoid memory-starving attention layers, and enough compute to train transformers at batch sizes that saturate modern optimisers.

Rent NVIDIA Tesla P4 on Cherry Servers from $0.16/hr — check live availability and deploy.