Tesla P40
A data-center GPU from NVIDIA based on the Pascal architecture with 24GB of GDDR5 memory; its low second-hand price makes it commonly used for large-model inference.
Timeline (last 90 days)
141 submissions and 2,156 files over three weeks, covering 45 graphics cards, from Tesla P40 to RTX 5090.
Dual-card comparison experiments: Pro 4500 running Q4 went from 61.5 tok/s on a single card to 111.0 tok/s on dual cards (1.81x); Tesla P40 went from 19.6 to 34.8 tok/s (1.78x); the other two groups were 1.48x and 1.22x.
All Facts (2)
141 submissions and 2,156 files over three weeks, covering 45 graphics cards, from Tesla P40 to RTX 5090.
50%UnverifiedDual-card comparison experiments: Pro 4500 running Q4 went from 61.5 tok/s on a single card to 111.0 tok/s on dual cards (1.81x); Tesla P40 went from 19.6 to 34.8 tok/s (1.78x); the other two groups were 1.48x and 1.22x.
50%