GPUs / NVIDIA

RTX 3080

Memory
GDDR6X
Bus
320-bit
Published peak
760 to 912 GB/s
Capacity
from 10 GB

Measured

Read
610.0GB/s
Write
548.2GB/s
Copy
576.1GB/s
Random
48.1GB/s
Latency
274ns
Sustain
572.0GB/s

Medians of 2 verified runs from 1 operator.

84.52 verified runs, median 84.684.6

Local AI speed

ModelParameters4-bit, tokens/s8-bit, tokens/s
Llama 3.1 8B8.03B≤ 152≤ 76
Qwen 2.5 14B14.7B≤ 83Does not fit
gpt-oss-20b21B, 3.6B activeDoes not fitDoes not fit
Mistral Small 24B24BDoes not fitDoes not fit
Qwen 2.5 32B32.5BDoes not fitDoes not fit
Llama 3.3 70B70.6BDoes not fitDoes not fit
gpt-oss-120b117B, 5.1B activeDoes not fitDoes not fit
Upper bounds: median read ÷ bytes read per token. Mixture-of-experts models read only their active experts. Fit checks weights against the smallest 10 GB configuration; the KV cache needs room on top.

Top operators

#OperatorScoreReadDate
1OP-4AB83C84.6614.4 GB/s2026-09-27

Full RTX 3080 leaderboard