Tenstorrent Blackhole p150 Llama-3.1-8B token-gen (vendor b32)
33.1 tok/s per user at batch-32; 1059 t/s agg
Fields
id
benchmarks/comm-tt-blackhole-p150-llama31-8b-tgMetric
tok_s_per_user
Value
33.1
Unit
tok/s per user at batch-32; 1059 t/s agg
Model
Llama-3.1-8B
Methodology
Vendor tt-metal model matrix, batch-32 decode ISL=128; marked '*' = Blackhole sw under active dev
Measured by
vendor
Thermal state
unknown
notes
VENDOR-MEASURED, sw maturing. 2x p150 DP: 1190 t/s.
Notes
VENDOR-MEASURED, sw maturing. 2x p150 DP: 1190 t/s.