Decode · tok/s
153.4
Prefill
2844
tok/100W
21.9
Strata v0.1.42 on 2× RTX 3090 (48GB VRAM) + 64GB (DDR5 6000MHZ) RAM! Qwen3.8-Flash-Next IQ3_S Cold prefill: 2,844 tok/s (191K context) Peak prefill: 3,113 tok/s Sustained decode: 120+ tok/s Peak decode: 153.4 tok/s 262K context, CPU/RAM offload, pipeline-windows=2.
First-person 2× 3090 update on v0.1.42. Rank is the quoted peak 153.4. Sustained 120+ and the prefill figures stay in the note. Prior pair on the same cards was 122.5 peak.
