deepseek-v4.1IMAGE
DeepSeek-V4.1-Flash-1017 EXL3 on DGX Spark: TP3 vs TP2 gains ~34%
With identical 2.9 bpw weights and clocks capped at 2,000 MHz, 3 Sparks (TP3) vs 2 (TP2) gave 117 vs 88 tok/s on code, 82 vs 61 on prose and 156 vs 120 on 4 streams, roughly 29-34% faster.
Thomas Eastham
@easthambuildsArchived 9 days agoOriginally posted 2026.10.22
Thomas Eastham @easthambuildsOriginally posted 2026.10.22
Overnight speed test: DeepSeek-V4.1-Flash-1017 EXL3 on DGX Spark, with @alexellisuk rigmark, sandboxed, clocks capped at 2,000 MHz. Same 2.9 bpw weights, 3 Sparks (TP3) vs 2 (TP2): code 117 vs 88 tok/s (+34%) prose 82 vs 61 (+34%) 4 streams 156 vs 120 (+29%) 64K prefill: level
x.com/easthambuilds/status/2113288165330927950Archived 2026.10.23 12:07