Running 200B Models Locally: 2x RTX PRO 6000 Beats DGX Spark by 10x

TheZachMueller · x · 2026-08-09

A developer's benchmarks reveal that a 200K-context agentic session takes roughly 2-3 minutes on 2x RTX PRO 6000s, compared to 22 minutes on 2x DGX Sparks.

Concurrency and Tensor Parallelism also scale significantly worse on the DGX Spark, making the two systems fundamentally unequal for heavy workloads.

Related event: Dual RTX PRO 6000 Outperforms DGX Spark by 10x in Local LLM Tests(2 posts)→

Original post →

More from Infra

Infra channel →