Seeking benchmarks: 4x DGX Spark cluster vs 2x cards

Gobra_Slo · reddit · 2026-08-23

Asks for first-hand experience comparing a cluster of four DGX Sparks to two, specifically regarding available models. Discusses the prosumer hardware scene, software stacks, and real-world performance of LLMs like Qwen, DeepSeek V4F, GLM 5.2/5.3, MiniMax M3, and Tencent Hy3, inquiring about intelligence, stability, and OOM issues.

Original post →

More from Infra

Infra channel →