Unfortunately, If you don’t use 2 sparks, you’re basically wasting $1500, because you’re not using the infiniband module you already have. Also training scales linearly with cluster expansion.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| GLM 4.6V works on Spark! | 12 | 1828 | January 22, 2026 | |
| DGX Spark performance | 50 | 3093 | February 27, 2026 | |
| Two-Spark cluster with vLLM using tensor-parallel-size 2 causes one node to drop while the other's GPU goes 100% forever | 36 | 856 | February 13, 2026 | |
| Running GLM-4.7-FP8 (355B MoE) on 4x DGX Spark with SGLang + EAGLE Speculative Decoding | 32 | 1001 | February 26, 2026 | |
| New bleeding-edge vLLM Docker Image: avarok/vllm-nvfp4-gb10-sm120 | 35 | 2210 | December 31, 2025 | |
| GLM-4.7-Flash-NVFP4 was just released, but for Transformers 5.0 + vLLM 0.14...? | 90 | 3843 | February 27, 2026 | |
| Help: Running NVFP4 model on 2x DGX Spark with vLLM + Ray (multi-node) | 18 | 1949 | December 25, 2025 | |
| Make GLM-4.7-Flash go BRRRRR | 17 | 1714 | February 5, 2026 | |
| Install and Use vLLM for Inference on two Sparks does not work | 159 | 4424 | December 9, 2025 | |
| Setting up vLLM, SGLang or TensorRT on two DGX Sparks | 16 | 1356 | December 7, 2025 |