NVIDIA DGX Spark — first boot and local LLM inference
Unboxing, OTA update, NVIDIA NIM setup and benchmarking Llama 3.1 8B and Qwen3 32B on DGX Spark. Real-world experience including the problems I hit along the way.
2 posts
Unboxing, OTA update, NVIDIA NIM setup and benchmarking Llama 3.1 8B and Qwen3 32B on DGX Spark. Real-world experience including the problems I hit along the way.
The host CPU type enables Spectre/Meltdown mitigations inside Windows, dramatically slowing down RAM operations. The fix is x86-64-v3.