NVIDIA DGX SPARK CLUSTERS NOW REACH 8 UNITS AND 1 TB OF MEMORY

What happened: Nvidia's DGX Spark local AI systems now support clusters of up to eight units, doubling the previous limit of four, with unified memory reaching 1 TB in the most advanced setups. Key numbers: Up to 8 units per cluster, previously 4. Up to 1 TB unified memory. Why it matters: Nvidia says the added memory lets teams run large-scale models with long contexts and distributed development tools without depending on outside servers, and the software work with vLLM and llama.cpp targets smoother local inference.

NVIDIA DGX Spark clusters now scale to 8 units with 1 TB memory.

NVIDIA DGX Spark clusters now scale to 8 units with 1 TB memory. — impact context not separately stored; see What changed.