Artificial Analysis benchmarks NVIDIA RTX Spark N1X laptop on local agent workloads
Article timeline
The articles in this story. Times are ET.
Artificial Analysis@ArtificialAnlysOfficialArtificial Analysis publishes NVIDIA RTX Spark AgentPerf local benchmark analysisAIArtificial Analysis publishes an analysis of NVIDIA RTX Spark performance on its AgentPerf-Local benchmark. The post links a laptops and workstations leaderboard and the open-source aa-agentperf-local tool for benchmarking local machines.
- Artificial Analysis@ArtificialAnlysOfficialArtificial Analysis benchmarks NVIDIA RTX Spark laptop on local agent workloads
AIArtificial Analysis tested a prototype 15-inch Surface Laptop Ultra with an RTX Spark N1X chip, 128 GB unified memory, and $5,900 price on its AA-AgentPerf-Local workload. The laptop completed Qwen3.8 27B in 18.7 minutes, close to DGX Spark at 17.3 minutes and 2.0x faster than MacBook Pro M5 Pro at 37.6 minutes.
- Artificial Analysis@ArtificialAnlysOfficialArtificial Analysis prices RTX Spark laptop at $5,900 among local AI systems
AIArtificial Analysis says the $5,900 RTX Spark laptop costs more than the MacBook Pro (~$3,700) and Ryzen AI Halo (~$4,000), but less than DGX Spark (~$6,000). Its GeForce RTX 5090 desktop, about $9,500 with a card of about $7,000, is the most expensive system listed. The firm notes prices shift weekly and says the most expensive systems currently complete its workload fastest on every model tested.
- Artificial Analysis ArticlesOfficialArtificial Analysis tests NVIDIA RTX Spark laptop on local agent workloads
AIArtificial Analysis tested a prototype 15-inch Surface Laptop Ultra with an RTX Spark N1X chip and 128 GB of memory, priced at $5,899.99. On its AA-AgentPerf-Local benchmark, the laptop finished the Qwen3.8 27B workload in 8 to 19 minutes depending on the model. Each system ran its own vendor-recommended recipe, so runtime, speculative decoding and weight format differ between machines.