RTX Spark vs Mac Studio: Same $6,000, Very Different Speed

Two AI boxes, the same 128 GB of memory and about the same $6,000: a Mac Studio with Apple's M5 Max, and NVIDIA's brand new RTX Spark with "1 petaflop" on the box. NVIDIA's spec page lists the cores, the power and the video encoder, but not the one number that decides how fast your model writes. I looked it up, tested the rule on my own 8 GB graphics card, and lined up the benchmarks that actually exist. Chapters 0:00 Same price, same 128 GB 0:52 What the RTX Spark actually is 1:53 The hype: 1 petaflop, “6.2× faster” 2:35 How a model really writes 3:18 I tested the rule on my RTX 5060 3:55 The number NVIDIA didn’t print 5:00 The race: gpt-oss-120b 6:01 Qwen3.8-27B: almost 4× apart 6:58 The twist: where the petaflop wins 8:08 Where the Spark simply wins 8:54 Who each one is for 9:36 What I’d actually do 10:24 The number that matters #localai #rtxspark #macstudio #LocalAI #RTXSpark #MacStudio