The 16-inch MacBook Pro equipped with Apple’s M5 Max chip and the Asus ProArt P16 laptop, equipped with the RTX Spark, Nvidia’s first full chip for Windows laptops, underwent the same tests. In the comparison published by Frandroid, the Mac wins most of the duels, but not all.
Both machines are based on the same logic: a CPU and a GPU combined on a single chip, drawing from a shared memory pool. The RTX Spark pairs a 20-core Arm processor, co-designed with MediaTek, with a Blackwell GPU, with up to 128 GB of shared memory. The M5 Max features 18 CPU cores and 40 GPU cores, with 128 GB of memory in the tested version. The RTX Spark is in its first generation, while Apple Silicon is in its fifth since the M1 chip of 2020.
CPU and memory: advantage to the Mac
On Geekbench 7, the MacBook Pro scores 3,759 points in single-core and 35,526 points in multi-core, which is 43% and 51% more than the RTX Spark, despite having two fewer cores. In Cinebench 2026, which simulates 3D rendering, the gap remains around 45% in multi-core. On the memory side, the bandwidth, i.e., the amount of data a chip can read per second, reaches 558 GB/s on the GPU side for the Mac, versus 249 GB/s for the RTX Spark.
Local AI: the Mac writes faster, the RTX Spark reads faster
A language model produces its text in small pieces called tokens, which correspond to a word or a fragment of a word. Across eight models tested with llama.cpp, the MacBook Pro writes faster, by 1.46 to 1.96 times. With Qwen3 8B, it reaches 98.7 tokens per second, compared to 53.1 for the RTX Spark. On gpt-oss-120b, OpenAI’s open model, the figures are 107 tokens per second on the Mac versus 63.3 on the RTX Spark.
For reading a text of 16,384 tokens, equivalent to a long report, the situation reverses. The RTX Spark reads 54% faster than the Mac with Qwen3 8B, thanks to its Tensor cores, specialized for AI computations. In image generation with Flux.2 Klein 4B in ComfyUI, the Mac takes 36.2 seconds for a 1,024 x 1,024 pixel image, versus 16.2 seconds for the RTX Spark. According to the test author, Nvidia’s Tensor cores natively handle the NVFP4 format, while the Mac calculates in 16-bit.
Gaming, battery life, and price
In Cyberpunk 2077, at 2,560 x 1,600 pixels, on Ultra settings and without ray tracing, the MacBook Pro reaches 57.3 frames per second, compared to 44.9 for the RTX Spark. However, the game catalog remains much richer on Windows. In web browsing, under the protocol applied to both machines, the Mac lasts about 16 hours 30 minutes, versus 9 hours 45 minutes for the RTX Spark. Apple advertises up to 22 hours of battery life for the 16-inch M5 Max, a figure from its own testing conditions, which the tester did not replicate.
The tested MacBook Pro, with 128 GB of memory and 2 TB of storage, costs 8,399 € on the Apple Store. The Asus ProArt P16 RTX Spark, in the same configuration, is listed at 7,599.99 € as a suggested price. With equal memory and storage, the Mac therefore costs 800 to 960 € more. Yet, relative to price, it offers, according to the test, about 50% more text AI per euro spent on gpt-oss-120b.
