Apple’s new Mac Studio reaches customers and stores on September 22, with M5 Max and M5 Ultra versions that push the compact desktop deeper into high-end AI, video and software-development work. Apple opened preorders on August 25.
The M5 Max configuration starts at $2,499 in the US; the M5 Ultra starts at $5,499. Apple says M5 Max has an 18-core CPU, an up-to-40-core GPU and support for up to 128GB of unified memory. M5 Ultra can scale to a 36-core CPU, an 80-core GPU and 512GB of unified memory, although Apple says the 512GB configurations will not arrive until late October.
Two tiers for different workloads
Apple is positioning M5 Max for creative and development work that benefits from a powerful small desktop. The company says it offers up to 614GB/s of unified-memory bandwidth, third-generation hardware-accelerated ray tracing, and a media engine that handles H.264, HEVC, ProRes and AV1 decoding.
M5 Ultra is the option for workloads that need more CPU, GPU and memory headroom. Apple says it has 1.2TB/s of memory bandwidth and twice the video encode and decode blocks of M5 Max, enabling up to 33 simultaneous 8K ProRes 422 streams at 30 frames per second. Those figures describe Apple’s specified configurations and benchmarks, rather than independent testing.
Apple puts local AI at the center
The company says M5 Ultra reaches up to 4.3 times the peak AI compute performance of M3 Ultra, while M5 Max reaches up to 3.9 times the prior generation’s AI performance. It also says the new systems can use Thunderbolt 5 and RDMA to cluster multiple Mac Studios, with four systems delivering up to three times the AI-inference performance of one system. Those are Apple measurements.
Connectivity is another part of the update. The machines add Wi-Fi 7 and Bluetooth 6, offer Thunderbolt 5, and Apple says the new SSD architecture is up to twice as fast as before. They can support up to eight displays, or four Studio Display XDR units at full 5K resolution and 120Hz.
Early reviews from outlets including PCMag and The Verge have focused on the machine’s substantial local-AI capability, but also on how quickly a fully configured M5 Ultra system can become expensive.