The new M5 Ultra Mac Studio is on a par with graphics cards like the Nvidia RTX 5080 in a PC, and the whole package is just as cost-effective for AI loads.
The first reviews of Apple's new Macs are starting to surface, and so too are the benchmarks. With the M5 Ultra being Apple's newest high-performance chip, the Mac Studio update is of particular interest to those in the field of AI.
If those first benchmarks are to be believed, the M5 Ultra is certainly as high-performance as ever. To the level of some premium graphics cards.
As raised by Overclocking.com, searches for the Mac17,15 in Geekbench 7 brings up GPU results for the M5 Ultra, with Metal test results in the general range of 350,000 to 360,000. For reference, the previous Mac Studio with M3 Ultra and 80 GPU cores scored 249,564 under the same Metal test.
When it comes to comparing against a consumer graphics cards, a good benchmark is the Nvidia RTX 5080. However, we have to look at OpenCL results instead of Metal.
Geekbench's OpenCL listings puts the GeForce RTX 5080 as scoring 250,711, which is very impressive. The Mac Studio results, however, show OpenCL scores around 210,000 to 220,000.
While not quite toppling the 5080 in GPU performance, it is still a very impressive improvement for the Mac Studio. Under the OpenCL test, the M3 Ultra with 80 GPU cores scored 140,821.
The results should be of great benefit to people working in the field of AI research and development, which commonly uses high-powered graphics cards as part of expensive computing rigs.
On GPU performance alone, it is evident that the Mac Ultra can still hold its own in the competitive field. When you consider the Dual Neural Engine and that each GPU core has a Neural Accelerator to assist in AI processing, that certainly helps the Mac Studio's position.
But then there's power to consider, which is the Mac Studio's wheelhouse. High-end graphics cards like the RTX 5080 have massive power needs, and computers using them can often pull over 600W under load.
With multi-GPU setups, this can approach and even exceed 1,000W.
By comparison, Apple's approach has been to enhance energy efficiency where possible. Previous models of the Mac Studio have been found to draw at considerably lower levels, closer to 100W and well below 200W at a push.
While raw processing power is a key element of building an AI development setup, other factors like power consumption certainly help Apple get a bit of an edge on rivals working in the space.
