Despite Apple’s claims, the new M1 Ultra chip’s GPU cannot outperform that of Nvidia’s flagship RTX 3090 chip.
The graph Apple showed off during the M1 Ultra’s announcement (above) shows the chip outperforming the “highest-end discrete GPU” at a fraction of the power. That discrete GPU is the RTX 3090.
The problem is the focus of the graph. It’s not inaccurate, just misleading. Apple is measuring GPU performance against its power draw, which speaks to the efficiency of the M1 Ultra. And it’s a highly efficient chip. The problem is that Apple can also claim that the M1 Ultra is faster than the RTX 3090 when it isn’t.
And that’s proven in tests conducted by The Verge, which have shown that M1 Ultra is no match for RTX 3090 system. Apple Silicon fails to offer better performance in gaming and compute tests featuring OpenCL API.
The Mac Studio review compared a 20-core M1 Ultra with 64 GPU cores against a desktop powered by an RTX 3090 and an Intel Core i9-10900 (both systems had 64 GB RAM).
Bearing in mind, Apple’s test system for its M1 Ultra performance graph actually featured a much-more powerful CPU with the Intel Core i9-12900K, although these tests are all about the GPU part anyway.
It’s not unsurprising that the Mac Studio with M1 Ultra falls behind in the frame-rate test, although the results are still more than impressive: 60 FPS at 4K in Shadow of the Tomb Raider is truly impressive.
However, the M1 is trounced by the GeForce RTX 3090 in the OpenCL synthetic benchmark, and there are enough test entries on Geekbench now to show that this is not an outlier result.
Apple’s marketing text for the M1 Ultra comments in the notes that “performance was measured using select industry‑standard benchmarks,” and there is the fair argument that the new chip performs exceptionally well considering the much-lower power draw. But a little more transparency would go a long way in what seems to be a typical marketing maneuver in the tech world.
