5nm process, 114 billion transistors, 20 core CPU, 64 core GPU... In Apple's publicity, in the eyes of many Apple fans, the M1 Ultra chip has become a second-in-a-second existence, whether it is performance or power consumption, whether it is AMD Ryzen™ 9 5950X, Intel Core i9-12900K processor or NVIDIA RTX 3090 graphics card, are not opponents.
However, the actual measurements show that in terms of GPU performance, M1 Ultra is still a huge gap compared to RTX 3090, GeekBench 5 computing performance lags behind by more than 50%, and game performance is much worse.
What about CPU performance? Can this guy with three times the size of AMD Ryzen really kill the x86 flagship?
Here is a summary of some of the tests of foreign media, let's take a look:
In the CineBench R23 test, the Ryzen 9 5950X single-core and multi-core were 4% and 21% ahead of the M1 Ultra, respectively, while in the GeekBench 5 test, the M1 Ultra single-core and multi-core had an advantage of 4% and 34%, respectively.
Which one should I look at? You should know that the GeekBench test is more favorable to Apple, often referred to as "Applebench".
Replaced by the i9-12900K, the CineBench R23 single multi-core is far ahead of the M1 Ultra by 28% and 27%, and the GeekBench 5 test is also 12% ahead of the single core, but the multi-core is losing 37%.
Another special test multi-core, the situation is similar, GeekBench 5 is still Apple's world, CineBench R23 is the original form, but only about 10% behind, I have to say that it is still strong enough.
To be a little practical, Adobe Premiere PRO 4K H.264 video transcoding test, M1 Ultra compared to M1 Max is significantly more advanced, a full double faster (after all, two-in-one), but also much faster than the combination of Ryzen™ 9 5950X, i9-12900K and RTX 3090, obviously very suitable for professional creation.
Third, in the CineBench R23 test, the M1 Ultra faced the i9-12900K, the single-core performance lagged behind by 18.5%, and the multi-core performance lagged behind by 8.9%, while compared with the M1 Max, the single-core performance was basically unchanged, and the multi-core performance doubled.
In the PugetBench Premiere Pro test, the M1 Ultra was comparable to the combination of the i9-12900K and RTX 3080 Ti, and the M1 Max was only improved by 5.3%.
In the PugetBench Photoshop test, the M1 Ultra performed average, lagging more than 20% behind the combination of the i9-12900K and RTX 3080 ti.
In short, the M1 Ultra is indeed strong enough, and the benchmark running score is lagging behind but definitely worthy of AMD and Intel's vigilance, and it is also very advantageous in specific workloads, but if you praise it now, it is still a bit exaggerated.
There are rumors that Apple also plans to stitch two more M1 Ultra together, so that even if the frequency is reduced, the multi-core performance is enough to completely pick out AMD and Intel, but this is not good for single-core performance.