At the press conference in the early hours of this morning, Apple launched the dream engine Mac Studio host, equipped with a processor M1 Ultra, equivalent to two M1 Max chips, 5nm process under the 114 billion transistors, which crushed all the current x86 processors.
As for the performance of the M1 Ultra, we also reported the test results in GeekBench 5, single-core 1793 points, multi-core 24055 points, multi-core score and 64 core Threadripper 3990X comparable, only 4% difference, basically close to the top server processor results.
Is the performance of the M1 Ultra really to the point of eliminating AMD and Intel? Looking at a GK5 score alone is certainly not enough, and the GK benchmark has been criticized as Apple's royal use and not friendly to x86.
The big V@wieken on the push gave another test, running the R23 test commonly used by CPU players, comparing the performance of the Core i9-12900K, of which M1 Ultra single-core, multi-core is 1562, 24566, 12900K single-core, multi-core is 2019, 27669 points.
12900K is 16 cores and 24 threads, M1 Ultra is 20 cores, and the multi-core performance is 3000 points worse than 12900K, which is about 1/8 of the performance difference, which is very close.
In terms of single-core performance, the 12900K is about 1/4 ahead of the M1 Ultra, and this gap is still somewhat obvious.
For this achievement, Intel and AMD can not be taken lightly, because the M1 Ultra suffers in the low frequency, it should be around 3.2GHz, and the 12900K single-core frequency exceeds 5GHz.
On the other hand, AMD and Intel have no way to compare with Apple, because chips such as M1 Ultra are too expensive, and the cost of preliminary analysis is not less than $1,000, which is 1-2 times more expensive than AMD and Intel processors, and 12900K is much cheaper in comparison.
