Recently, the brand-new macbook air m2 version has been launched, and the mold has been upgraded, but the price has also increased a lot. Therefore, the most cost-effective one is the old air in 2020, and Xianyu charged more than 5,000 for an almost brand-new one.
So, can the most beggar macbook air in 2020 still use MATLAB smoothly?
1. Test platform
Test tool: MACBOOK AIR 2020, 8G+256G, 8-core CPU (4 large cores + 4 small cores), 7-core GPU.
Test software: MATLAB_R2022a, downloaded from the official website, using the genuine license purchased by the school.
2. CPU theoretical running score
2.1 Benchmark software
This time, GEEKBENCH5 is used as the running score software, the version is 5.4.5.
When running the score, it is observed that the software is still running in the translation environment of ROSETTA2, which is not supported by native M1, but the running score is still normal.
2.2 Running score measurement
Below is a screenshot of my running score, which is in line with expectations.
Normal detection of system configuration
Remember to turn off other software before running the score, the score is for reference only.
2.3 Comparison with other processors
To be more intuitive, let's compare INTEL's processors, i7-8750h, i7-9750H, i7-10750H. Everyone should be very familiar with it. The classic 6-core 12-thread processor on the notebook side is also the most widely used.
Among them, 10750H, like M1, is a product of 2020. The data comes from cpu_monkey .
single core
multi-core
I am afraid that 6 cores will be unfair to 8 cores, and then compare with 10875H (8 cores and 16 threads)
In fact, the INTEL processor with 6 cores and 12 threads can already meet my usual requirements for scientific computing. M1 can have this performance in running points, which has met my expectations. Its multi-core performance is roughly equal to INTEL's 8-core 16-thread processor in 2020. In addition, it is worth mentioning that the power consumption of the Intel processor exceeds 50W when running the score, while the power consumption of the M1 is around 5W.
In the end, the running score is just a picture, everyone just read it. If you pursue absolute performance, you can choose the 12th generation Core machine.
3. MATLAB actual measurement
3.1 Adaptation of MATLAB
At present, MATLAB is still not fully adapted to M1, and it is still in the testing stage. You can go to this website to find out whether your application has been adapted. Whether the application is equipped with Apple's self-developed chips
It can also be seen in the activity monitor that MATLAB (R2022A) is still running in the INTEL environment and needs to be translated using ROSETTA2.
3.2 Test method: Use the bench that comes with MATLAB, that is, enter bench in the command line mode.
3.3 The test results are tested with a WINDOWS notebook equipped with a 10750H processor and a MAC notebook equipped with an M1 chip.
M1 running score results
10750H running score result
The words on the picture are a bit small. To sum up, the relative speed of M1 is roughly 30, while the relative speed of 10750H is roughly 35. It can be seen that translation still has an impact on the performance of M1.
3.4 Follow-up
In fact, the MATLAB test version that natively supports M series chips has been released, and the test will continue until 2022/8/31. Looking forward to future updates, Apple's appeal is still possible.
Four. Summary
After the above experiments, it can be concluded that the performance of M1 in MATLAB after translation is slightly weaker than that of INTEL or AMD 6-core 12-thread processors, but their performance is on the same order of magnitude. If M1 is not competent for the job, then the traditional 6-core processor of x64 architecture cannot do the same. The biggest advantage of M1 is its power consumption and battery life, which is why I choose it.
MATLAB is a very powerful software, the above tests are only applicable to ordinary calculations.
If you're using features beyond the above, more testing may be required before drawing reasonable conclusions.