M1 Ultra vs RTX 3080ti for Machine Learning
Table of Contents:
- Introduction
- Comparison of M1 Ultra and Intel Core i9
- Machine Learning Test by Thomas Capel
- Setup and Environment
- Testing the GPU Performance
- Power Draw and Fan Speed
- Results of the Machine Learning Test
- Comparing M1 Ultra with Other Macs and Dedicated Nvidia Graphics Card
- Apple's Comparative Performance Claims
- Conclusion
Introduction
In this article, we will be comparing the performance of the M1 Ultra and an Intel Core i9 machine in a machine learning test. The M1 Ultra is equipped with a 48-core GPU, which is considered to be the best Apple silicon GPU. On the other HAND, the Intel Core i9 machine has an Nvidia RTX 3080 Ti graphics card, which is a powerful discrete graphics card. We will be using a test developed by Thomas Capel, which utilizes TensorFlow. This test will help us evaluate the capabilities of the M1 Ultra and its unified GPU compared to a system with a dedicated graphics card.
Comparison of M1 Ultra and Intel Core i9
Before diving into the machine learning test, let's take a closer look at the specifications of the M1 Ultra and the Intel Core i9 machine. The M1 Ultra boasts a 48-core GPU, which is an impressive feature for an Apple silicon GPU. On the other hand, the Intel Core i9 machine is equipped with the powerful Nvidia RTX 3080 Ti graphics card. This discrete graphics card offers excellent performance for demanding tasks.
Machine Learning Test by Thomas Capel
The machine learning test we will be conducting is developed by Thomas Capel. He has created a repository where the test can be found, and we will provide a link to it. The test utilizes TensorFlow, a popular machine learning framework. It is worth mentioning that at the time of writing, PyTorch does not have GPU support, but once it becomes available, we will also conduct that test. Thomas Capel's repository allows for easy tracking of results from different users.
Setup and Environment
To conduct the machine learning test, we have set up an Ubuntu environment for the Intel Core i9 machine and a macOS environment for the M1 Ultra. Additionally, we have installed Miniconda on the macOS environment. If You need guidance on setting up Miniconda, we have a separate video tutorial that you can refer to. Both machines have already cloned the repository, so We Are ready to proceed with running the test.
Testing the GPU Performance
In order to evaluate the GPU performance of the M1 Ultra and the Intel Core i9 machine, we will be running the machine learning test according to the instructions provided in the repository. However, we will be adding a time command to measure the execution time of the test. Additionally, we need to specify the GPU name in the command. For the M1 Ultra, the GPU name is "M1 Ultra," and for the Intel Core i9 machine, it is "RTX3080Ti."
Power Draw and Fan Speed
During the test, it is important to monitor the power draw and fan speed of both machines. The power draw of the M1 Ultra is around 117 watts, while the Intel Core i9 machine fluctuates between 300 and 500-plus watts. The fan speed of the M1 Ultra remains consistent at around 1300 rpm, regardless of the GPU load. This may have implications for performance and heat management.
Results of the Machine Learning Test
After conducting the machine learning test on both the M1 Ultra and the Intel Core i9 machine, we have obtained the following results. The Intel Core i9 machine with the Nvidia RTX 3080 Ti completed the test in 5 minutes and 59 seconds. On the other hand, the M1 Ultra with its 48-core GPU took 14 minutes and 59 seconds, almost three times longer. This significant difference in performance raises questions about the effectiveness of the M1 Ultra for GPU-related activities like machine learning.
Comparing M1 Ultra with Other Macs and Dedicated Nvidia Graphics Card
When comparing the M1 Ultra to other Macs, the performance difference will vary. If you are interested in seeing a comparison with other Macs, please let us know in the comments section. However, when it comes to comparing the M1 Ultra with a dedicated Nvidia graphics card like the RTX 3080 Ti, the difference is clear. The dedicated graphics card outperforms the M1 Ultra, leaving no competition.
Apple's Comparative Performance Claims
It is worth noting that if Apple claims that the M1 Ultra has equivalent performance to the Nvidia RTX 3090, it is not Based on this machine learning test. The results of this test clearly demonstrate the superiority of the dedicated Nvidia graphics card. Therefore, it is important to critically evaluate claims made by manufacturers and conduct independent tests to verify performance claims.
Conclusion
In conclusion, the machine learning test conducted on the M1 Ultra and the Intel Core i9 machine highlights the significant difference in performance between the two. While the M1 Ultra's 48-core GPU is impressive for an Apple silicon GPU, it falls short in comparison to a dedicated graphics card like the Nvidia RTX 3080 Ti. When it comes to demanding tasks like machine learning, the M1 Ultra's unified GPU struggles to match the performance of a discrete graphics card.