[ad_1]
While Apple still needs to fully optimize the M1 processor and its software for the task, a 13-inch MacBook Pro with Apple Silicon performed almost as well in a machine learning test as the 16-inch MacBook Pro with dedicated Radeon graphics.
The benchmarks for the M1 processor have been impressive so far, with scores rivaling even the most expensive Intel MacBook Pro configurations. These are the early days as the software continues to optimize for the processor, so some tasks and processes will experience big jumps in speed as developers take advantage of the hardware.
One space in which the M1 processor should excel is machine learning (ML) processes. As with Apple’s A-series chips like the A12Z Bionic, the M1 has a dedicated neural engine that is used for ML and complex data processing. Apple says the M1 neural engine can handle up to 11 trillion operations per second when in use.
However, this processor is not the best in its class in terms of machine learning, as dedicated GPUs from companies like Nvidia boast even higher numbers for neural operations. The first generation of Macs running Apple Silicon have only the M1 processor to rely on; no additional GPU options available.
The Roboflow developers wanted to compare the new Apple machines with the older Intel variants. The processor transition has only just begun for Apple, so tools like TensorFlow have not yet been optimized to run in a full benchmark.
The testers chose to use Apple’s native tool called CreateML, which allowed developers to train a machine learning algorithm with object-based learning and without written code. The tool is available on M1-based Macs, so the testers believe that it should have been properly optimized for testing.
They chose to compare the 13-inch MacBook Pro with an M1 processor and 8GB of RAM to the 13-inch MacBook Pro with Intel Core i5 and 16GB of RAM that has a dedicated Intel Iris Plus Graphics 645 card. The 16-inch MacBook Pro was also tested. inches with an Intel Core i9 processor, 64GB of memory, and a dedicated Radeon Pro 5500M.
The Roboflow team decided to run the test with a codeless object recognition task. They used Microsoft’s COCO object detection dataset of 121,444 images, then exported the assets using Roboflow software to convert them to the Create ML format. They ran CreateML software using a YOLOv2 object detection model for over 5,000 epochs with a batch size of 32.
The COCO dataset used is a large database of object images that a 4-year-old should easily recognize and is used to test machine learning algorithms. YOLOv2 is a type of image recognition that uses bounding boxes to show where an object is in an image. An epoch is a cycle of a test, and a batch size is the number of objects that are executed in each cycle.
Basically, computers will be shown a series of images and will have to decide what to show based on what they have learned from what was previously shown. As you view more images of a given object, the more accurate it is to identify that object in other random images.
- M1-based MacBook took 149 minutes to finish test with 8% GPU utilization
- The MacBook with Intel Core i5 took 542 minutes to run the test, although it did not use Intel Iris Plus Graphics 645
- The MacBook running Intel Core i9 with Radeon Pro took 70 minutes and used 100% of the GPU during the test.
The team notes that CreateML was able to use 100% of the discrete Radeon GPU, but didn’t bother to use Intel Iris at all and only 8% of the integrated M1 GPU. This hit time and is probably because Apple needs to further optimize the toolset for the M1 processor.
According to this benchmark, the Apple M1 is 3.64 times faster than the Intel Core i5. However, the M1 machine is not fully utilizing its GPU and so far underperforms the i9 with discrete graphics.
Apple is expected to continue optimizing its CreateML framework and is working with TensorFlow to successfully port its toolset to M1. Future M-series processors may have even more powerful neural motors and processors, as rumors already indicate that a 32-core M-series chip could be in a future desktop Mac.
