title
Apple's New M1 Chip is a Machine Learning Beast (M1 vs Intel MacBook speed test)

description
At the start of the year I bought a nearly top spec Intel-based 16-inch MacBook Pro to edit videos and write code. Then Apple released their M1-powered MacBooks and the performance graphs looked insane. I decided to run a few speed test comparisons to compare my MacBook Pro to the new M1 MacBooks. Read about the experiments and see the results - https://www.mrdbourke.com/m1-macbook-vs-intel-macbook-speed-comparison/ Tests run: 1. Final Cut Pro video export 2. CreateML machine learning model training 3. TensorFlow machine learning model training Links: Learn ML (beginner-friendly courses I teach) - https://www.mrdbourke.com/ml-courses/ ML courses/books I recommend - https://www.mrdbourke.com/ml-resources/ Read my novel Charlie Walks - https://www.charliewalks.com Timestamps: 0:00 - Intro 1:00 - Mac Specifications 1:23 - Tests we’re running 1:46 - Test 1 (video rendering) start 3:39 - Starting battery life(s) 5:15 - 16-inch test 1 results 6:47 - 13-inch Pro test 1 results 8:41 - 13-inch Air test 1 results 9:57 - Test 2 (CreateML) start 13:35 - 13-inch Air test 2 results 14:55 - 13-inch Pro test 2 results 18:38 - 16-inch Pro test 2 results 19:30 - Test 3 (TensorFlow code) start 21:13 - TensorFlow code tests outline 21:50 - Basic CNN test start 23:03 - Transfer learning test start 24:53 - tensorflow_macos benchmark start 28:40 - Ending battery life(s) 29:44 - Test 3 (TensorFlow code) results #machinelearning #m1 #tensorflow

detail
{'title': "Apple's New M1 Chip is a Machine Learning Beast (M1 vs Intel MacBook speed test)", 'heatmap': [{'end': 77.138, 'start': 54.778, 'weight': 0.755}, {'end': 608.191, 'start': 585.99, 'weight': 0.729}, {'end': 1443.627, 'start': 1423.341, 'weight': 0.9}, {'end': 1808.302, 'start': 1763.567, 'weight': 0.702}], 'summary': "Compares the speed and performance of m1 macbooks with intel macbook pro, highlighting m1's machine learning potential, faster video rendering, and superior battery life during ml tasks, with macbook air outperforming the others and achieving 68% battery after a task.", 'chapters': [{'end': 68.828, 'segs': [{'end': 68.828, 'src': 'embed', 'start': 44.451, 'weight': 0, 'content': [{'end': 54.778, 'text': "i bought this bad boy this is almost a top of the line, 16 inch about nine ten months ago, and i'm like, did i just spend all this money on this big,", 'start': 44.451, 'duration': 10.327}, {'end': 57.46, 'text': 'powerful macbook pro and these bad boys?', 'start': 54.778, 'duration': 2.682}, {'end': 61.143, 'text': 'so this is a m1 chip, 16 gigabyte macbook pro.', 'start': 57.46, 'duration': 3.683}, {'end': 67.307, 'text': 'uh, 16 gigs of ram, that is, and this is an m1 chip uh, 16 gigabytes of ram macbook air.', 'start': 61.143, 'duration': 6.164}, {'end': 68.828, 'text': 'can i replace this big dog?', 'start': 67.307, 'duration': 1.521}], 'summary': 'The speaker purchased a top-of-the-line 16-inch macbook pro with an m1 chip and 16gb of ram about nine to ten months ago and is now considering replacing it with a macbook air with the same specifications.', 'duration': 24.377, 'max_score': 44.451, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq044451.jpg'}], 'start': 0.389, 'title': 'M1 chip speed comparison', 'summary': 'Discusses a speed comparison between a 16-inch macbook pro and two m1 macbooks, highlighting the performance and potential of the m1 chip.', 'chapters': [{'end': 68.828, 'start': 0.389, 'title': 'M1 chip speed comparison', 'summary': 'Discusses a speed comparison between a 16-inch macbook pro and two m1 macbooks, highlighting the performance and potential of the m1 chip.', 'duration': 68.439, 'highlights': ['The chapter focuses on comparing the speed of a 16-inch MacBook Pro with two M1 MacBooks, emphasizing the performance of the M1 chip.', 'The speaker questions the value of their previous investment in a top-of-the-line 16-inch MacBook Pro compared to the performance of the M1 chip in the newer models.', 'The chapter mentions the specifications of the devices being compared, including the 16 gigabytes of RAM in both the 16-inch MacBook Pro and the M1 MacBook Air.']}], 'duration': 68.439, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0389.jpg', 'highlights': ['The chapter focuses on comparing the speed of a 16-inch MacBook Pro with two M1 MacBooks, emphasizing the performance of the M1 chip.', 'The speaker questions the value of their previous investment in a top-of-the-line 16-inch MacBook Pro compared to the performance of the M1 chip in the newer models.', 'The chapter mentions the specifications of the devices being compared, including the 16 gigabytes of RAM in both the 16-inch MacBook Pro and the M1 MacBook Air.']}, {'end': 574.183, 'segs': [{'end': 117.583, 'src': 'embed', 'start': 90.705, 'weight': 4, 'content': [{'end': 98.228, 'text': 'Because, okay, the benchmarks are amazing for these, but do they perform doing the actual tasks I would do every day,', 'start': 90.705, 'duration': 7.523}, {'end': 101.529, 'text': 'which is edit videos and write code?', 'start': 98.228, 'duration': 3.301}, {'end': 104.03, 'text': "So that's what we're gonna test them out for.", 'start': 102.349, 'duration': 1.681}, {'end': 108.347, 'text': "The first one we've got is, A video rendering test.", 'start': 104.07, 'duration': 4.277}, {'end': 112.39, 'text': "So I've set them all up to have the same setup here.", 'start': 108.647, 'duration': 3.743}, {'end': 114.972, 'text': "We've got Final Cut Pro, the latest version.", 'start': 112.81, 'duration': 2.162}, {'end': 117.583, 'text': "They're all running Mac OS Big Sur 11.0.", 'start': 114.992, 'duration': 2.591}], 'summary': 'Testing performance of laptops for video rendering and coding tasks using final cut pro and mac os big sur 11.0.', 'duration': 26.878, 'max_score': 90.705, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq090705.jpg'}, {'end': 168.892, 'src': 'embed', 'start': 141.732, 'weight': 2, 'content': [{'end': 147.117, 'text': "So the file, who knows how large it's gonna be, I think it's around 10 gigabytes once it's fully exported.", 'start': 141.732, 'duration': 5.385}, {'end': 153.963, 'text': 'So this would be a great test because this is something that I would actually use these computers for, not like a Geekbench store.', 'start': 148.358, 'duration': 5.605}, {'end': 157.446, 'text': "So let's just get started.", 'start': 155.885, 'duration': 1.561}, {'end': 164.65, 'text': "I'm probably most interested to see how the MacBook Air goes with no fan versus a big dog.", 'start': 157.886, 'duration': 6.764}, {'end': 168.892, 'text': 'I mean, when we were pre-rendering it, the fans on this were just screaming like crazy.', 'start': 164.67, 'duration': 4.222}], 'summary': "Testing computers with large file export, macbook air's performance without fan, and excessive fan noise during pre-rendering.", 'duration': 27.16, 'max_score': 141.732, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0141732.jpg'}, {'end': 217.519, 'src': 'embed', 'start': 190.358, 'weight': 0, 'content': [{'end': 194.72, 'text': "So Why don't you click in this one? Ready? We're going to count down on three.", 'start': 190.358, 'duration': 4.362}, {'end': 196.7, 'text': 'Three, two, one.', 'start': 195.16, 'duration': 1.54}, {'end': 198.241, 'text': 'Alright, there we go.', 'start': 197.401, 'duration': 0.84}, {'end': 200.102, 'text': "We've got the export up here.", 'start': 198.861, 'duration': 1.241}, {'end': 201.983, 'text': 'Export up here.', 'start': 201.262, 'duration': 0.721}, {'end': 204.227, 'text': 'And export up here.', 'start': 203.246, 'duration': 0.981}, {'end': 206.749, 'text': "The important thing is that they're all running off battery life.", 'start': 204.247, 'duration': 2.502}, {'end': 215.217, 'text': 'And so the battery life for the MacBook Pro 13-inch has actually been rated as basically insane.', 'start': 207.61, 'duration': 7.607}, {'end': 217.519, 'text': "I think it's like 20 hours or something like that.", 'start': 215.457, 'duration': 2.062}], 'summary': 'Macbook pro 13-inch has an impressive battery life of around 20 hours.', 'duration': 27.161, 'max_score': 190.358, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0190358.jpg'}, {'end': 364.318, 'src': 'embed', 'start': 337.963, 'weight': 1, 'content': [{'end': 353.298, 'text': 'the macbook pro 16 inch with the dedicated gpu intel chip i9 finished rendering the video when this was on 65 finished and the macbook air is on 55 finished with 83 of battery left so far.', 'start': 337.963, 'duration': 15.335}, {'end': 356.421, 'text': "so that's actually a pretty impressive performance.", 'start': 353.298, 'duration': 3.123}, {'end': 364.318, 'text': "like i mean it's not like this wildly finished before these, i mean this has got 35% left and this one's just over halfway.", 'start': 356.421, 'duration': 7.897}], 'summary': 'Macbook pro 16 inch finished video rendering at 65%, macbook air at 55% with 83% battery, impressive performance.', 'duration': 26.355, 'max_score': 337.963, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0337963.jpg'}], 'start': 68.828, 'title': 'Macbook performance comparison', 'summary': "Delves into macbook air's video rendering performance compared to other machines, focusing on export time, file size, and the impact of a fanless design. it also compares battery life and performance of macbook pro 16-inch, macbook pro 13-inch, and macbook air during video export.", 'chapters': [{'end': 189.577, 'start': 68.828, 'title': 'Macbook video rendering performance test', 'summary': 'Explores the performance of macbook air and other machines in video rendering with a 2 hour 40 minute video, with emphasis on the export time and file size, highlighting the impact of a fanless design on performance.', 'duration': 120.749, 'highlights': ["The MacBook Air's performance in exporting a 26.5 gigabytes estimated file without a fan is compared to other machines, indicating the impact of fanless design on export time and file size.", 'The chapter discusses the test setup using Final Cut Pro, Mac OS Big Sur 11.0, and a 2 hour 40 minute video, providing context for evaluating the performance of the machines in real-world tasks.', 'The relevance of the video export test is emphasized as a practical use case for evaluating the performance of the machines in editing videos, providing insight into their suitability for everyday tasks.']}, {'end': 574.183, 'start': 190.358, 'title': 'Macbook battery life comparison', 'summary': 'Compares battery life and performance of macbook pro 16-inch, macbook pro 13-inch, and macbook air during video export, with the 16-inch having the highest power consumption and the 13-inch showing the best battery performance.', 'duration': 383.825, 'highlights': ['MacBook Pro 16-inch with dedicated GPU consumes more power, losing half a percent of battery for every percent increase in render export. The MacBook Pro 16-inch with dedicated GPU loses half a percent of battery for every percent increase in render export, indicating high power consumption.', 'MacBook Pro 13-inch maintains 83% battery after export, demonstrating superior battery performance compared to the other models. The MacBook Pro 13-inch maintains 83% battery after the video export, showcasing its superior battery performance compared to other models.', 'MacBook Air, with M1 chip and no fan, finishes rendering a 2-hour, 40-minute video in 38 minutes, losing only about 15% of battery. The MacBook Air, with M1 chip and no fan, finishes rendering a 2-hour, 40-minute video in 38 minutes and loses only about 15% of battery, demonstrating impressive performance and efficiency.']}], 'duration': 505.355, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq068828.jpg', 'highlights': ['MacBook Pro 13-inch maintains 83% battery after export, demonstrating superior battery performance compared to the other models.', 'The MacBook Air, with M1 chip and no fan, finishes rendering a 2-hour, 40-minute video in 38 minutes, losing only about 15% of battery, demonstrating impressive performance and efficiency.', "The MacBook Air's performance in exporting a 26.5 gigabytes estimated file without a fan is compared to other machines, indicating the impact of fanless design on export time and file size.", 'MacBook Pro 16-inch with dedicated GPU consumes more power, losing half a percent of battery for every percent increase in render export, indicating high power consumption.', 'The chapter discusses the test setup using Final Cut Pro, Mac OS Big Sur 11.0, and a 2 hour 40 minute video, providing context for evaluating the performance of the machines in real-world tasks.', 'The relevance of the video export test is emphasized as a practical use case for evaluating the performance of the machines in editing videos, providing insight into their suitability for everyday tasks.']}, {'end': 942.019, 'segs': [{'end': 628.942, 'src': 'heatmap', 'start': 585.99, 'weight': 0, 'content': [{'end': 590.312, 'text': 'And when it comes to machine learning, performance is spectacular.', 'start': 585.99, 'duration': 4.322}, {'end': 597.997, 'text': 'Thanks to the neural engine, ML is up to 11 times faster than the previous generation, which means for on-device.', 'start': 591.153, 'duration': 6.844}, {'end': 608.191, 'text': 'Something went wrong, as with all live demos.', 'start': 605.31, 'duration': 2.881}, {'end': 611.453, 'text': "This isn't really live, but that's the beauty of video editing.", 'start': 608.712, 'duration': 2.741}, {'end': 614.134, 'text': "We've now got the same test set up on all three.", 'start': 611.753, 'duration': 2.381}, {'end': 621.578, 'text': '7, 500 training images, 2, 500 testing images, 10 different food classes using CreateML.', 'start': 614.154, 'duration': 7.424}, {'end': 627.041, 'text': "We've got augment data turned on, flip and rotate on each of them.", 'start': 622.499, 'duration': 4.542}, {'end': 628.942, 'text': 'Oh, nearly stuffed up on this one.', 'start': 627.261, 'duration': 1.681}], 'summary': 'Machine learning performance is up to 11 times faster with neural engine, using 7,500 training images and 2,500 testing images for 10 food classes.', 'duration': 54.219, 'max_score': 585.99, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0585990.jpg'}, {'end': 702.484, 'src': 'embed', 'start': 679.52, 'weight': 3, 'content': [{'end': 688.671, 'text': 'And just a little quick update the MacBook Air is about a thousand images in front of the 13-inch MacBook Pro,', 'start': 679.52, 'duration': 9.151}, {'end': 695.719, 'text': 'and the 13-inch MacBook Pro is about three times as many images through as the 16-inch MacBook Pro.', 'start': 688.671, 'duration': 7.048}, {'end': 702.484, 'text': "Now, I'm not entirely sure of what algorithm Apple runs when they use CreateML.", 'start': 696.48, 'duration': 6.004}], 'summary': 'Macbook air leads by 1000 images, 13-inch pro has 3x more than 16-inch pro', 'duration': 22.964, 'max_score': 679.52, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0679520.jpg'}, {'end': 884.346, 'src': 'embed', 'start': 852.949, 'weight': 1, 'content': [{'end': 863.034, 'text': "But the results here, I mean, we're nearly a full minute and a half in front for the MacBook Air.", 'start': 852.949, 'duration': 10.085}, {'end': 865.695, 'text': 'All right, training starting on the MacBook Pro 13 at.', 'start': 863.634, 'duration': 2.061}, {'end': 870.073, 'text': '14 minutes and 50 seconds.', 'start': 868.731, 'duration': 1.342}, {'end': 873.876, 'text': 'Model converged after 10 epochs, same as the MacBook Air.', 'start': 870.453, 'duration': 3.423}, {'end': 877.66, 'text': "Now we're doing the testing and we're done.", 'start': 874.717, 'duration': 2.943}, {'end': 881.103, 'text': 'So this one still has 69% of battery left.', 'start': 878.46, 'duration': 2.643}, {'end': 884.346, 'text': "I'm just gonna take a screenshot of the results there.", 'start': 881.143, 'duration': 3.203}], 'summary': 'Macbook air leads by 1.5 minutes, macbook pro 13 trained in 14m 50s, 69% battery remaining.', 'duration': 31.397, 'max_score': 852.949, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0852949.jpg'}], 'start': 574.723, 'title': "Apple's machine learning performance and macbook air vs macbook pro createml test", 'summary': "Discusses apple's machine learning performance, highlighting that the neural engine makes ml up to 11 times faster than the previous generation. it also compares the processing speed and battery consumption of macbook air, 13-inch macbook pro, and 16-inch macbook pro in training and testing a machine learning model, where the macbook air outperforms the other two models, finishing the task in under 12 minutes with 68% battery left.", 'chapters': [{'end': 628.942, 'start': 574.723, 'title': "Apple's machine learning performance", 'summary': "Discusses apple's machine learning performance, highlighting that the neural engine makes ml up to 11 times faster than the previous generation and provides specific details about the test setup and the use of createml for training.", 'duration': 54.219, 'highlights': ['The neural engine makes ML up to 11 times faster than the previous generation, showcasing significant performance improvements.', 'Test setup includes 7,500 training images, 2,500 testing images, and 10 different food classes using CreateML, with augment data turned on, providing specifics about the test environment and methodology.', 'Reference to the beauty of video editing, indicating the demonstration is not live, providing insight into the presentation process.']}, {'end': 942.019, 'start': 629.622, 'title': 'Macbook air vs macbook pro: createml test', 'summary': 'Compares the processing speed and battery consumption of macbook air, 13-inch macbook pro, and 16-inch macbook pro in training and testing a machine learning model, where the macbook air outperforms the other two models, finishing the task in under 12 minutes with 68% battery left.', 'duration': 312.397, 'highlights': ['The MacBook Air outperforms the 13-inch MacBook Pro and the 16-inch MacBook Pro in training a machine learning model, finishing the task in under 12 minutes with 68% battery left, while the other two models lag behind in processing speed and battery consumption.', 'The 13-inch MacBook Pro lags behind the MacBook Air in processing speed, not even finishing extracting features from its images, let alone commencing training, while the 16-inch MacBook Pro is not even halfway through pre-processing the images after over 30 minutes and 8% battery left.', 'The MacBook Air processes about 1300 images out of 7500 ahead of the other models, with the 13-inch MacBook Pro about a thousand images behind the MacBook Air and the 13-inch MacBook Pro about three times as many images behind the 16-inch MacBook Pro.', 'The MacBook Air finishes testing a machine learning model in 12.13 minutes, about four minutes quicker than the 13-inch MacBook Pro, with 67% battery left, while the 16-inch MacBook Pro struggles to process half the images with 19% battery left and a warning for low battery.']}], 'duration': 367.296, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0574723.jpg', 'highlights': ['The neural engine makes ML up to 11 times faster than the previous generation, showcasing significant performance improvements.', 'The MacBook Air outperforms the 13-inch MacBook Pro and the 16-inch MacBook Pro in training a machine learning model, finishing the task in under 12 minutes with 68% battery left, while the other two models lag behind in processing speed and battery consumption.', 'Test setup includes 7,500 training images, 2,500 testing images, and 10 different food classes using CreateML, with augment data turned on, providing specifics about the test environment and methodology.', 'The MacBook Air processes about 1300 images out of 7500 ahead of the other models, with the 13-inch MacBook Pro about a thousand images behind the MacBook Air and the 13-inch MacBook Pro about three times as many images behind the 16-inch MacBook Pro.']}, {'end': 1153.767, 'segs': [{'end': 981.309, 'src': 'embed', 'start': 946.303, 'weight': 0, 'content': [{'end': 949.985, 'text': 'Referring back to the MacBook Air, that one finished in 11 minutes.', 'start': 946.303, 'duration': 3.682}, {'end': 958.388, 'text': "I mean, Apple must have, again, I'm not sure what chip or what processor the CreateML app runs off.", 'start': 950.125, 'duration': 8.263}, {'end': 961.55, 'text': "I'm guessing on the M1 machines.", 'start': 958.969, 'duration': 2.581}, {'end': 968.173, 'text': "it runs on the Neural Engine, which is Apple's dedicated part of the M1 chip, which is specific for machine learning code.", 'start': 961.55, 'duration': 6.623}, {'end': 971.757, 'text': "If you've ever had experience with machine learning,", 'start': 969.053, 'duration': 2.704}, {'end': 981.309, 'text': 'you know GPUs and specific chips designed for machine learning are a lot faster at creating machine learning models than just pure CPU,', 'start': 971.757, 'duration': 9.552}], 'summary': 'The createml app finished on macbook air in 11 minutes using the neural engine on m1 chip, faster than just pure cpu.', 'duration': 35.006, 'max_score': 946.303, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0946303.jpg'}, {'end': 1060.096, 'src': 'embed', 'start': 994.867, 'weight': 1, 'content': [{'end': 1004.393, 'text': "So I think this has been run off pure CPU, which might explain why it's so much slower in comparison to a machine that costs three times as less.", 'start': 994.867, 'duration': 9.526}, {'end': 1008.216, 'text': 'Three times as slower, but costs three times as more.', 'start': 1005.374, 'duration': 2.842}, {'end': 1018.881, 'text': 'to the wire.', 'start': 1018.14, 'duration': 0.741}, {'end': 1022.686, 'text': "This is the most enthralling video you're going to watch today.", 'start': 1019.582, 'duration': 3.104}, {'end': 1025.89, 'text': "This is the most nerd thing you've ever seen.", 'start': 1022.706, 'duration': 3.184}, {'end': 1027.012, 'text': 'Battery versus Michelle.', 'start': 1025.91, 'duration': 1.102}, {'end': 1027.992, 'text': '7, 500 out of 7, 500 images pre-processed.', 'start': 1027.031, 'duration': 0.961}, {'end': 1029.535, 'text': '1% of battery left.', 'start': 1028.012, 'duration': 1.523}, {'end': 1039.911, 'text': "Is the model going to start training? We're 42 minutes in.", 'start': 1029.555, 'duration': 10.356}, {'end': 1042.07, 'text': "We're 42 minutes in.", 'start': 1040.631, 'duration': 1.439}, {'end': 1043.09, 'text': 'Nearly 42 minutes.', 'start': 1042.111, 'duration': 0.979}, {'end': 1043.972, 'text': 'Training started.', 'start': 1043.152, 'duration': 0.82}, {'end': 1045.652, 'text': "If it goes black here, I'm done.", 'start': 1043.992, 'duration': 1.66}, {'end': 1048.453, 'text': "It's making its way through.", 'start': 1047.173, 'duration': 1.28}, {'end': 1049.913, 'text': 'You can do it.', 'start': 1049.253, 'duration': 0.66}, {'end': 1052.014, 'text': 'My trusty MacBook Pro.', 'start': 1050.634, 'duration': 1.38}, {'end': 1055.155, 'text': 'Has it done it? Training started.', 'start': 1053.815, 'duration': 1.34}, {'end': 1056.175, 'text': '11.57am Training completed.', 'start': 1055.175, 'duration': 1}, {'end': 1057.616, 'text': '12.39pm Testing is starting.', 'start': 1056.195, 'duration': 1.421}, {'end': 1060.096, 'text': "It's testing on 2, 500 images.", 'start': 1057.676, 'duration': 2.42}], 'summary': 'Video processing on cpu: 3x slower, 3x cheaper, 7500 images pre-processed, 1% battery left, training completed in 42 mins, testing on 2500 images.', 'duration': 65.229, 'max_score': 994.867, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0994867.jpg'}, {'end': 1160.686, 'src': 'embed', 'start': 1131.823, 'weight': 2, 'content': [{'end': 1133.383, 'text': "So that's astonishing.", 'start': 1131.823, 'duration': 1.56}, {'end': 1138.024, 'text': 'That is, I mean, the MacBook Air took just 11 and a half minutes.', 'start': 1133.963, 'duration': 4.061}, {'end': 1142.965, 'text': 'And by the way, the MacBook Pro has 65% of battery.', 'start': 1139.144, 'duration': 3.821}, {'end': 1147.626, 'text': 'Mind you, the screen has been on the whole time this has been training, so an extra half an hour.', 'start': 1143.005, 'duration': 4.621}, {'end': 1149.647, 'text': '65% of battery left.', 'start': 1147.646, 'duration': 2.001}, {'end': 1153.767, 'text': "And same thing with the MacBook Air, the screen's been on and it has 64% of battery left.", 'start': 1150.407, 'duration': 3.36}, {'end': 1159.445, 'text': "Well We're going to have to plug this one into charge before the next test.", 'start': 1153.787, 'duration': 5.658}, {'end': 1160.686, 'text': "That's the penalty as well, mate.", 'start': 1159.465, 'duration': 1.221}], 'summary': 'Macbook air completed a task in 11.5 minutes, with 64% battery left.', 'duration': 28.863, 'max_score': 1131.823, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01131823.jpg'}], 'start': 946.303, 'title': "Macbook air's machine learning performance", 'summary': "Discusses the macbook air's performance in running a machine learning model using the cpu instead of a dedicated gpu, resulting in slower processing. it also highlights the number of images pre-processed for the task. additionally, it captures the process of training and testing a machine learning model on macbook pro and macbook air, with the macbook air notably performing faster despite similar remaining battery levels.", 'chapters': [{'end': 1027.992, 'start': 946.303, 'title': 'Macbook air machine learning performance', 'summary': "Discusses the macbook air's performance in running a machine learning model, specifically highlighting the use of the cpu instead of the dedicated gpu, resulting in slower processing compared to machines with specialized chips. it also mentions the number of images pre-processed for the task.", 'duration': 81.689, 'highlights': ['The CreateML app on the MacBook Air finished the task in 11 minutes, running on the CPU only, which is slower compared to machines with specialized chips, despite being three times more expensive.', 'The task involved pre-processing 7,500 images for machine learning.']}, {'end': 1153.767, 'start': 1028.012, 'title': 'Macbook training and testing', 'summary': 'Captures the process of training and testing a machine learning model on a macbook pro and macbook air, with the macbook air notably performing significantly faster despite both devices having similar remaining battery levels.', 'duration': 125.755, 'highlights': ['The MacBook Air completed the testing in less than a minute, while the MacBook Pro, with 65% battery, took 43 minutes and 10 seconds, showcasing the significant difference in performance between the two devices.', 'The training was completed, but the testing did not finish, indicating that both devices achieved similar results due to the assumption of using the same machine learning algorithm under the hood.', 'The process involved testing the model on 2,500 images, demonstrating the scale of the testing process and the computational load on the devices.', 'The chapter begins with the MacBook Pro having only 1% battery, highlighting the tense and uncertain nature of the process and the potential risk of the device shutting down during testing.']}], 'duration': 207.464, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq0946303.jpg', 'highlights': ['The MacBook Air finished the task in 11 minutes, running on the CPU only, slower compared to machines with specialized chips.', 'The task involved pre-processing 7,500 images for machine learning.', 'The MacBook Air completed the testing in less than a minute, while the MacBook Pro, with 65% battery, took 43 minutes and 10 seconds, showcasing the significant difference in performance between the two devices.', 'The process involved testing the model on 2,500 images, demonstrating the scale of the testing process and the computational load on the devices.']}, {'end': 1309.615, 'segs': [{'end': 1273.021, 'src': 'embed', 'start': 1250.854, 'weight': 0, 'content': [{'end': 1258.976, 'text': "So Apple and TensorFlow both wrote a blog post saying There's been an incredible speed up on running TensorFlow code natively on the M1 chip.", 'start': 1250.854, 'duration': 8.122}, {'end': 1260.217, 'text': "so that's what we're going to test.", 'start': 1258.976, 'duration': 1.241}, {'end': 1261.677, 'text': 'Specifically, we have three.', 'start': 1260.377, 'duration': 1.3}, {'end': 1265.919, 'text': 'We have a basic convolutional neural network to do some image classification,', 'start': 1262.217, 'duration': 3.702}, {'end': 1270.68, 'text': "the same one as the architecture you'll find on the CNN explainer website.", 'start': 1265.919, 'duration': 4.761}, {'end': 1273.021, 'text': 'so just a multi-class classification there.', 'start': 1270.68, 'duration': 2.341}], 'summary': "Apple's m1 chip shows incredible speed up running tensorflow code for image classification.", 'duration': 22.167, 'max_score': 1250.854, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01250854.jpg'}], 'start': 1153.787, 'title': 'Testing tensorflow on m1 chip', 'summary': 'Discusses the setup and testing of tensorflow code on m1 machines, emphasizing battery usage, running native tensorflow code, and performance improvements based on three specific tests.', 'chapters': [{'end': 1309.615, 'start': 1153.787, 'title': 'Testing tensorflow on m1 chip', 'summary': 'Discusses setting up and testing tensorflow code on m1 machines, highlighting battery usage and running native tensorflow code, with a focus on three specific tests and the expected performance improvements.', 'duration': 155.828, 'highlights': ['The M1 machines, despite not being charged, are being used for testing TensorFlow code, with one machine at 70% battery and another at 40%, showcasing the efficiency and power usage of the M1 chip.', "The chapter focuses on testing native TensorFlow code on M1 machines, highlighting Apple's claim of incredible speed up in running TensorFlow code natively on the M1 chip, which is to be tested through three specific experiments including basic convolutional neural network, transfer learning, and a benchmark.", 'The tests include running a basic convolutional neural network, transfer learning with an efficient net, and a benchmark on the TensorFlow macOS, aiming to gauge the performance improvements of running TensorFlow code natively on the M1 chip.']}], 'duration': 155.828, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01153787.jpg', 'highlights': ['The tests include running a basic convolutional neural network, transfer learning with an efficient net, and a benchmark on the TensorFlow macOS, aiming to gauge the performance improvements of running TensorFlow code natively on the M1 chip.', "The chapter focuses on testing native TensorFlow code on M1 machines, highlighting Apple's claim of incredible speed up in running TensorFlow code natively on the M1 chip, which is to be tested through three specific experiments including basic convolutional neural network, transfer learning, and a benchmark.", 'The M1 machines, despite not being charged, are being used for testing TensorFlow code, with one machine at 70% battery and another at 40%, showcasing the efficiency and power usage of the M1 chip.']}, {'end': 1521.872, 'segs': [{'end': 1371.386, 'src': 'embed', 'start': 1340.392, 'weight': 1, 'content': [{'end': 1342.814, 'text': "I mean, you've lost test one already.", 'start': 1340.392, 'duration': 2.422}, {'end': 1345.076, 'text': 'Wow, this is incredibly fast.', 'start': 1343.294, 'duration': 1.782}, {'end': 1347.717, 'text': "So they're taking about eight seconds per epoch each.", 'start': 1345.236, 'duration': 2.481}, {'end': 1353.879, 'text': 'So the MacBook Pro, only five epochs on 750 images with a relatively small convolutional neural network.', 'start': 1347.937, 'duration': 5.942}, {'end': 1357.061, 'text': "But in terms of running, wow, that's finished, done.", 'start': 1354.039, 'duration': 3.022}, {'end': 1361.882, 'text': 'So about eight seconds, seven seconds, eight seconds, seven seconds, seven seconds per epoch.', 'start': 1357.161, 'duration': 4.721}, {'end': 1367.185, 'text': 'And the MacBook Air has got eight seconds, seven, seven, seven, seven.', 'start': 1362.903, 'duration': 4.282}, {'end': 1371.386, 'text': 'So the MacBook Air, slightly faster per epoch than the MacBook Pro.', 'start': 1367.205, 'duration': 4.181}], 'summary': 'Macbook pro: 8 secs/epoch, 5 epochs, 750 images. macbook air: 7 secs/epoch, slightly faster.', 'duration': 30.994, 'max_score': 1340.392, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01340392.jpg'}, {'end': 1452.529, 'src': 'heatmap', 'start': 1423.341, 'weight': 0, 'content': [{'end': 1424.262, 'text': "You're seeing this live.", 'start': 1423.341, 'duration': 0.921}, {'end': 1432.953, 'text': "This is, we got five seconds per epoch, seven seconds per epoch, and this one's about 25 seconds per epoch.", 'start': 1424.303, 'duration': 8.65}, {'end': 1435.385, 'text': "okay, we've just finished.", 'start': 1434.165, 'duration': 1.22}, {'end': 1443.627, 'text': 'so that took 66 seconds on the big dog versus 24 seconds and 21 seconds on the MacBook.', 'start': 1435.385, 'duration': 8.242}, {'end': 1448.168, 'text': 'the MacBook Air is going faster than the MacBook Pro.', 'start': 1443.627, 'duration': 4.541}, {'end': 1448.888, 'text': 'this is amazing.', 'start': 1448.168, 'duration': 0.72}, {'end': 1452.529, 'text': "so these are they've got one more Epoch left.", 'start': 1448.888, 'duration': 3.641}], 'summary': 'Testing showed the macbook air outperformed the macbook pro, with the former taking 24 seconds and the latter taking 66 seconds to complete a task.', 'duration': 50.666, 'max_score': 1423.341, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01423341.jpg'}], 'start': 1310.016, 'title': 'Tensorflow and efficientnet b0 performance', 'summary': "Evaluates tensorflow performance on m1 max, with macbook air exhibiting faster per-epoch training times of 7-8 seconds compared to macbook pro's 8 seconds. it also involves transfer learning using efficientnet b0, achieving 66 seconds per epoch on a macbook pro and 21 seconds per epoch on a macbook air, with 750 training images and 2,500 testing images.", 'chapters': [{'end': 1384.134, 'start': 1310.016, 'title': 'Tensorflow performance comparison on m1 max', 'summary': "Evaluates the performance of tensorflow on m1 max, comparing the training times of macbook pro and macbook air using the same data and architecture, with the macbook air displaying faster per-epoch training times of about 7-8 seconds compared to the macbook pro's 8 seconds, for a relatively small convolutional neural network.", 'duration': 74.118, 'highlights': ["The MacBook Air displays faster per-epoch training times of about 7-8 seconds compared to the MacBook Pro's 8 seconds, for a relatively small convolutional neural network.", "The chapter also mentions the error encountered by the 'big dog' during training, resulting in the inability to test its performance with the same data and architecture used for the other devices.", 'The MacBook Pro completes training with only five epochs on 750 images, while the MacBook Air finishes training relatively faster, demonstrating its efficiency in handling the workload.']}, {'end': 1521.872, 'start': 1384.274, 'title': 'Transfer learning with efficientnet b0', 'summary': 'Involves performing transfer learning using an efficientnet b0 pre-trained model on a dataset, achieving 66 seconds per epoch on a 16-inch macbook pro and 21 seconds per epoch on a macbook air, with 750 training images and 2,500 testing images.', 'duration': 137.598, 'highlights': ['Transfer learning using an EfficientNet B0 pre-trained model on a dataset, achieving 66 seconds per epoch on a 16-inch MacBook Pro and 21 seconds per epoch on a MacBook Air, with 750 training images and 2,500 testing images.', 'Noticing the need to lower the batch size significantly on M1 machines to fit the pre-trained model into memory.', 'The MacBook Air outperforming the MacBook Pro in terms of processing speed for the current task.']}], 'duration': 211.856, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01310016.jpg', 'highlights': ["MacBook Air exhibits faster per-epoch training times of 7-8 seconds compared to MacBook Pro's 8 seconds.", 'MacBook Pro completes training with only five epochs on 750 images, while MacBook Air finishes training relatively faster.', 'Transfer learning using EfficientNet B0 achieves 66 seconds per epoch on a MacBook Pro and 21 seconds per epoch on a MacBook Air, with 750 training images and 2,500 testing images.', 'Need to lower the batch size significantly on M1 machines to fit the pre-trained model into memory.']}, {'end': 1882.374, 'segs': [{'end': 1610.443, 'src': 'embed', 'start': 1578.329, 'weight': 5, 'content': [{'end': 1582.571, 'text': 'So this is specifically telling the M1 chip to run this experiment on its GPU.', 'start': 1578.329, 'duration': 4.242}, {'end': 1588.853, 'text': "It'll be interesting to see if we get sped up results by taking advantage of this machine's GPU.", 'start': 1583.411, 'duration': 5.442}, {'end': 1593.275, 'text': 'We are almost finished on the feature extraction test on this one.', 'start': 1590.054, 'duration': 3.221}, {'end': 1599.118, 'text': 'Okay, finally, we can go to test three on the 16-inch MacBook Pro.', 'start': 1594.076, 'duration': 5.042}, {'end': 1600.779, 'text': 'This is where this might shine.', 'start': 1599.138, 'duration': 1.641}, {'end': 1610.443, 'text': "um, because we're using apple's framework ml compute, to tell it to use the g, the dedicated gpu, on this machine.", 'start': 1601.978, 'duration': 8.465}], 'summary': "Running experiment on m1 chip's gpu for faster results, testing on 16-inch macbook pro.", 'duration': 32.114, 'max_score': 1578.329, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01578329.jpg'}, {'end': 1679.742, 'src': 'embed', 'start': 1652.325, 'weight': 0, 'content': [{'end': 1655.308, 'text': 'Python is using 50% of the CPU on the MacBook.', 'start': 1652.325, 'duration': 2.983}, {'end': 1660.132, 'text': 'And 70% of the CPU on the MacBook are 16-inch.', 'start': 1656.469, 'duration': 3.663}, {'end': 1663.275, 'text': 'That just jumped about 500%.', 'start': 1660.893, 'duration': 2.382}, {'end': 1664.255, 'text': 'So making some headway.', 'start': 1663.275, 'duration': 0.98}, {'end': 1666.858, 'text': 'The 16-inch is finally clawing its way back.', 'start': 1664.275, 'duration': 2.583}, {'end': 1668.888, 'text': "I'm gonna get out of activity monitor.", 'start': 1667.366, 'duration': 1.522}, {'end': 1671.551, 'text': 'So the MacBook Pro is finished.', 'start': 1669.369, 'duration': 2.182}, {'end': 1676.378, 'text': "So it's finished test three, the benchmark, averaged out at about 26 seconds per epoch.", 'start': 1671.812, 'duration': 4.566}, {'end': 1679.742, 'text': 'The MacBook Pro 16 inch is catching up to the MacBook Air.', 'start': 1676.458, 'duration': 3.284}], 'summary': 'Python usage on macbook: 50%, 16-inch at 70%, 500% jump, macbook pro finished test three with 26 seconds per epoch, 16-inch catching up to macbook air.', 'duration': 27.417, 'max_score': 1652.325, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01652325.jpg'}, {'end': 1808.302, 'src': 'heatmap', 'start': 1744.471, 'weight': 3, 'content': [{'end': 1753.619, 'text': "feeling like I mean it's hard to justify the cost of this machine for the things that I would want to be doing in terms of day to day.", 'start': 1744.471, 'duration': 9.148}, {'end': 1758.182, 'text': 'Maybe it edges out slightly, definitely on the video editing, on the export.', 'start': 1754.139, 'duration': 4.043}, {'end': 1763.567, 'text': "It lost on the CreateML app, so if you're using the CreateML app at all, maybe not the.", 'start': 1758.402, 'duration': 5.165}, {'end': 1767.129, 'text': "I mean, I think Apple's M1 chip has been optimized for the CreateML app.", 'start': 1763.567, 'duration': 3.562}, {'end': 1770.692, 'text': 'Battery life, these two just absolutely crushed it.', 'start': 1767.75, 'duration': 2.942}, {'end': 1777.317, 'text': "And then if you're running native TensorFlow code, I mean, the M1 laptops, even a MacBook Air is able to run.", 'start': 1771.193, 'duration': 6.124}, {'end': 1782.345, 'text': 'native TensorFlow code at a respectable pace.', 'start': 1779.043, 'duration': 3.302}, {'end': 1790.929, 'text': "So I'm gonna probably also do these experiments on a Colab instance and then put together the results in a table, so stay tuned for that.", 'start': 1782.945, 'duration': 7.984}, {'end': 1797.552, 'text': "But what do you reckon, mate? Camera wham? You've seen all these experiments happen, mate.", 'start': 1792.189, 'duration': 5.363}, {'end': 1808.302, 'text': "What's your feelings about them? Yeah, two laptops.", 'start': 1798.052, 'duration': 10.25}], 'summary': "Apple's m1 chip excels in video editing, export, battery life, and running native tensorflow code, but loses on the createml app.", 'duration': 46.458, 'max_score': 1744.471, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01744471.jpg'}, {'end': 1852.569, 'src': 'embed', 'start': 1824.39, 'weight': 6, 'content': [{'end': 1828.693, 'text': "I mean, they're both, all three of them are going to be able to write wordpressing documents.", 'start': 1824.39, 'duration': 4.303}, {'end': 1836.555, 'text': "I'm probably going to permanently set this one up as a desktop and then travel around with a MacBook Pro or a MacBook Air with the M1.", 'start': 1829.153, 'duration': 7.402}, {'end': 1837.636, 'text': "I haven't quite decided yet.", 'start': 1836.575, 'duration': 1.061}, {'end': 1843.637, 'text': 'I mean, in terms of MacBook Pro versus MacBook Air M1, the only difference was on the video render.', 'start': 1837.676, 'duration': 5.961}, {'end': 1845.398, 'text': "But we'll put the numbers together.", 'start': 1843.877, 'duration': 1.521}, {'end': 1852.569, 'text': "I'll put them into some sort of article, and I mean this one failed the first test.", 'start': 1846.865, 'duration': 5.704}], 'summary': 'Comparison between macbook pro and macbook air m1 for wordpressing, with focus on video render performance.', 'duration': 28.179, 'max_score': 1824.39, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01824390.jpg'}], 'start': 1522.592, 'title': 'Macbook and m1 laptop performance', 'summary': 'Presents a benchmark test of training a machine learning model on different macbook models, with quantifiable data on the training time, gpu and cpu utilization, and battery consumption. it also compares the performance of m1 laptops in various tasks, highlighting the superiority in video editing and battery life, but encountering issues in running createml app, and mentions the ability to run native tensorflow code and plans for future experiments.', 'chapters': [{'end': 1744.471, 'start': 1522.592, 'title': 'Macbook performance benchmark on ml training', 'summary': 'Presents a benchmark test of training a machine learning model on different macbook models, with quantifiable data on the training time, gpu and cpu utilization, and battery consumption, showcasing the performance differences among the devices.', 'duration': 221.879, 'highlights': ['The MacBook Pro finished test three with an average of 26 seconds per epoch, outperforming the MacBook Air, while the 16-inch model was catching up. The MacBook Pro achieved an average of 26 seconds per epoch, showcasing its superior performance compared to the MacBook Air, while the 16-inch model was also showing progress.', 'Python 3.8 on the MacBook Pro was utilizing about 88% of the GPU, demonstrating significant GPU usage during the experiment. Python 3.8 on the MacBook Pro utilized about 88% of the GPU, indicating substantial GPU usage during the experiment.', 'The battery consumption was detailed, with the MacBook Pro and MacBook Air still having 35% and close to 40% battery left, respectively, while the 16-inch model experienced a large power consumption, dropping from 100% to 65%. The battery usage was highlighted, with the MacBook Pro and MacBook Air still retaining significant battery percentages, while the 16-inch model exhibited a substantial power consumption, dropping from 100% to 65%.']}, {'end': 1882.374, 'start': 1744.471, 'title': 'M1 laptop performance review', 'summary': 'Compares the performance of m1 laptops in various tasks, highlighting the superiority in video editing and battery life, but encountering issues in running createml app. additionally, it mentions the ability to run native tensorflow code and plans for future experiments.', 'duration': 137.903, 'highlights': ['The M1 laptops excel in video editing and offer impressive battery life, outperforming other tasks like running native TensorFlow code.', 'The CreateML app performance is subpar on the M1 laptops, indicating potential optimization issues with the M1 chip.', 'The chapter plans to conduct further experiments on a Colab instance to compare results, providing valuable insights for readers.', 'The author expresses uncertainty between using a MacBook Pro or MacBook Air M1 for daily use, emphasizing the importance of video render performance in the decision-making process.']}], 'duration': 359.782, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/oqnq8miWVq0/pics/oqnq8miWVq01522592.jpg', 'highlights': ['The MacBook Pro achieved an average of 26 seconds per epoch, showcasing its superior performance compared to the MacBook Air, while the 16-inch model was also showing progress.', 'Python 3.8 on the MacBook Pro utilized about 88% of the GPU, indicating substantial GPU usage during the experiment.', 'The battery usage was highlighted, with the MacBook Pro and MacBook Air still retaining significant battery percentages, while the 16-inch model exhibited a substantial power consumption, dropping from 100% to 65%.', 'The M1 laptops excel in video editing and offer impressive battery life, outperforming other tasks like running native TensorFlow code.', 'The CreateML app performance is subpar on the M1 laptops, indicating potential optimization issues with the M1 chip.', 'The chapter plans to conduct further experiments on a Colab instance to compare results, providing valuable insights for readers.', 'The author expresses uncertainty between using a MacBook Pro or MacBook Air M1 for daily use, emphasizing the importance of video render performance in the decision-making process.']}], 'highlights': ["The M1 chip's performance is compared to a top-of-the-line 16-inch MacBook Pro, questioning the value of the previous investment.", 'MacBook Air with M1 chip finishes rendering a 2-hour, 40-minute video in 38 minutes, losing only about 15% of battery, demonstrating impressive performance and efficiency.', 'The neural engine makes ML up to 11 times faster than the previous generation, showcasing significant performance improvements.', 'The MacBook Air outperforms the 13-inch MacBook Pro and the 16-inch MacBook Pro in training a machine learning model, finishing the task in under 12 minutes with 68% battery left.', 'The tests include running a basic convolutional neural network, transfer learning with an efficient net, and a benchmark on the TensorFlow macOS, aiming to gauge the performance improvements of running TensorFlow code natively on the M1 chip.', "MacBook Air exhibits faster per-epoch training times of 7-8 seconds compared to MacBook Pro's 8 seconds.", 'The MacBook Pro achieved an average of 26 seconds per epoch, showcasing its superior performance compared to the MacBook Air, while the 16-inch model was also showing progress.']}