George Hotz | Programming | coding in UOps on AMD MI350X tickets.comma-con.com/events/comma/1859964
Date of the stream 1 Nov 2025. COMMA_CON 2025 - https://tickets.comma-con.com/events/comma/1859964 from $999 buy https://comma.ai/shop/comma-3x & best ADAS system in the world https://openpilot.comma.ai & - https://tinycorp.myshopify.com Source: - https://github.com/tinygrad/tinygrad - https://docs.tinygrad.org Follow for notifications: - https://twitch.tv/georgehotz Support George: - https://twitch.tv/subs/georgehotz Order tinybox: - https://tinycorp.myshopify.com Chapters: TBD Official George Hotz communication channels: - https://geohot.com - https://twitter.com/realGeorgeHotz - https://instagram.com/georgehotz - https://tinygrad.org - https://geohot.github.io/blog - https://github.com/geohot We archive George Hotz and comma.ai videos for fun. Follow for notifications: - https://twitter.com/geohotarchive Thank you for reading and using the SHOW MORE button. We hope you enjoy watching George's videos as much as we do. See you at the next video.
Video Chapters
- 3:03 Benchmarking AMD's new MI350X GPU
- 4:21 Hitting the manual to unlock tensor cores
- 5:38 The simple mistake holding back all the performance
- 7:43 Finding the magic instruction for 2x performance
- 8:36 Writing our first GPU kernel in tinygrad
- 10:18 Building a matrix multiply from absolute scratch
- 11:28 Debugging and fixing our custom matmul
- 12:47 The ultimate GPU programming environment?
- 14:52 tinygrad vs. TVM: Kernel syntax comparison
- 16:09 Deep dive: How the 'after' keyword works
Original Output
3:03 Benchmarking AMD's new MI350X GPU 4:21 Hitting the manual to unlock tensor cores 5:38 The simple mistake holding back all the performance 7:43 Finding the magic instruction for 2x performance 8:36 Writing our first GPU kernel in tinygrad 10:18 Building a matrix multiply from absolute scratch 11:28 Debugging and fixing our custom matmul 12:47 The ultimate GPU programming environment? 14:52 tinygrad vs. TVM: Kernel syntax comparison 16:09 Deep dive: How the 'after' keyword works Timestamps by StampBot 🤖 (340-george-hotz-programming-coding-in-uops-on-amd-mi350x-tickets)
Unprocessed Timestamp Content
0:10 Welcome back after a long break from my new Hong Kong residency 1:08 My quick thoughts on internet toxicity and recent massive tech layoffs 2:20 Comparing passionate rocket engineers to people who just showed up in tech 3:03 Benchmarking the AMD Instinct MI350X GPU's raw petaflop performance 4:21 Diving into the official AMD manual to understand how tensor cores work 5:38 The simple mistake that was preventing the tensor core from activating 7:43 Discovering a new GPU instruction that gives double the performance for free 8:36 Let's write our first custom kernel using the new tinygrad syntax 10:18 Building the absolute stupidest matrix multiply from the ground up 11:28 Debugging why the custom matmul failed verification and fixing the code 12:47 This is the best development environment for programming GPUs hands down 14:52 Comparing the custom kernel syntax to TVM and highlighting key differences 16:09 A detailed explanation of how the 'after' keyword controls operations Timestamps by StampBot 🤖 (340-george-hotz-programming-coding-in-uops-on-amd-mi350x-tickets)