George Hotz | Programming | can tinygrad whisper? | AMD Rant | Giving up on AMD | Speech to Text
george hotz archiveCited 11× on HN, 2023-2024
95K views · 3 years ago · other · news
A long George Hotz live-stream from June 2023 in which he works on tinygrad and rants about AMD's drivers, ultimately declaring he is giving up on AMD. It also touches on hardware and compiler topics.
Mostly cited as the source of Hotz's AMD driver rant, which some commenters found validating and others found unflattering.
3:08:46George Hotz: Hacking the Simulation & Learning to Drive with Neural Nets | Lex Fridman Podcast #132
Lex Fridman
1.6M views • 6 years ago
1:05:28HC2023-K2: Hardware for Deep Learning
hotchipsvideos
15K views • 3 years ago
10:50:05George Hotz | Programming | Livecoding SLAM | twitchslam | Part1
george hotz archive
2M views • 8 years ago
1:10:58Trends in Deep Learning Hardware: Bill Dally (NVIDIA)
Paul G. Allen School
32K views • 2 years ago
40:37ai.bythebay.io: George Hotz, Self-Driving Lessons from Comma AI
FunctionalTV
123K views • 9 years ago
42:13Clojure is Not Afraid of the GPU - Dragan Djuric
ClojureTV
7K views • 9 years ago
1:59:37George Hotz: Comma.ai, OpenPilot, and Autonomous Vehicles | Lex Fridman Podcast #31
Lex Fridman
875K views • 7 years ago
1:00:34Chuck Moore's Fireside Chat 2011
GreenArraysInc
6K views • 14 years ago
2:55:13Oral History of Gary Davidian, part 2 of 2
Computer History Museum
2K views • 7 years ago
3:50:00George Hotz | Programming | we do coronavirus science aka hacking nature #lockdown part1 | COVID-19
george hotz archive
286K views • 6 years ago
2:03:05GTC March 2024 Keynote with NVIDIA CEO Jensen Huang
NVIDIA
27.4M views • 2 years ago
2:35:53Tensorflow and deep learning - without a PhD by Martin Görner
Devoxx
610K views • 10 years ago
1:15:09Cliff Click @ VMSS16: Bits of Advice For the VM Writer
softdevteamuk
4K views • 10 years ago
25:57GPT3: An Even Bigger Language Model - Computerphile
Computerphile
446K views • 6 years ago
1:37:12[7] Jim Keller, Silicon Wizard 🧙♂️
TechTechPotato
67K views • 5 years ago
2:50:14Simple Code, High Performance
Molly Rocket
463K views • 5 years ago
2:25:52The spelled-out intro to neural networks and backpropagation: building micrograd
Andrej Karpathy
4.4M views • 4 years ago
1:05:25Robocalypse Now: Using Deep Learning to Combat Cheating in Counter-Strike: Global Offensive
GDC Festival of Gaming
35K views • 6 years ago
3:29:25THE NVIDIA AI GPU BLACK MARKET | Investigating Smuggling, Corruption, & Governments
Gamers Nexus
3.5M views • 1 year ago
51:50[11] Mike Clark, AMD Zen Chief Architect
TechTechPotato
27K views • 5 years ago
11 comments from Hacker News
So for reference, this geohot is George Hotz who has a company Tiny Corp [0] in the space. Among a long list of things, he had a moment in the spotlight recently for a long rant [1] where he "gave up on AMD" because their drivers sucked more than he expected. The situation is reasonably complex - AMD have some programmers working on ROCm who seemed to be operating at standard fare when that is really not what AMD needs right now (I personally, suspect there is/was a PM in a key position who didn't "get it", although I am not sure what it is either). As far as I know it got the attention of Lisa Su. I'm cheering him on, even though his complaints were a little melodramatic. My experience is the driver technically supports everything I could possibly want. The problem is if I spend an evening trying to do anything with OpenCL or ROCm the kernel hard-locks and I go to bed early. If the problem inside is what it looks like from the outside (repeating myself, a key manager somewhere just doesn't get the space) they really need some pressure from grumpy customers like George to realign their software development process. That context might be related to this particular case. I see a suspicious folder called "crash" in this repo. [0] link [1] ▶ this video - as I recall he ran the demo suite and it crashed.
> according to George Hotz In this video from June, George Hotz says to go with "3090s over 4090s. 3090s have NVLink... 4090s are $1600 and 3090s are 750. RAM bandwidth is about the same." Has he changed his recommendations since then? ▶ this video
They're talking about the meltdown he had on stream [1] (in front of the mentioned pirate flag), that ended with him saying he'd stop using AMD hardware [2]. He recanted this two weeks after talking with AMD [3]. Maybe he'll succeed, but this definitely doesn't scream stability to me. I'd be wary of investing money into his ventures (but then I'm not a VC, so what do I know). [1] ▶ this video [2] link [3] link
He flamed out, then is back after Lisa Su called him (lmao) link ▶ this video link link On a personal level that youtube doesn't make him come off looking that good... like people are trying to get patches to him and generally soothe him/damage control and he's just being a bit of a manchild. And it sounds like that's the general course of events around a lot of his "efforts". On the other hand he's not wrong either, having this private build inside AMD and not even validating official, supported configurations for the officially supported non-private builds they show to the world isn't a good look, and that's just the very start of the problems around ROCm. AMD's OpenCL runtime was never stable or good either and every experience I've heard with it was "we spent so much time fighting AMD-specific runtime bugs and specs jank that what we ended up with was essentially vendor-proprietary anyway". On the other other hand, it sounds like AMD know this is a mess and has some big stability/maturity improvements in the pipeline. It seems clear from some of the smoke coming out of the building that they're cooking on more general ROCm support for RDNA cards, and generally working to patch the maturity and stability issues he's talking about. I hate the "wait for drivers/new software release bro it's gonna fix everything" that surrounds AMD products but in this case I'm at least hopeful they seem to understand the problem, even if it's completely absurdly late. Some of what he was vi
Hmm. I found ▶ this video and link I feel a lot better about my journey with AMD now; there seemed to be some major issues with their GPU drivers. Now I know it wasn't just me.
Even with this very notable reconciliation, the freewheeling YT video: ▶ this video And a harder look at the ML space outside of PyTorch eager mode has made me more skeptical of tinygrad.
short answer is they have somewhat competent hardware but software sucks or you can watch george hotz rant about how amd driver sucks ▶ this video
Here is a timestamped clip of George Hotz Twitch stream from 6 days ago making the same point about rationalizing his own emotional state of powerlessness in Industrial Society and its Future ▶ this video
When I got to slide 5 I LMAO when they wrote "LLVM doesn’t remove the dead loop" in bright red it reminded me so much of the geohot drama on his batshit twitch archive on youtube where some AMD GPU driver author says something about 'remove the dead loop' in their patch to fix the driver which was causing hotz's linux to actually kernel panic. Please read slide 5 here and then watch ▶ this video it's hilarious especially if you never heard 'remove dead loop' before these two times. To be honest, I still don't know what 'remove dead loop' actually means.
It seems he reached out to AMD and Su, but got back a horrible patch and wasted a lot of time and he "gave up on AMD", "this driver will never, ever be good". ▶ this video
> "... aww shit they have alignment, ok well, we don't want any alignment. our AIs are unaligned" lmao ▶ this video