Top all-time · #16133
Henry Ko
Personal blog · henryhmko.github.io/postsaihardwareprogramming
Henry Ko's personal blog at henryhmko.github.io covers ai, hardware and programming. Since 2025, 4 different Hacker News readers have linked it 5 times, most often “TPU Deep Dive”. No feed was found, so read it on the site itself.
Too few sampled comments (3) to say how often HN recommends it.?
Why HN links it
HN readers link Henry Ko's deep dives into Google's TPU hardware, covering the chip's tray design, tensor cores versus GPUs, and the XLA compiler and VLIW scheduling work.
Summarised by Claude from a sample of 3 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- TPU Deep Dive2 people · 3 mentions · first linked 2025-06-22
What Hacker News says?
does that cooling channel have a NEMA stepper on it as a pump or metering valve?[0] If so, wild. That seems like overkill. [0]: https://henryhmko.github.io/posts/tpu/images/tpu_tray.png
Comment linking /posts/tpu/images/tpu_tray.png
on “TPU Deep Dive”Submitted TPU Deep Dive as a story · 451 points
HN discussionTensor cores can help a lot for matrix maths, sure, definitely. They made a big splash in 2017 & have been essential. https://developer.nvidia.com/blog/programming-tensor-cores-c... But it's still something grafted onto the existing architecture, of many grids with many blocks with many warps, and lots and lots of coordination and passing intermediary results around. It's only a 4x4x4 unit, afaik. There's still a lo…
Comment linking /posts/tpu/tpu.html
on “TPUs vs. GPUs and why Google is positioned to win AI race in the long term”Submitted Show HN: Optimizing DeepSeek's NSA for TPUs – A Kernel Worklog as a story · 2 points
HN discussionI also enjoyed https://henryhmko.github.io/posts/tpu/tpu.html https://news.ycombinator.com/item?id=44342977 . The work that XLA & schedulers are doing here is wildly impressive. This feels so much drastically harder to work with than Itanium must have been. ~400bit VLIW, across extremely diverse execution units. The workload is different, it's not general purpose, but still awe inspiring to know not just that they b…
Comment linking /posts/tpu/tpu.html
on “Touching the Elephant – TPUs”
