The Chip Letter but smaller, cuter and more fun!
TLDR: Announcing a new series of shorter posts that are opt-in if you want to receive them as emails
First of all, a warm welcome to all new readers who have subscribed after reading Demystifying GPU Compute Architectures, which has turned out to be my most popular post so far!
New readers who want to know a bit more about what this newsletter is all about might like to read A Technology of Wonder. It’s a short post that sets out what I’m aiming to achieve and has some recommendations for further reading.
I also wanted to share that I’ve been working on a deep dive into the origins of Google’s ‘Tensor Processing Unit’. It’s proving to be a completely fascinating topic. I hope that you’ll find it just as interesting when it appears in the (hopefully) not too distant future.
Now for a small announcement.
Regular readers will have noticed that posts have been getting longer. The first five posts of 2024 have averaged well over 5,000 words each. That amounts to a third of a book in a little bit more than a month!
This has created a couple of dilemmas, though. First, writing these posts is quite time consuming, which in turn limits the range of topics we can cover. Second, some topics are really interesting, but don’t really merit a 5,000 word post.
So, just as the Semiconductor industry has innovated using ‘chiplets’ we’re going to follow suit with …
“Chiplet” posts which will be a new kind of, much shorter, post.
A Chiplet post might aim to:
Share something of interest : an architecture, a computer, an important document or event, or even a great Substack or blog post;
A follow-up to an earlier post with supplementary information;
Initiate a discussion or ask readers for your help or opinions on a topic.
They will be a mix of historical and current topics. Useful and just interesting. Serious and fun.
The focus will still be on hardware and semiconductors, but I expect that posts will sometimes range more widely across the computing landscape.
I also expect that they will be a bit more informal. They will definitely be shorter!
In some weeks there might be a couple of Chiplets and in others, none at all, depending on what I’ve found over the week and my own time constraints.
Chiplets we be a combination of free and paid but free posts will usually be emailed to and unlocked for paid supporters (thank you to you all) first.
I expect to continue the pace of longer posts at around once a week, although they may not all be 5,000 words!
There is one important thing to note though (sorry for the large print but this is important!) I really, really don’t want to fill inboxes with too many emails.
So receiving ‘Chiplets’ will be opt in.
If you want to receive Chiplets in your Inbox then you’ll need to take action.
It’s easy though…
Just navigate to this link
And toggle the button highlighted below.
You can always ‘opt-out’ again at any time if the volume of emails is too much!
I’ll highlight recent Chiplets in regular ‘full-length’ posts and a link to Chiplets will appear in their own section in the header at the top of the front page of thechipletter.substack.com.
The first Chiplet is definitely at the fun end of the spectrum, and is free for all to read today. It has a young Satya Nadella, demonstrating IBM AS/400 to Excel data transfer, all the way back in 1993.
Chiplets in (various stages of) draft include:
Jean Hoerni’s Patent Notebooks for the ‘planar process’;
RISC-III and RISC-IV, the Berkeley follow-ups to RISC-I and RISC-II;
Chips and early Venture Capital : A brief review of ‘The Power Law’;
The Mostek 5065 - ‘Great Uncle’ of the 6502;
And suggestions from readers for future ‘Chiplets’ are very welcome. A big thank you to all readers who have already made suggestions for topics to cover.
The Chip Letter is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Finally, a big thank you to Fritzchens Fritz for his pictures of AMD Zen4 Chiplets. Readers can view the full collection of his images here.