Chip Wars by Chris Miller — Part II beating Russia with silicon

Just about everything in this post is from Chip Wars by Chris Miller, some are direct quotes, others are paraphrases.   A few things are my own, but they’re pretty obvious.

Even though I was vitally interested in computers as a neurologist starting in the early 70s and got one as soon as I could afford it (an Alpha Micro which I really couldn’t), Chip Wars covering the early history of silicon based computers taught me a lot.  It starts with Shockley’s invention of the transistor in 1948 and goes from there.   I won’t try to summarize that, but if you want an extremely well written early history of the period, read this book.  It isn’t dry and the personalities of the main characters are well fleshed out.

What I didn’t realize was just how much of the early development of silicon computers was driven by the military.  In particular Defense Advanced Research Projects Agency (DARPA) funded a lot of academics and their research into computation. Today every chip company uses tools from one of 3 chip design companies founded and built by graduates of DARPA programs

Guided munitions during the early Vietnam war used vacuum tubes that were hand soldered (Sparrow III) they broke down 2/3 of the time only 10% hit their target.  Bombs fell 420 feet from target.    Some 800 bombs had tried to take out a bridge in Vietnam and failed.   A set of wings was added to direct the bomb’s flight along with a laser guidance system which worked as follows.  A small silicon wafer was divided into 4 quadrants and placed behind a lens.  The laser reflecting off the target would shine through the lens onto the silicon,.  If the bomb veered off course one quadrant would receive more of the laser’s energy than the others and circuits would move the wings to reorient the bomb’s trajectory so the laser was shining straight through the lens.    A simple laser sensor and a few transistors made the bomb accurate.  This was 1972.  Spoiler alert — we still lost.

Photolithography is a crucial technology for drawing circuits on silicon, and it is mentioned many times throughout the book.   It’s basically a simple idea— turning a microscope lens upside down to make something big look smaller.  It was initiated by  Lathrop at Texas Instruments in 1958  It took thousands of experiments and a lot of interaction with suppliers etc to make it work.   There will be much more about photolithography in the next posts.

When owning a copy machine was a crime and no one without security clearance could use a computer, Russia had no way to educate a truly huge number of computer programmers.  So they basically used espionage to copy our technology.

This didn’t work.   The Soviet copy it strategy was flawed, because they couldn’t scale up the manufacturing process reliably, something Grove at Intel and Chang at Texas Instruments fixated on and spent countless hours improving.  Moreover the US had access to technology in optics, chemistry purified materials.  They knew the temperature at which chemicals needed to be heated or how long photoresist should be exposed to light.  Every step of the process of making chips involved specialized knowledge that was rarely shared outside a specific company — often not written down.    So it couldn’t be stolen.

As Silicon valley crammed more transistors onto silicon chips, building them became steadily harder.    Russia stole the equipment to make them, but they had no way to get spare parts.  The Russian military didn’t trust the chips produced in country, so they minimized the use of electronics and computers in military systems.    The math they put into guidance computers was simpler to minimize the strain on the onboard computer.

The copy it strategy left the Russians 5 years behind.

Russia’s defense chief of staff Ogarkov knew this and said in ’83 to Leslie Gelb  “The Cold War is over and you have won”.  Interestingly, Star Wars begun the same year is nowhere mentioned in the book.  It was derided as implausible, but the Russians knew they couldn’t match it.

The Soviets only customer for computers was the military, while the US had a large civilian market which created companies with a wide variety of expertise in everything needed for them (pure silicon wafers, advance optic for lithography).  The Russians also had no international supply chain. The Gulf War ’91 on Saddam Hussein — US smart weapons (Paveway) using more advanced electrons decimated the best equipment of Russia.

As I said in part I (https://luysii.wordpress.com/2024/03/10/chip-wars-by-chris-miller-part-i/) Chip Wars is really about manufacturing, not the abstract computational problems and programming I was interested in.

The denouemont came in 1991 with the Gulf War. US smart weapons (Paveway) using more advanced electronics decimated the best equipment of Russia.

Things haven’t changed much in Russia. p335

“The fact that Russia faced shortages of guided cruise missiles within several weeks of attacking Ukraine is partly due to the sorry state of its semiconductor industry.”

Next up: The coming competition with China

The neuron as motherboard

Back in the day when transistors were fairly large and the techniques for putting them together on silicon were primitive by today’s standards, each functionality was put on a separate component which was then placed on a substrate called the motherboard. Memory was one component, the central processing unit (CPU) another, each about the size of a small cellphone today. Later on as more and more transistors could be packed on a chip, functionality such as memory could be embedded in the CPU chip. We still have motherboards today as functionality undreamed of back then (graphic processors, disc drives) can be placed on them.

It’s time to look at individual neurons as motherboards rather than as CPUs which sum outputs and then fire. The old model was to have a neuron look like an oak tree, with each leaf functioning as an input device (dendritic spine). If enough of them were stimulated at once, a nerve impulse would occur at the trunk (the axon). To pursue the analogy a bit further, the axon has zillions of side branches (e.g,. the underground roots) which than contact other neurons. Probably the best example of this are the mangrove trees I saw in China, where the roots are above ground.

How would a contraption like this learn anything? If an impulse arrives at an axonal branch touching a leaf (dendritic spine) — e.g. a synapse, the spine doesn’t always respond. The more times impulses hit the leaf when it is responding to something else, the more likely the spine is to respond (this is called long term potentiation aka LTP).

We’ve always thought that different parts of the dendritic tree (leaves and branches) receive different sorts of information, and can remember (by LTP). Only recently have we been able to study different leaves and branches of the same neuron and record from them in a living intact animal. Well we can, and what the following rather technical description says, its that different areas of a single neuron are ‘trained’ for different tasks. So a single neuron is far more than a transistor or even a collection of switches. It’s an entire motherboard (full fledged computer to you).

Presently Intel can put billions of transistors on a chip. But we have billions of neurons, each of which has tends of thousands of leaves (synapses) impinging on it, along with memory of what happened at each leaf.

That’s a metaphorical way of describing the results of the following paper (given in full jargon mode).

[ Nature vol. 520 pp. 180 – 185 ’15 ] Different motor learning tasks induce dendritic calcium spikes on different apical tuft branches of individual layer V pyramidal neurons in mouse motor cortex. These branch specific calcium spikes cause long lasting potentiation of postsynaptic dendritic spines active at the time of spike generation.