Twelve years ago, my daughter Allie and I conspired to spoof about 500 attendees at an IBM VM user conference in Orlando where I was the keynote speaker. After talking about industry trends and positing the mainframe value proposition for an hour, for my wrap-up I brought my very precocious 8-year-old on stage to pontificate about the future of processor technology. Our quasi-scripted interchange went like this:
Me: Welcome, Allie. I’d like to start by asking you about Moore’s Law. Do you foresee processor chips hitting the performance speed limit anytime soon?
Allie: First of all, I’d like to point out Moore’s Law isn’t really a law, it’s just an observation. The question is about how much longer computer chip manufacturers can keep doubling the chip’s density every 18 months to two years.
OK, so at this point, the attendees started scratching their heads with a quizzical look on their faces, as if to say, “Say what?” We continued.
Me: So, Moore’s Law is an observation about circuit density doubling and not about performance?
Allie: Correct, but when you double the density, performance improvements follow.
Me: So, what will determine if they will be able to continue that pace?
Allie: Fabrication technology must improve in many areas with each successive process generation, such as the move from 0.25-micron design to 0.18.
Me: So, how small is that?
Allie: .25 microns (25/100ths) is about 250 times smaller than the diameter of a human hair.
Me: How can they work with things that small?
Allie: One critical process is photolithography. Photolithography is where short-wavelength light sources are focused with a number of precision lenses and shone through small, transparent masks containing circuit details. This exposes the photoresist on a wafer's surface, which is chemically removed, leaving microscopic details of the circuit pattern on the wafer.
Next, I was supposed to tell Allie that we were out of time but ask if she would be around afterward to answer people’s questions, to which she would have replied, “Actually, this represents 100 percent of my knowledge on the topic, so there will be no need for any questions.” But I simply forgot, so many attendees marveled at what they thought was a child prodigy.
For our script, I took a paragraph on fabrication technology from a 1999 article in PC Magazine and added the Moore’s Law portion. At the time, prognostications about Moore’s Law and hitting processors’ speed limits were plentiful as the next generation of chips would introduce 0.13-micron processes, 193-nanometers (nm) wavelengths, and the use of excimer lasers as the light source. A few years later, we were expecting 0.09-micron process using 157-nm wavelength excimer lasers, and the longer term projection was that by 2011, photolithography would likely require Extreme-UV (EUV) light sources with a wavelength of only 13nm. Much of that actually did happen, but some big challenges in corralling EUV have been finding suitable transparent mask materials for etching that would allow such short wavelengths to pass through, and also to find new reflective lithography processes and optics. As 2011 came and went, all the chip makers grappling with EUV, including Intel, had a track record of slipping delivery dates and revising estimates of how small they hope to go with it, but the persistence will be paying off soon.
What triggered my flashback about Allie’s spoof was the “say what?” factor in something I read recently about memristor technology, which included this from Wikipedia: Hysteresis is the dependence of a system not only on its current environment but also on its past environment. This dependence arises because the system can be in more than one internal state. To predict its future development, either its internal state or its history must be known.
I learned that memristor memory technology was birthed 40 years ago by University of California, Berkeley and that HP has recently created a version that uses alternating layers of titanium dioxide and platinum and lays them crosswise to make itsy bitsy cubes for storing information. Each side of the cube measures 2 to 3 nm. Hence, I found myself daydreaming about being on stage with one of my precocious young granddaughters and asking, “So how big is a nanometer?” and she would reply, “It’s one-billionth of a meter, or roughly one ten-thousandth the thickness of a human hair.” Don’t worry, I would most certainly have her explain one of the resulting properties of memristors, which is the existence of a pinched hysteresis effect. Say what?