We have just started 2024 and there are already exciting announcements. The clever people at Georgia Tech in Atlanta have built the first scalable semiconductor using a graphene base. Graphene, a wonder product, is not a scalable semiconductor on its own, so they bonded silicon carbide, or what we call carborundum, to a layer of graphene creating the necessary bandgap to have a working switch. A switch means binary and from there they can make wafers like those currently used in the chip manufacturing process to make CPUs and other devices.
- This new graphene approach has a number of benefits. Power requirements are lower and the graphene structure allows for better conductivity, which means better speed. The other previously limiting factor was a heat dissipation problem. Graphene is excellent at that as well. There needs to be more development before we will see it in our homes as silicon is the dominant force, but chips based on this technology may appear in the next 3-5 years. This could also save Moore's Law.
- I recently watched a presentation from physicist Sabine Hossenfelder on IBM's new 1121 qubit chip with roughly a 50% increase in qubit density from their last one. Readers will remember that a qubit is the operating heart of a quantum computer. Sabine pointed out that the number of qubits "isn't all that relevant". The underlying question is if you can operate all of them as you like. This means the ability to set up a problem definition and know in advance how bad the error rate might be. IBM didn't reveal that info other than to note it has similar performance as the earlier chip which had 433 qubits.
- Interestingly, IBM still spends most of its time referring to the even older 133 qubit version of the chip. IBM revealed in 2018 that the number of qubits is not necessarily a good measure of capability and some time back coined a new term "quantum volume". Again, their quantum volume updates ended in 2022 back with their 16 qubit chip, so no one knows what you can actually do with this new chip. According to Sabine, for a commercially viable quantum computer, you will need about a million well-behaved qubits. Looks like that predicted quantum revolution is some way off in the future for now. It also means that quantum AI is also just a concept at this point.
- If you've been around computers as long as I have then the name Nicklaus Wirth, the creator of the Pascal programming language, will be a familiar one. His Algorithms + Data Structures = Programs book, of which I still have a copy somewhere, is a whopping AUD$375 (8,800 baht) on Amazon Australia. The Swiss scientist died on New Year's Day a bit before his 90th birthday. He had asteroid 21665 named after him and he won the Turing Award in 1984. Wirth's Law, also known as the law of computer design, states that "software is getting slower more rapidly than hardware becomes faster" and was first coined in 1995. Note that this law contradicts Moore's Law which famously states that "the number of transistors on a chip double even as prices are halved".
- Wirth's point was that software was getting too bloated as memory became less expensive because of lazier programmers. He also noted that complex software was a marketer's dream because every marketer knows that sophisticated is better. He also correctly pointed out that engineers at the time got less and less design and planning time as the demand for quicker software releases increased. He also noted, and any programmer will identify with this, that customers tend to confuse features (nice to have) with functionality (necessary to have). Software vendors and managers try to make their customers happy by inserting all the features they demand. This results in fat software where customers pay for all features but actually use only a handful. Microsoft Office is a classic example. Early in his career he worked on early ALGOL releases, then Pascal when he split from the ALGOL-68 team, and then he went on to create the Modula, Modula-2 and Oberon series of programming languages. Wirth was a true computer giant.
James Hein is an IT professional with over 30 years' standing. You can contact him at jclhein@gmail.com.
