Turns out metals (in particular copper) are already incredibly good conductors.
This is not true anymore. At leading edge process nodes (i.e., smaller conductor pitch), the bulk resistivity of copper changes dramatically because it is increasingly dominated by ballistic electron scattering on the interfaces and grain boundaries: surface area has decreased as a fraction of volume and the average grain size has shrunk to fit inside the interconnect lines.
Alternative metals like Cobalt and Ruthenium have been proposed and to some extent used in production as an alternative to Cu for the low interconnect layers, but they just arrest the trend -- the fact is that the interconnect resistance is much higher than it used to be.
It's also why a lot of the big players like TSMC and Intel are investing heavily in backside power delivery.
Here is an accessible article on the topic:
https://www.fabricatedknowledge.com/p/backside-power-deliver...
(Also: high-voltage DC power is commonly used for long-range transmission and does not have coupling issues)
Resistance in the metal wires is not so much a problem for power as it is for signal propagation delay. That is the biggest problem with R skyrocketing in M0/1.
And backside power delivery is to relieve congestion in the interconnects. That's typically one of the limiting factors in complex logic designs now, transistor densities of under 70% aren't uncommon.
That's what "Removing the power signal and signal line to just a signal line would free up space for more transistors." line is about in the link. Power delivery losses aren't nothing, but they aren't a big fraction of energy in today's CPUs.
Eliminating most wire delay would be a huge benefit though, and could be a pretty big revolution. Not just within the core or on the die. You're still going to have to switch those transistors and burn current on leakage though, so no cold CPUs just yet.
Technically true but not applicable... Super conducting CPUs already exist (at below room temp), and they don't use semiconductors. They don't have an exact transistor equivalent but use digital logic circuits such as Quantum Flux Parametron (QFP) and rapid single flux quantum (RSFQ), which are comprised of Josephson Junctions [0] which are made from the same super conducting material. Relying on quantum effects of a gap between two conductors rather than specific material properties.
So while it is true that CMOS are limited by the resistance of both the conductor interconnects AND the semiconductors. Super conducting circuits only appear to need a conductor... There are other drawbacks, such as larger feature sizes due to the nature of super conducting materials, and QFP and RSFQ might be more complex? but this is potentially offset by huge advantages: zero resistance buys not only power efficiency and minimum heat generation (as required by landauer limit), but also allows circuits to be driven at a higher clock, so far demonstrated to 770GHz for one CPU [1]
[0] https://en.wikipedia.org/wiki/Josephson_effect
[1] https://spectrum.ieee.org/superconductor-logic-goes-lowpower
That's not a CPU, it's a simple circuit. From the article, "Simple superconducting logic circuits have been shown to operate at speeds of up to 770 gigahertz."
Exotic semiconductor transistors can switch that fast, record being over 800GHz I think. Regular transistors of the type in your phone can probably do around 200-300GHz. Not to say there could not be revolutionary designs opened up with superconducting wires or switches, just the switching speed itself doesn't necessarily tell us one way or the other.
Actually it's probably more like 200 than 300, I haven't seen 3nm models but transistors are still getting faster.
I'd be interested to know from any experts whether there is potential for fundamentally higher switching speed in super conducting circuits. As you suggest, they are not necessarily directly indicative of the clock speed of a specific real world circuit or CPU, especially considering super conductors don't even use transistors and need to use different building blocks for equivalent logic... but as a rough indication based on order of magnitude of switching speed would be interesting.
And to this point, HVDC has been slowly rising as a viable alternative beyond just undersea transmission.
A mistaken belief is that AC is more efficient than DC. What AC is is more easy to transform from one voltage to another (until somewhat recently). That makes it easier to run AC at 1 MV and then step it down to 240V for residential applications.
The cost is as AC voltage goes up, capacitive resistance increases. This is part of the reason why high voltage lines have such huge towers with the lines far apart.
An HVDC line, however, can be put underground (or water) without suffering power losses. It's voltage can go well above that of AC voltages with the only limit being how much insulation we need for the line.
Super conductors are nice, but only in the "now we can run 1 billion amps at 100V" sense to avoid the capactive resistance. Without any sort of special materials you get most of the benefits of super conductors by using HVDC. The only real downside is high voltage DC is still super dangerous. Cut the insulation and you've got something that can arc meters whereas a superconductor at lower voltages would be about the same danger as any other conductor at lower voltages.
That’s a pretty high current density compared to what’s feasible with copper — don’t get me wrong — but a billion amps would still require a pretty huge cable even with a superconductor.
So unless someone comes up with super conducting welding I guess that would mean no junctions.
DC is also used because a conductor can carry more current per kg than AC due to not having any skin loss. You can use smaller wires, less metal, less dollars.
Finally someone is speaking English in these room-temperature superconductor threads that I can actually understand and get excited about!
> None of the things you listed are limited by the conductors in them.
Never mind. :(
So maybe just just t normal conductors like copper, silver, graphene (assuming the latter can be commercially made to surpass silver in RF surface resistance).
So what do we stand to gain?
At least a slight improvement in efficiency of transmission. And a huge improvement due to simpler heat dissipation needs.
So this might mean way faster battery charging - like 500 miles worth of charge in 5 mins. We’re mainly limited by how hot the batteries get during charging.
(1) https://www.tue.nl/en/news-and-events/news-overview/13-07-20...
Earlier discussion: https://news.ycombinator.com/item?id=36954783
Or store in coils, as "magnetic energy"(forgot the name for it)
Yes, but superconductors don't have that limitation, do they? You just dump current into them.
Currently they're only feasible as high quality power sources for fabs and other industrial uses because of the operating costs of cooling the superconductors.
https://www.eaton.com/gb/en-gb/catalog/electronic-components...
I have 4 at home and blows people's minds when you melt a wire with it. Incredible rapid current delivery.
You also get losses from practical usage - i.e. no one can build a 3V supercapacitor that has decent endurance (you can totally build one which will work, but you're rating it knowing that every cycle is damaging it).
Heat?
... you cannot just dump unlimited current through a superconductor. Once you exceed the critical current density, your superconductor becomes a regular conductor.
Similar speeds for charging are impractical because of the power spikes required even if the battery could take it.
You cannot have instant charging, that's not pedantic, that's the discussion, and in any case there is a practical limit to how fast it can happen for similar reasons.
All the work being done on smart home EV chargers that automatically schedule the right time (controlled by grid) to charge overnight are because even at current speeds this wreaks havoc on the electrical subsystems and grid if everyone plug their EVs in at the same time in the evening...
What we need is sufficient infrastructure for ev charging.
As often in engineering, the sweet spot is where multiple factors have ~equal contributions, so even a 50% win in one of the factors can't give you that much benefit.
For high voltage AC power lines it's Ohmic losses, corona discharge losses, and inductive losses.
It could definitely open up the road to long distance high capacity power lines, but somewhere along the line reality is going to make things difficult. It's not magic after all, just sufficiently advanced technology.
Doesn't sound too spectacular but it took them over a year to fix all the damage.
External researchers have not even begun to purify and test the limits of this material.
Can't superconductor help here?
It's hyper-specialized tech, so it'd probably take over a decade from now to be seen in useful, everyday technologies.
But yet there's still so much that could be accelerated even with current hardware, but isn't.