I had this questions come into my head becasue I was watching a video of someone zooming into a microchip and they pass a human hair and continue zooming in an incredible amount. I’ve heard that some of the components in microchips are the size of DNA strands which is mind boggling. I also watched a video of the world’s smoothest object in which they stated that normal objects are no where near as smooth because if you blew them up in size the imperfections would be the size of Mount Everest. Like if you blew a baseball blew up to the size of earth it would have huge valleys and mountains. It wouldn’t be perfectly smooth across. So my question is how are these chip components the size of DNA not affected by these imperfections. Wouldn’t transistors not lay flat on the metal chip? How are they able to make the chips so smooth? No way it’s a machine press that flattens the metal out that smooth right? Or am I talking about two different points and we haven’t gotten that small yet?
In: Engineering
They aren’t.
For instance, Intel does ‘binning’ on its processors. They have more than one chip now, but in the good old days the i3, i5 and i7 were all the exact same chip. The only difference was in performance testing – the chips that ended up binned as i3 failed to measure up but were within the tolerances of their i3 line, while the i7 are the best ones. Same with overclocking; chips which have the performance to be overclocked are unlocked, so the most near-perfect chips ended up as i7k, while the most flawed but still commercially acceptable chips ended up as locked i3s.
You’d have to get more specific with exact chips to know what they do with ones that are out of acceptable tolerances, though; some are destroyed or recycled as much as possible, rather than sold or shipped as lesser versions of the main chip line because their tolerances are very, very strict and specific.
They don’t. The error rate on microchips is fairly high, precisely because they’re so hard to manufacture. They are, by a pretty wide margin, the most complex mass manufactured devices devised by humanity.
Some chips fail outright. Some don’t work as well as others at speed, and that’s how we get different speed chips.
Nothing lays flat on the chip; they’re complex 3D structures when you zoom in. They are manufactured by insanely sophisticated equipment.
At this level the work, transistors are not placed. The transistors are atoms thick and wide and engineered directly into the silicon wafers by doping the silicon in those locations to turn the silicon in that location into a transistor.
The process is refined, but thinking there are no imperfections is not correct. Many CPUs you buy are the exact same chip, it just failed it primary tests, so they ‘turn’ the chip down, maybe deactivate a core, cut out 1 level of cache and now it passes a lower test and gets sold as a lower chip. That i5 you buy, may not have been planned to be that specific i5.
They make it look like there are no imperfections, it not that. They are just real good at not wasting ‘bad’ chips by selling them as lower tier items.
How they are made is wild. There are videos on it. But it works similarly to how old 35mm film is developed on paper with enlargers in dark rooms. But the opposite. In a nutshell the cpu is built using layers of chemicals that change composition when hit with powerful UV light. A large version of the cpu is printed as a filter in front of this light and then uses lenses that focused much smaller on the surface of the cpu. Then another chemical layer is applied that is slightly different and uses a different filter and the process repeats. The results are super tiny transistors laid out how they need to be. And as others said, failed chips are thrown out, poor performing but working chips are sold for cheap, and the more perfect chips are priced much higher.
Here is absolutely fantastic video that breaks down the manufacturing process for microchips Into semi-ELI5 concepts:
Hope you find it as interesting as I did if you have the time to watch it.
The tolerances for manufacturing are essentially “perfection”. Yes everything is held perfectly level. Yes the wafers are polished to absolute flatness. All of the machines are mounted to resist any sort of external vibration. You don’t dare bring anything containing copper or that touched copper anywhere near the non-copper areas of the factory. It isn’t something that someone woke up one day and said “let’s make a billion transistors for one chip”. It was all iterative and learning how to do things better, and better technology created tighter tolerances and more complex and larger designs.
And there’s still defects but there’s also redundancy built into the design so you don’t have to throw the whole thing away just because a handful of transistors caught a defect.
None of them are perfect.
Your lower tiered chip may very well have been a newer higher tiered chip with too many imperfections.
When you hear about overclocking a gpu it’s because they are all different and capable of more than it is limited to. The idea is to push yours as hard as it can go.
This is often called the “silicon lottery” because you might just be lucky and have the best ever manufactured, or one that just barely makes the cut.
They don’t. I know a guy that made his zillions building equipment that chip manufacturers use to test whether their chips are faulty.
Chips are made in ultra-clean facilities, and they still have a defect rate high enough to serve as the input into other industries, such as photovoltaics.
Monocrystalline solar cells are often made of chip wafers which have too high of a defect rate to be worth slicing up into individual chips. They abrade off all of the chip etchings, and convert the recycled wafer into a high efficiency photovoltaic unit. PV materials don’t need the ultra high purity monocrystalline silicon used in chip manufacturing to work, though the monocrystalline silicon has substantially higher performance than polysilicon PV material. However, it is not cost effective to make such ultra purity silicon for photovoltaics, so they take the rejects from the chip industry, which are more than good enough for PV use, and recycle them as PV materials. This is a win-win arrangement. The chip makers don’t end up wasting high purity silicon, and the PV makers don’t have to grow monocrystalline silicon from scratch.
# Wafer World | [The Rise of Silicon Wafer Recycling in Semiconductor Manufacturing ](https://www.waferworld.com/post/the-rise-of-silicon-wafer-recycling-in-semiconductor-manufacturing)
That’s the neat part, you don’t.
Well of course they try to do it as perfectly as possible, but its unavoidable. This is why you see the big chip makers designing their chips a lot differently in modern times compared to how they used to.
It wasn’t that long ago really, where if you wanted to make the best, fastest chip, you just manufactured the biggest one you can. It’s called a “monolithic die” for that reason. If you take the heatsink off a high end GPU, you’ll find a giant piece of silicon. If you take one off a low end one, you’ll find a small one. Part of what makes a huge monolithic die expensive is if there’s a defect on it, they have to toss out (pulling a number out of my hat) 25% of a wafer for the one mistake. But if the wafer is all small chips, they might only be tossing out 5% of the wafer.
The new hotness is instead of making one monolithic die, you make smaller chips and connect them together, to make up for the fact they are individually worse than a big chip.
As for avoiding flaws to begin with, that’s more in their blackbox of trade secrets. There’s videos for like, 90s era chips on youtube that can explain some of it, but they were working on much larger process nodes.
Latest Answers