6 ms·
Most likely silicon cruicibles. The reason there is no backup power is because this would've taken building an own power plant.
by baybal2 9y ago
Most likely silicon cruicibles. The reason there is no backup power is because this would've taken building an own power plant.
- CaliforniaKarl 9y agoLink? I've not heard of those before, and am interested to learn more!
- selectodude 9y agoThey're just the things that the silicon crystals are grown in. The furnace cools off and the whole thing needs to be re-started/replaced.
- baybal2 9y agoIt does not work like that. The whole thing needs to be a single crystal. If a minute temperature change creates a second grain in the melt, it is all gone, along with the crucible.
- CaliforniaKarl 9y agoIs there any place online, on a book, you suggest I go to in order to watch or read more about this? That all sounds remarkably interesting, especially for something which impacts all of our lives so much.
- baybal2 9y agohttps://youtu.be/8QKzS_w_Ko0 https://youtu.be/8QKzS_w_Ko0
- HankB99 9y agoWhen I worked at what was then Inland Steel (Indiana, US) they had a coal fired 60 Hz power station. It did not produce all of their power but it did produce a significant amount. The reason it was there was that they could use steam after it had passed through a portion of the turbine for other plant processes. Chip fabs are in a different situation. My understanding (and I am far from an expert) is that typical power interruptions such as the momentary loss when a breaker trips and recloses will interrupt production. A quarter second might as well be 30 minutes as it will cause all motor controllers to drop out. Chip fabs need something akin to a giant UPS (or alternative power feed that they can switch to in a few milliseconds. Such equipment exists and is costly. The business case is the cost of the power conditioning vs. the expected lost production due to potential outages. Perhaps Samsung did have something that could ride out a short outage but not the one they faced. In that case there is a vast difference between 1/4 second drop and a 30 minute outage.
- greglindahl 9y agoLots of big power users have their own power plants, with the grid as backup. This is especially good if there's someone nearby who can use the waste heat from the power plant. Looks like the melting point of pure silicon is high enough that it's way above the temperature of the power plant's waste heat.
- wbkang 9y agoOk people keep saying there was no backup power but there was, it was just not enough to power everything. Source: http://m.yna.co.kr/kr/contents/?cid=AKR20180316130000003 http://m.yna.co.kr/kr/contents/?cid=AKR20180316130000003 maximum estimated amage was around $50mil usd. It says it's their internal transformer that failed. Up to 60k wafers were affected but not all of them are unusuable.
- avs733 9y agoNot likely. I don't believe Samsung makes their own bare wafers and I know of no fab that makes wafers on site. I don't think the demand model would work to make that vertically integrated. A couple crucibles can easily supply a Fab and the bulk cycles of crucibles is uite different from the more unitized fab production.
- baybal2 9y agoWell, this is the only explanation I have for loss on such scale, and the fact that wafer manufacturing is the only thing that can't be usually backed up with UPSes. All DRAM fabs use a lot of proprietary pixie dust treatment for wafers these days, and it is not looking improbable that they had a wafer manufacturing on site. If they were to loose power on a stepper line, they would've only lost wafers being currently processed. On my memory, there was a number of wafer shortages when Taiwanese wafer makers had similar incidents. As far as I was told, if a crystal forms on a crucible wall, touches it, or moreover the whole thing freezes, a big part of equipment becomes a write off, and after that it will take few months for a company to resume production.