Power Outage At Samsung's Fab Destroys 3.5 Percent of Global NAND Flash Output (anandtech.com) 103
An anonymous reader quotes a report from AnandTech: A half-hour power outage at Samsung's fab near Pyeongtaek, South Korea, disrupted production and damaged tens of thousands of processed wafers. Media reports claim that the outage destroyed as much as 3.5% of the global NAND supply for March, which may have an effect on flash memory pricing in the coming weeks. The outage happened on March 9 and lasted for about 30 minutes, according to a news story from Taiwain-based TechNews that cites further South Korean reports. The report claims that the outage damaged 50,000 to 60,000 of wafers with V-NAND flash memory, which represent 11% of Samsung's monthly output. The report further estimates that the said amount equates to approximately 3.5% of global NAND output, but does not elaborate whether it means wafer output or bit output. Samsung uses its fab near Pyeongtaek to produce 64-layer V-NAND chips used for various applications. The fab is among the largest flash production facilities in the world and therefore any disruption there has an effect on the global output of non-volatile memory. Meanwhile, since production lines have not been damaged and the fab is back online, the significance of such an effect is limited.
heh - are you telling me (Score:3)
Re: (Score:2)
The only thing I can think of is the power disruption led to some sort of contamination. However with Fukishima, the problem wasn't the age of the plant but the design was not meant to handle both an earthquake and a 50ft tsunami at the same time.
The design worked well as soon as the earthquake hit, all reactors were immediate put into shutdown mode. The earthquake also hit the power grid but the diesel generators kicked in. The problem was that it would take about 24 hours of active cooling to get the rea
Re: (Score:3)
I have. I was IN a fab that lost complete power for several hours at a time once, largely due to human error. It took us weeks to recover our factory to normal operation.
Re: I've never heard of a FAB losing production du (Score:4, Insightful)
Since I forgot to mention details, it was one of STMicroelectronics's fabs in Phoenix, AZ, January 2009. Total blackout due to human error, fab was without power for 4 hours.
Re: (Score:2)
140 milliseconds... not bad for 100 megawatts...
http://www.news.com.au/technology/environment/thats-a-record-south-australias-tesla-battery-responds-to-coalfired-plant-failure/news-story/d9e02c0dbf6774ffea948a1b919f3b7f [news.com.au]
Re: (Score:2)
I've never heard of a FAB losing production due to a power loss
I have never heard of an industry that has not at some point lost production due to a power loss. I'm going to assume you spend too much time living in a cage.
Re: (Score:2)
10 seconds of power loss is still fatal to wafers in process. Toshiba had a power loss of *70 milliseconds* at a fab in December 2010, that caused material loss and a measurable drop in flash output in the following two months.
Re: (Score:2)
70 ms isn't a heck of a lot. To put it in perspective, at 60Hz, 1 cycle is 16.7ms (at 50Hz, 20ms). That means the fab lost power for a mere 3.5-4.25 cycles of AC power.
Most UPSes that aren't online ones switch within 2 cycles (that gives it time for it to detect the failure and switch on the inverter, let it stabilize and switch over)
There's a reason why fabs are located in places where the country is highly advanced - the power requirements are extremely high.
Re: (Score:2)
'At a lot of these places, the term "power outage" actually doesn't make sense - the population have never experienced more than a brief flicker of light. '
And yet they still build fabs in the USA.
What a coincidence (Score:5, Interesting)
Re: (Score:1)
Time to short AMD!
Re: (Score:2)
Not everything is a conspiracy. You can put away the tinfoil now.
Re:What a coincidence (Score:4, Insightful)
Then explain why ram has increased 200% and GPUs 500%? Cases and power supplies keep going up as $1500 in 2015 gets you a much better PC then ,2018 :-(
Greed!
Re: (Score:2)
Cryptomining. So, yes, greed.
Re: (Score:2)
Then explain why ram has increased 200% and GPUs 500%?
Basic supply and demand.
Cases and power supplies keep going
They are also doing wonders meeting new efficiency regulations. Give me a current PSU over an older one any day.
2015 gets you a much better PC then ,2018
It really doesn't. It'll only get you a better GPU and a smidgen more RAM. The PC itself is far better in 2018.
Re: (Score:2)
Then explain why ram has increased 200% and GPUs 500%?
Basic supply and demand.
Cases and power supplies keep going
They are also doing wonders meeting new efficiency regulations. Give me a current PSU over an older one any day.
2015 gets you a much better PC then ,2018
It really doesn't. It'll only get you a better GPU and a smidgen more RAM. The PC itself is far better in 2018.
I am trying to find the link. Either gamers nexus or Jays2cents on youtube did a comparison. Prices keep going way up on all parts except the CPU. Yes supply and demand but DDR 4 was claimed iphone 8 used same ram by 2017 the problem would be fixed etc ... gee no change.
FYI the SEC busted the ram makers before for price fixing. With Trump in charge they know they can do whatever they want as the Chinese are now investigating because they believe in regulation.
I smell a rat and if you remember in 2011 hard d
Re: (Score:2)
Prices weren't "getting low." Stop making shit up.
Re: (Score:2)
Hey now. If Twitler can make shit up, so can I.
Who's going to stop me, you and the Marine Core?
another mysterious fire with no video or pictures (Score:3)
I remember when a ram plant caught on fire a long time ago. Ram prices went through the roof. No video or pictures that 'event' either. A quick google search yielded very little on this disaster. Most of what I found was from a battery fire in 2017.
Re: (Score:3)
I've never seen pictures of you, so I'm reasonably sure that you're merely one of Alex Jones' AI chatbots and can be safely ignored.
BTW: Don't bother linking to a picture. I know about Photoshop, you sneaky bot you...
Re:another mysterious fire with no video or pictur (Score:5, Funny)
Here is a photo of the blackout [wikimedia.org].
Re: (Score:2, Funny)
That's a fake. If it were really a picture of the blackout, there would be stars visible in the sky.
Re: (Score:2)
NO I AM NOT BOT. BEEP BOP!
Re: (Score:2)
We did have out battery supplier lose their entire factory to a fire, forcing us to delay our own products until while we searched for and adapted to a different battery vendor. So the stuff does happen. Those guys weren't raising prices though, they had no product to sell.
Similarly, there was a shortgage of one processor we were using, and the explanation was that the fabs were all overbooked. As in all the big name suppliers of small embedded MCUs were in the same boat. It could have been lying, but s
Re: (Score:2)
No video or pictures that 'event' either.
I know right! Last time the power went out at my work costing many millions of dollars per day of outage my first reaction too was that I need to go and get a photography permit filled out, get the plant manager to sign it and take a photo! /facetous.
sabotage? (Score:2)
Hard Drive Redux (Score:4, Insightful)
Hackerzz? (Score:2)
More likely it was storms or squirrels or changes.
Re: (Score:2)
If there's even less video cards on the market, prices will rise even more.
Also, this wouldn't affect the current video cards already used in mining setups.
Backups? (Score:2)
Re: (Score:3)
I think it's safe to assume they had backup power and it didn't cut in. We had that exact problem at work last year - the UPS gets tested monthly along with the genny, but the transfer switch does NOT, and of course that didn't work when it needed to. So our genny started right up after some yutz hit a pole outside our building, purring happily while we were running the datacenter on a large UPS (30 minute runtime, normally only needed for about 40 seconds to start and stabilize the genny, switch to it fr
Re: (Score:2)
A critical system isn't redundant unless everything is redundant. Someone forgot the redundant transfer switches!
Re: (Score:3, Informative)
I work at Intel. Our fabs are comparable to Samsung's fabs in terms of equipment and scale. Ain't no UPS in the world that can keep a fab up and running if a blackout hits, even with backup generators. Most fabs are extremely sensitive to power glitches that last mere seconds, nevermind for a half hour.
Re: (Score:2)
Seems like you'd be talking to Elon. That Australian battery farm has been accomplishing switches within cycles.
Do you have any clue what about losing the power would damage 50K+ wafers?
Re:What Is a "Fab"? (Score:5, Informative)
If you didn't understand any of the abbreviates or acronyms then this isn't the site for you.
Re:What Is a "Fab"? (Score:4, Informative)
Fab is a common term of art for a facility that produces chips.
Re: (Score:2)
Here's a photo of a fab [spudsmart.com].
Soon to be sold on Amazon (Score:2)
How damaged? (Score:5, Interesting)
I’m curious how loss of power would destroy already-fabbed wafers. Do they need to be kept in conditions that require power to maintain or something?
Re:How damaged? (Score:5, Informative)
You should realize that these wafers build up 64 layers of flash storage and so include hundreds of precisely deposited layers of deposited and etched materials. Now imaging losing power in the middle of depositing or etching a layer so that the respective materials either fail to be deposited in the required amounts or dwell on the partially completed layer for too long.
Re: (Score:2)
Now imaging losing power in the middle of depositing or etching a layer so that the respective materials either fail to be deposited in the required amounts or dwell on the partially completed layer for too long.
Sounds plausible, only thing is, the articles talk of 'processed wafers'. Maybe they cooled off too fast?
Re: (Score:2)
But Anandtech is merely rereporting from TechNews, which doesn't mention anything about "processed" wafers. None of the other stories that I've seen concerning the outage mention "processed" wafers. So no, I'm not crediting one word that apparently was inserted by an author who's only acknowledged source is another author's work that doesn't include that fact.
Re: (Score:2)
I'm not sure processed means finished. As parent post mentions, 64 layers of flash storage could mean hundreds of layers of material. Each layer requires multiple steps [wikipedia.org], and most steps require both time and temperature.
Imagine trying to bake a layer of cake with a required precision of 1/100th inch of bread rise. If you kill the power before it's done, the cake is too flat or too puffy. They actually call them recipe's in the fab, and it's what all the work of manufacturing is - adjusting recipe's for m
Re: (Score:1)
It's a pipeline of processing steps. If the power fails, every wafer that is in a production step is immediately and irreversibly damaged. Restarting the production pipeline will probably also result in significant start-up losses until all machines are recalibrated after cleanup and repairs.
Re: (Score:1)
As someone who works in the industry, those wafers were probably all in production at the time of the power outage. With this kind of fabrication, losing power mid-process will affect the final product enough that the wafers cannot be salvaged (I've had this happen during R&D runs enough times to know that it is really frustrating!). For a fab with that kind of through-put, everything is automated for a reason and any effort to try and re-evaluate every wafer in an attempt to salvage some of them woul
Re: (Score:2)
Doubtful that wafers that are sitting idle will be damaged. However, depending on how many wafers are in-process concurrently, (and without going into specifics) a power glitch can EASILY cause damage to thousands of wafers. Power glitches are the easiest way to cause a clean room to suddenly NOT be a clean room.
Re: (Score:2)
Re: (Score:2)
I would assume they were only partially fabbed, and interrupting their manufacturing destroyed them.
What happened to the back up power? (Score:1)
This is why we need baseload power (Score:2)
In countries that the Greens have ruined, or prevented from industrializing in the first place, they keep claiming that all needed power can come from the sun and the wind. The reasoning is that if your urban economy is "post-industrial" a shop that consists of software developers can just knock off and go to the pub when the wind stops blowing at a time when there happens to not be enough sunshine.
But post-industrial countries can exist only when they can import the products of heavy industry from places l
Re: (Score:2)
I don't believe anyone suggests that there be no baseload power. The only question is can renewables take care of it (geothermal/hydroelectric) or can we store enough on peak production in batteries/flywheels to get us through the downtime.
Re: (Score:2)
Only a few countries (Norway, Iceland, Switzerland) are fortunate enough to have a renewable hydro or geothermal baseload. Because the rest of us have been using mostly coal and gas as the baseload, a serious effort to move away from carbon will require that we transition it from fossil to nuclear. To which the Greens reply, and there are several of them who will make that argument right here, that we no longer need baseload at all because we're "postindustrial."
Re: (Score:2)
I consider myself "Green" and pro-nuclear. Just a reminder that we exist. (granted, I do recognize it has a number of problems, the biggest involving waste & politics)
Re: (Score:2)
To which the Greens reply, and there are several of them who will make that argument right here
That doesn't seem terribly likely. If they don't materialise, will you withdraw your comment?
Re: (Score:2)
My point is they have already materialized, as you can see in any of those breathless threads about how fast the cost of installed wind is falling. This thread is in response to Amory Lovins' claim that wind makes the idea of a baseload obsolete:
https://hardware.slashdot.org/... [slashdot.org]
Re: (Score:2)
My point is they have already materialized
You said they'd reply right here. they haven't.
Re: (Score:2)
Right here in this forum, not right here in this thread.
Re: (Score:2)
LOL OK. The problem is then your argument degenerates to "someone said something stupid on slashdot", which falls into the same kind of general category as "water is wet".
I am appaled... (Score:3)
Re: (Score:2)
I'm actually more impressed that the power went out. What happened with all those recalled batteries? That would have made quite an incredible UPS.
Elon would like a word (Score:1)
Re: Materials Managers: (Score:3)
In the industry you cannot afford to stockpile millions of dollars worth of equipment just in case. It's cheaper long run to absorb the price increases (it's not like there will be nothing available) than to sit on aging stock.