SK Hynix admits to DRAM defects, smacks down rumour it botched big batches
South Korean chip maker SK Hynix has admitted some of its DRAM components included defects, though it says accounts of the issue are overblown.
[...] The impact of the defects may be less substantial than the damage done to SK Hynix's reputation, as South Korean newswire Yonhap reports that rumours have circulated to the effect that 240,000 wafers of DRAM are defective. SK Hynix is thought to have the capacity to produce around 1.8 million wafers a month, around 80 per cent of which are dedicated to DRAM. The whispering therefore describes a serious situation at a time the world's hunger for silicon can't be satiated.
SK Hynix's missive to The Register continued: "The scale of the potential losses mentioned in the rumour is absolutely not true and exaggerated."
The company has therefore requested a police investigation into whoever is spreading this gossip.
SK Hynix Admits to Some DRAM Production Flaws, Calls the Cops
SK Hynix admitted that, while it was dealing with some faulty DRAMs, the damages did not impact 240,000 wafers. The company claimed that these rumors are meant to injure its reputation and called the local police to file libel and misinformation charges against its unknown assailant, thus spurring a police investigation into the rumors.
(Score: 0) by Anonymous Coward on Friday June 11 2021, @04:53PM (5 children)
In the past when a RAM stick was bad it just wouldn't boot = no problem because you know it's a dud, send it back. So what's the big deal? Can a bad RAM stick silently corrupt your data/programs without crashing?
(Score: 1, Informative) by Anonymous Coward on Friday June 11 2021, @06:09PM
Yes, it can. Data is in the memory too.
(Score: 1, Informative) by Anonymous Coward on Saturday June 12 2021, @02:54AM (2 children)
Yup, it absolutely can, and it's happened to me when one of my two sticks of RAM went bad. I was running a bunch of Matlab simulations. Everything worked fine until my data sets got beyond a certain size, at which point the simulation would throw really bizarre errors - as in, Matlab wouldn't actually crash, but would throw error messages that should have been mathematically impossible, based on what I was asking it to do. After spending several futile hours trying to debug my code, I ram a memory test, and it found problems with my second DDR module. It turned out my simulations were getting messed up when they started to get large enough to need to put data in that part of RAM. After I replaced it, my simulations worked fine.
(Score: 0) by Anonymous Coward on Saturday June 12 2021, @09:14PM
Shit. Better do a memtest. I buy the cheapest Chinese non-EEC garbage I can find.
(Score: 0) by Anonymous Coward on Saturday June 12 2021, @09:48PM
Shit. Where is memtest?! Looks like it stopped working with UEFI. What the fuck are we supposed to do now, Linus?
(Score: 3, Interesting) by Reziac on Saturday June 12 2021, @03:03AM
Yes. My old 286 has a bad RAM chip. This manifested two ways:
-- crash when WordPerfect was asked to select a font.
-- random character substitution in my BBS message editor (plain text).
After I locked out the affected memory address, both problems stopped.
And there is no Alkibiades to come back and save us from ourselves.