EVGA power supply kills 22TB of customer storage then denies responsibility
tomshardware.com
tomshardware.com
In these cases 80% of retail box and the power supply itself should be covered with red warning label. "PINOUT CHANGED! WILL CAUSE PERMANENT SYSTEM DAMAGE"
EVGA are quoting warranty conditions, but surely this just comes under common tort liability, at least for the material cost of the damaged drives?
The redesigned power supplies are probably otherwise fine, but at the very least EVGA need to ensure the pin out doesn't let previous gen cables be plugged in.
Otherwise they're just asking for liability issues like happened here. :(
Those cable sets are ostensibly compatible with almost every decent/good psu that evga made (the entire g2/t2/p2/g3/g5/g6/g7/gq family theoretically share a set) and are marketed as compatible, traded on eBay, etc. They’ve polluted the whole shebang, 20 years of cables need to be tested or replaced.
https://www.evga.com/products/product.aspx?pn=101-CK-0850-B9
The only guaranteed safe way out is to either recall all cable sets from all units in those series and then ship out new ones marked as safe, or ship a free tester board to anyone with any compatible cable string (forever into the future) to validate that a given string is actually safe.
Evga isn’t going to do either of those because they’re basically on the way out of business already. They’re down to a skeleton crew etc and certainly are in no financial shape to do a giant recall, even if the big boss wanted to (and I’m guessing he’s the problem).
(The really fortunate thing is that it’s just the SATA string so far… but the other PERIF strings do plug into the same ports and are supposed to be compatible too! (I am guessing they fit and would blow up your drives too.) But modern users don’t use the SATA strings that often, because modern hdds use the 3.3v as a “power off” toggle to spin down the drives… so ideally you use molex adapters anyway. But still - I have at least 5x G2 units in my various machines, and I’ve mixed and matched cable sets without thinking - they’re sold as compatible, why wouldn’t I? I’ve used the SATA strings recently on some ssds etc. Not happy about this at all.)
We were fixing old QNAP server, and it used normal PSU keying. The thing is - it blew PSU capacitors(with a nice small explosion) after being turned on.
we suspected that the old QNAP model had proprietary pin layout for the PSU. No labels of course.
No clearer case of criminal incompetence. What a bunch of fuckwits. Noone at that company should ever work in engineering again.
(but still not the worst fuckwittery I've seen)
Wow, the drive manufacturer's warranty most certainly wouldn't apply. There was no fault with the drives at all.
Very poor look for EVGA.
Dollars to doughnuts that it doesn't (https://www.evga.com/warranty/power-supplies/), and likely has language confirming the opposite for liability reasons (it does). Commonplace for consumer gear.
Sure the good, and human thing to do would be to cut the guy a check, that's inarguably true. It's probably even a good business decision to, given that bad press is likely more expensive than a couple grand worth of drives. Legally though, they're probably in the clear.
I would bet dollars to donuts that they lost a lot more than the replacement cost would have been.
Here's a question: If you're going to change the pinout of the connector, which is proprietary anyway, why is the connector not keyed for the pinout it's supposed to use so you can't plug in the wrong cable?
Because it's cheaper to reuse the connectors they already have rather than to make a new one. /s
EDIT: Added a /s just to be on the safe side.
And it is not like they could not be used in some other model later...
It's insane that PSU manufacturers still haven't standardised the PSU-end of modular PSUs to this day. Or at the very least, standardised within one manufacturers offerings.
But I suppose I know the reason, it's more profitable if people can't reuse cables but have to get new ones every time they get a new PSU...
https://www.citizensadvice.org.uk/consumer/somethings-gone-w...
I'm sure that will totally fly in a court. Not one with real lawyers or jurists, but _a_ court.
This wasn't a mistake or freak accident. This was a reasonably foreseeable outcome of EVGA's decisions. They could have prevented it at multiple steps along the way. The only things the customer could have done to avoid this (i.e., replace the cables) they were directly advised not to do by EVGA.
Who could have prevent this happening, and how?
The end user could have prevented it by... not making use of the warranty he was entitled to? Ignoring EVGA's advice to keep his cables and instead replace them at his own expense without even knowing that the cables were in fact different?
EVGA could have prevented it by not sending him a PSU with a different pinout, by sending him the appropriate cables for the PSU he was sent, by providing a warning that the PSU was incompatible with previous cables and telling him to replace them, by...
You think there's no obligation here?
They could have easily rectified this by shipping new cables with the PSU and having sufficient warnings in multiple places.
And I guess the lesson learned is you RMA all parts when you RMA a product.
But EVGA said that they won't return the cables if you send the power supply with the cables. From the article:
After contacting EVGA, he sent the unit (at his own expense) to EVGA to get the problem resolved. But in typical fashion, EVGA told the Reddit post to keep all his accessories and power supply cables as they wouldn't be returned if shipped. This is not an unusual practice for most power supply manufacturers.
I haven't had to RMA a power supply yet, but I buy Seasonics that are gold or better.
Not sure I can hear those coil whines though.
also, I wish the EU / GOV would take action to ask HDD companies to provide affordable, repairable services for these PCB problems.
22 TB of storage - ~500 $
the data stored on the fried disks - may be priceless
Yeah, the old adage to always make backups applies here, but I guess many people (me included) back up stuff by keeping duplicate copies on different drives, because it's very rare for two drives to fail at the same time. Except if your PSU decides to fry them, of course...
Side note, but damn, imagine telling someone in the 90s that storage would get this cheap. We've come a long way.
83-93, disk prices dropped about 300x
93-03, about 1000x
03-13, about 20x
13-23, about 3x
SSDs are catching up but they're not going much faster.
Unless you are upstream-constrained or a data hoarder, just make offsite incremental backups. Preferably some redundant block storage.
S3 Glacier Deep Archive is 99c/TB. Just ship it.
It only costs $0.10 and $4/month to back up 1TiB, but $100-$350 to restore it, depending on how fast you want to restore. Still, $4/mo, or $88/month for 22TB is some peace of mind.
As a teenager I would have been perfectly capable and willing to pay 5/month to backup all my stuff. But I wouldn't have been able to afford 100-150 to restore it. So not knowing about the restoration fees would have given me a false sense of security and it might have been disastrous.
Regardless, it has good reliability properties.
But WD always were quite demanding for the power quality and the amount, ie some drives could work with 11.75 while WD probably wouldn't.
Chinesium PSU went way lower than 11v and... would you pay 5% more for the every HDD not to have a problem if your sub $20 PSU is sub $20 one?
You quite literally wouldn't jave a problem with any HDD if you invested $20 more in your PSU.
The sata power spec gives 3.3, 5, and 12 volt rails, but I believe most devices now don't use 3.3 or 5 volt rails at all because they want to use their own internal voltage regulators where power up scheduling is more predictable (ie. Ramp rates etc).
Given that, the drives probably only used 12v and gnd, any miswiring could only have been reverse polarity??
Never had an issue with seasonics and superflowers.
Sort of sarcasm, sort of not.
However, EVGA swapped voltages around on sata cables without changing the keying on the proprietary connector. That is fire-the-engineers grade stupid and should probably mark their exit from the power supply market. It drastically lowers confidence that the rest of the product is sensibly put together.
Say you buy USB-C brick. How do you test that is outputs correct 9V/12V when PD or PPS requests it? Do you inspect it with thermal camera under max rated load?
World cannot function if every fault is blamed to lack of due diligence by end user
There is no reasonable test that would have saved him.
The new supply even had the same model number!
https://reddit.com/r/DataHoarder/comments/1bilja1/egva_power...
But in any case, I was never blaming the guy. I'm sure I would have blown up my SATA drives in exactly the same way. I'm only saying that from now on, the paranoid, like me will always check the voltages on a new power supply output pins before I use them. He did actually check them after he realized the problem. It doesn't take long to check. You don't need to send it out to a lab as was suggested.