I try not to be too sanctimonious about clickbait stories because I'm guilty of that sin more than a few times. But one story in the International Business Times is really rubbing me the wrong way.
It opens with the hysterical headline of SSDs lose data if left without power for just 7 days, but if you read down you see a lot of very extreme qualifiers. It cites a presentation from a Seagate executive who claims that data will be retained on an SSD is halved for every 5 degrees Celsius rise in temperature in the area where the SSD is stored.
The writer said consumer-class SSDs can store data for up to two years before the standard drops, but enterprise SSDs will only retain data for a period of three months.
However, the devil is in the JEDEC slideshow details. The scenario described is an absolute, worst case scenario where if you write to the SSD at low temperatures, then store it at high temperatures (which is completely backwards from the usual scenario) and if the SSD is very close to its end of life (which is harder than you think) then leave it unpowered, which never happens in a datacenter because the servers run 24/7, then you might lose data.
As Jim Handy, an analyst who specializes in memory and SSD technology put it, "That’s a lot of 'ifs'."
He added "I haven’t the slightest idea why JEDEC even specified 3-month data retention for enterprise SSDs. Who the heck is going to leave an enterprise SSD powered down for even a few days? It’s either working 24/7 or it’s being replaced by a newer model, never to be used again. Plus, enterprise SSDs are mainly used to cache hot data, so whatever is on them is also copied onto the HDD they are caching."
The JEDEC spec referred to in the article is an absolute worst case scenario of data written to the SSD at a very high temperature (55C) and then stored at a lower temperature (40C) on an old SSD with lots of wear and left unpowered for three months. The thing is, JEDEC wasn't issuing a warning, it was laying out a test scenario for its members to see how things will play out in all kinds of temperature scenarios.
Handy said that for client SSDs the write load is so small that wear doesn’t measurably impact retention. "I believe that the standard guarantee is five continuous years without power, and I would be good with that," he said.
I decided to look into it myself. My main PC uses an OCZ Vertex 4 SSD and has been running for two years now. Besides work, there is constant activity from the Rainmeter skins and a distributed computing client. I used to run Einstein@Home but switched to World Community Grid. In the evening, I unwind playing EverQuest, which usually causes more heartburn than it eases.
So this SSD is never idle. It has constant read and write activity. And after two years, SSD Life reports that after 17,704 hours, the drive health is excellent, I have 99% lifespan left and it should last until 2024.
Granted, that speaks to the durability of an SSD, not its ability to last unplugged. But it just so happened I had an old PC sitting in the garage for the past eight months. It had an OCZ Agility 3 SSD for the C: drive and it had been through the insane weather here in Orange County, from 94 degree days to 45 degree nights in an uninsulated garage. Shame on me for not putting it in the box at least.
So I hooked it up, turned it on, and it fired up like a champ. There wasn't a single issue. SSD Life put its lifespan at 100% and said it was good to 2025. I even gave the drive a workout by installing Windows 10 on it. The install went without a hitch.
SSD makers have busted their backs in the past decade to improve durability and wear leveling to gain the confidence of consumers and enterprises alike. SSD life is now put at a decade or more, and is rapidly surpassing that of hard disks. One test found SSDs can last a theoretical one thousand years.
So do yourself a favor and don't be startled by sensationalist headlines.