The Tech Report Endurance Test Ends Just After 2.4PB

Subject: Storage | March 16, 2015 - 07:00 AM |
Tagged: ssd, samsung 840 pro, Samsung, endurance

The Samsung 840 Pro was the last SSD standing in The Tech Report's experiment with a final score of over 2.4 petabytes written. Granted, only one (or two in the case of the Kingston HyperX) of each model participated, which means that one unit could have been top of its batch and another could have been bottom -- and can simply never know. What it does say, however, is that you really should not be worried about writing your SSD to death under normal (or even modestly abnormal) conditions.

View Full Size

This almost looks like one of our Frame Rating charts.

Again, that whole warning (above) about “this could be 100% binning luck” still holds true. Even so, here is the final ranking of contestants!

  1. Samsung 840 Pro (256GB)
  2. Kingston HyperX 3K (240GB with Compression)
  3. Corsair Neutron GTX (240GB)
  4. Samsung 840 (No Suffix and 250GB)
  5. Intel 335 (240GB)
  6. Kingston HyperX 3K (240GB)

The Tech Report notes that the Samsung drives did not warn users through SMART as much as their competitors. In both cases, death from write wearing was abrupt, albeit far into the future. I'd wonder what is next for them, but part of me expects that they never want to run anything like this again.

Source: Tech Report

March 16, 2015 | 09:45 AM - Posted by collie

This has been over a year and a half hasn't it? So sick that it lasted so long, that they ALL lasted so long. I shall nick-name the winning drive "Sammy-Eight-Fow-ty-P" and drink several drinks in Sammy's name!!!

YOU MAY TAKE OUR LIVES, BUT YOU WILL NEVER TAKE, OUR STORAGE!!!

et cetera

March 16, 2015 | 08:47 AM - Posted by Anonymous (not verified)

Its surprising to see that Kingston and Corsair drive last a lot longer than more expensive Intel drive. Intel really needs to get back to drawing board with regards to its consumer ssds.

March 16, 2015 | 09:03 AM - Posted by Anonymous (not verified)

Well the intel drive did run to its rated max and then went read only as opposed to the kingston and corsair drives which did just outright die.

March 16, 2015 | 09:07 AM - Posted by Anonymous (not verified)

That makes it sound like Intel may have limited the number of write in firmware. Probably to protect data from being lost as the drive passes its MTBF.

March 16, 2015 | 09:30 AM - Posted by Rustknuckle (not verified)

The insanely stupid thing about the Intel SSD is that once it goes read only and you reboot your PC it bricks it self permanently. Would have been better if the drive just stayed in read only mode.

March 16, 2015 | 10:28 AM - Posted by patrickjp93 (not verified)

Just like Apple and Samsung, equipment makers want to sell hardware. At least Intel outright protects your data instead of letting it potentially be lost forever.

March 16, 2015 | 11:59 AM - Posted by Rustknuckle (not verified)

I dont think you understood. One reboot and your data is also lost forever on the Intel SSD. What do most people do when something is not working right with a PC? They try a reboot and BAM their data is gone.

March 16, 2015 | 04:58 PM - Posted by Anonymous (not verified)

Well backups help solve that problem, and keeping track of the odometer should help the closer to MBTF the drive gets. Hopefully M$ and Intel have some driver features that can be set up for some write through/Mirroring between the SSD, and spinning rust that kicks in when the over provisioning space begins to dwindle, or the error correcting rates begin to approach a threshold. There is some good tiered storage software out there that is able to manage the storage pool between SSDs and hard drives, with backups to a mirrored file/backup share on a hard drive/s with plenty of options for keeping the SSD free of any stale files that may not be needed as quickly. Does Intel's driver software, not warn the OS long before that the end is near, and ask the user to make a backup, or warn the user that they are using the SSD on borrowed time?

March 17, 2015 | 05:49 PM - Posted by Master Chen (not verified)

Sooooo...just don't reboot your PC? %)

March 16, 2015 | 08:48 AM - Posted by BBMan (not verified)

Well, time to start up another batch with Samsung 850 etc. If Scott lives so long.

March 16, 2015 | 10:33 AM - Posted by patrickjp93 (not verified)

What's weird is there's no justification for this taking so long.
550MB/s * 24hr/day * 60min/hr * 60s/min = 47,520,000 MBs/day =47.52TBs/day. In other words, why the hell did it take so long?

2.5 PB can be written in 53 days. Why did it take a year? All of these drives have write speeds greater than or equal to 500MB/s

March 16, 2015 | 12:11 PM - Posted by Rustknuckle (not verified)

No drive writes at the top speed rated by manufacturers constantly once you throw a large continuous data stream at it. Then you add in data retention tests(leaving the drives unplugged for a week or more to see if they still keep the data), benchmarking and other data collection.

There is also a good chance the tests have not run 24/7 for 18 months as that would be irrelevant to the results and an unnecessary strain on the person doing the test.

March 17, 2015 | 09:35 AM - Posted by BBMan (not verified)

Look at the data. The actual speed was about 1/5 550. Moreover, to do this right, you have to take time out to verify that it isn't corrupted. So you have to read it at some point. Then you have interrupts between write streams, etc. 18 months is close to 550 days. When you consider everything, this isn't such a stretch.

Moreover, what someone pointed out was that the Samsung was "slower" than most. So it's possible that degradation is somewhat proportional to transaction speed- not necessarily design. Something to consider.

March 16, 2015 | 02:55 PM - Posted by Branthog

So one Samsung was the winner and the other was barely above the worst... and both Samsungs had MUCH slower average write-speeds than EVERY other drive.

March 16, 2015 | 06:05 PM - Posted by collie

no, the main take away was that SSDs have a lot more life to them than we thought they did 2 years ago. Before they started this test we were all rationing write cycles flash death fear, this test showed how silly that is, we can use our SSDs without fear. Who won and who lost is semantics, like the greyhound at the track, not necessarily the best just the winner of the day for those who were betting.

March 16, 2015 | 06:07 PM - Posted by collie

Actualy, forget that "No" you are abosulitly right, I just felt that you were glossing over the real point. Speed wasn't the test, it was to see how long it would take to kill an average ssd

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.