Welcome to the MacNN Forums.

NewsPoster · 03:20 PM

Although many are celebrating the long-awaited arrival of Trim support for third-party SSD storage that can be enabled by users, Apple's implementation of it should be considered preliminary, and comes with an unusual and strongly-worded warning that offers no support if the user should experience problems. In addition, some popular SSD models, notably Samsung's 840 and 850 lines, may be subject to a data-destroying bug if Trim is activated.

Trim -- often typset in all caps, but not actually an acronym for anything -- is a special ATA command that manages how files are deleted on an SSD to maximize the performance and efficiency of the device by optimizing the "garbage data handling," since SSDs erase cells before writing new files. SSDs used without Trim or some alternative form of garbage collection gradually slow down over time.

Apple has provided Trim management for OEM SSDs that come with machines such as the Retina MacBook line and the SSDs that are part of Fusion drives, but for the past four years has not supported Trim on aftermarket SSDs. Some drive makers, such as vendor OWC, offer drives that include a built-in alternative garbage-management scheme known as "over-provisioning" and don't require Trim support.

As of the betas and now public release of OS X 10.10.4, users can opt to enable Trim support on third-party SSD drives that may need it for long-term performance. While third-party tools have previously been made available that could turn on Trim usage in third-party drives, by default the system didn't and still doesn't enable it. Users who want to turn on Trim management must open Terminal, issue the command "sudo trimforce enable," and supply their admin-level password to activate Trim support in SSDs that can use it. The Mac will automatically reboot in order to implement the new support.

Before that happens, however, users in Terminal will be greeted with a warning that says the forcing Trim support "may result in unintended data loss or data corruption," and recommends a backup of all data. Apple specifically states in the warning that "use of the tool is at your sole risk, and that the entire risk as to satisfactory quality, performance, accuracy and effort is with you," specifically excluding the Trim usage from any existing warranties or coverage to the extent permitted by applicable law, suggesting that Apple has not tested its version of Trim with a wide variety of possible third-party drives, some of which (such as SSDs using Sandforce controllers) do not need it and advise against enabling it.

Older SSDs in particular can vary in the way they implement or support the Trim command, but Apple's force-enabling of Trim should work fine as a complement to existing garbage-collection routines on recent models of SSDs. MacNN has, however, seen reports that Samsung's line of "840" and "850" drives, including the Evo and Pro lines, along with some models of SSD by Crucial and Micron, can permanently write the wrong data due to a flaw in their implementation of Trim support, and thus it is not advised to force-enable Trim if a user has one of those drives until given the all-clear from the manufacturer.

The bug that affects the Samsung drives may only appear in Linux distributions, as Windows users do not seem to be affected, but this leaves Mac compatibility unclear. Samsung has said it is looking into the issue. Most other recent SSD models will work with Apple's Trim command in harmony with any existing management.

Should a user decide to reverse the force enabling of Trim in OS X 10.10.4, the command "sudo trimforce disable" will revert to the default. Externally-mounted SSDs connected by USB, eSATA, Thunderbolt or other connection methods are generally not effected by Trim or the Trim flaw. Because the use of Trim only produces a notable difference in performance after years of use, it can be difficult to detect if enabling Trim is worthwhile -- Samsung SSD owners, however, report that the bug can strike nearly immediately, or take years to happen.

cashxx · Jul 2, 2015, 04:10 PM

Using 500GB EVO 840 with trim enabled here on 10.11 and no problems yet.

Mike Wuerthele · Jul 2, 2015, 04:13 PM

Cash, we're running some tests with accelerated wear on the 840, and 850 EVO and Pro SSDs we've got here. We'll be back with some numbers in a few days.

bobolicious · Jul 2, 2015, 04:15 PM

...blackmagic may be of use to check drive speeds:
https://itunes.apple.com/us/app/blackmagic-disk-speed-test/id425264550?mt=12

Charles Martin · Jul 2, 2015, 06:44 PM

cashxx, please see this link: https://bugs.launchpad.net/ubuntu/+source/fstrim/+bug/1449005

cashxx · Jul 2, 2015, 08:05 PM

I'll be curious to see the results. I have a backup so not worried. I'm not sure what SSD firmware I am on though, I'll have to check. I was aware of Linux having issues.....I haven't noticed anything yet under OSX....knock on wood!

Mike Wuerthele · Jul 3, 2015, 12:41 PM

So far, we've moved 44TB on and off four 840 and 850 pro and evo SSDs. That's about a month of heavy use. So far, so good.

More testing to come, but this is just the start. As previously mentioned, this may take a while. We'll keep going.

ablatt · Jul 10, 2015, 03:42 PM

Mike, how's the testing going? Have quite a few Samsung drives so I'm very interested.

Mike Wuerthele · Jul 10, 2015, 03:45 PM

No failures yet, and we're up well over 250TB off and on each drive.

ablatt · Jul 10, 2015, 03:52 PM

Are you using the latest firmware on all test drives? Running them under Yosemite 10.10.4?

Mike Wuerthele · Jul 10, 2015, 04:08 PM

The first pass is on the first, less reliable firmware. The second (which starts tonight) will be on the newer.

The test has to be under 10.10.4, as it relies on the new TRIM support enabled in that version- its the whole point of the test!

ablatt · Jul 10, 2015, 04:22 PM

When you say first, less reliable firmware, I take it you are referring only to the 850 Pro?

Mike Wuerthele · Jul 10, 2015, 05:34 PM

The entire line has had slow data-recovery firmware patches since initial release. There's no great Mac tool to do the updates, other than a DOS USB stick, so many Mac users are still on original firmware.

Mike Wuerthele · Jul 10, 2015, 09:03 PM

304TB across all four drives. Firmware updated, let's go again!

ablatt · Jul 11, 2015, 10:23 AM

As far as I know, all new 850 Pro's have shipped with the latest firmware for some time and I don't believe an update has been released yet for the 850 Evo. The 840 Evo had the slowdown-over-time issue, and the 850 Pro firmware update bricked some drives and was pulled, even though new drives ship with it.

Mike Wuerthele · Jul 11, 2015, 01:01 PM

Yeah, I'm aware of all this, and had old stock with the original firmware to start the testing with - otherwise, the scientific method is shot, and doesn't reflect both original and new firmware users. The 850 evo firmware, with an update to the originally shipping firmware to the current firmware isn't widely available, but we do have it. It is a very small update, and honestly, I've got no idea what its got in it.

So, what we've got here is original shipping firmware, that the testing is complete on. We're now moving the same files, with the same scripts, to the drives with the last 840 series firmware, and the currently shipping 850 firmware.

I'm working on an update article for later this weekend. I didn't see any trim failures, but am seeing some cell failures consistent with usage, easily handled by drive firmware.

Spheric Harlot · Jul 11, 2015, 02:35 PM

Any thoughts on the Crucial m500 series? I have a 1TB m500...

Edit: Apparently, Crucial m500 drives have issues with queued TRIM, but are fine with the sequential TRIM implemented in OS X.
Right-ho. Enabled.

Mike Wuerthele · Jul 11, 2015, 02:56 PM

Glad we could help

jcwren · Aug 13, 2015, 11:08 PM

I just enabled trim using 'sudo trimforce enable' on a Samsung 850 EVO 500GB SSD, and after the system rebooted, it's not coming up. I don't know exactly which firmware is loaded in 850, but it was purchased just a few days ago.

Of course, I have backups, so this isn't a big deal, but it is an inconvenience.

What I'm not clear on is if this is a sticky setting in the drive, and after restoring the SSD, do I need to 'sudo trimforce disable', or is this purely an OS setting?

Mike Wuerthele · Aug 14, 2015, 08:02 AM

That's a new one. I know a firmware update does that, but we have yet to see a drive unavailability because of the Trim enabling. The setting is sticky in the OS.