Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Oh well, we will never know then, maybe BMD Speed-test was faulty, will never know now, I returned the USB enclosure and and now have a bare SSD -> now looking to buy Thunderbolt enclosure.

There's no question it was faulty, it was reporting a read speed that was literally impossible.

(And again, TRIM has nothing to do with read speeds.)
 
Oh well, we will never know then, maybe BMD Speed-test was faulty, will never know now, I returned the USB enclosure and and now have a bare SSD -> now looking to buy Thunderbolt enclosure.
If your SSD is a SATA blade M.2 type device, you should return that too. Otherwise you won't see any significant gain in performance even if you can find a Thunderbolt enclosure that uses those drives. You should be using an NVMe disk with a Thunderbolt enclosure for best performance.
SATA drives are limited to approximately a theoretical 600MB/second which is not significantly limited by a 10Gb USB 3.1 interface.
 
If your SSD is a SATA blade M.2 type device, you should return that too. Otherwise you won't see any significant gain in performance even if you can find a Thunderbolt enclosure that uses those drives. You should be using an NVMe disk with a Thunderbolt enclosure for best performance.
SATA drives are limited to approximately a theoretical 600MB/second which is not significantly limited by a 10Gb USB 3.1 interface.

You can get a Thunderbolt 2 or Thunderbolt 3 PCI expansion for about $200. They are bootable if you get a pic-e bootable NVME interface. The trick is finding one that is "Driver-less". It's actually not a bad choice but it is rather pricey in the end. If you are on TB2 Mac, you would need an Apple TB2->TB3 adapter. Make sure it's the Apple brand as it is bi-directional. This is what is needed to get that to work. The PCI-e cards can vary in price: I am not sure if the $20 ones are driverless but they start there and go as high as $100. Then there is the NVME card. It's pricey.

I have not seen TB2 or TB3 laptop drive enclosures in a while. I only know of one is still marketed right now and it comes with a drive. Sata only.

Maybe with USB 4.0, we will see more options. It will include the PCI-e bus in Thunderbolt. Basically bringing Thunderbolt to USB. (Thunderbolt 3 brought USB to Thunderbolt).

As for performance between TB2/TB3, you won't really notice a difference. The 40 GB/s is by decoupling the synchronous transfer of data. It is hard to explain but basically, Thunderbolt consists of two transfer channels. One is PCI and one is Video. Not both. TB1 took the two 10 GB channels which ran in parallel, and then split them two 1 way in TB2. SO basically it's still 10GB/s for PCI and 10 GB/S for video. TB3 theoretically bumped that up to 20 GB/s for PCI and 20 GB/s for video but there is a caveat - a huge one - it requires active cables. Otherwise you're not really going to see that. If you see a speed boost it's because you have a 6" cable and your device doesn't have much overhead.

So is it worth it? Is USB 3.0/3.1 good enough with garbage collection? Well... See my next post.
 
Chances are, TRIM isn't supported.

BUT... don't worry about it.
TRIM will make next-to-no difference at all. It's one of those issues that is "lots of smoke but no fire".

I've been booting and running a 2012 Mac Mini for almost 6 years using an SSD in a USB3/SATA dock. The lack of TRIM has never been an issue ... NEVER.
Having said that, I keep plenty of free space on the SSD, so the OS has "room to breathe".

WRONG - TRIM makes a difference.

When that problem-free SSD he keeps basing this declaration of TRIM's irrelevance on does eventually die it may have been twice as fast as it would have in a TRIM-enabled machine.

No, TRIM does not work over USB in macOS and yes, TRIM absolutely does make a difference in performance and significantly reduces write amplification.

The lifetime of an SSD is calculated in write cycles.

SORTA WRONG - TRIM is not just about life cycles. It is relative to the type of NAND, Cache, and other "stuff"

No. TRIM DOES NOT matter that much any more.
It's Sushi's "advice" that is misinformation.
I stand by this post.

An open request to anyone reading this that is currently booting and running an iMac or Mini via USB:
Has the lack of TRIM mattered much for you?
If so, have you been able to document the information?

For someone with an older Mac (2012-2015), who has only a platter-based hard drive inside, and who wants to "squeeze more years of use" from it, adding an external USB3 SSD is a cost-effective solution. No breaking open the machine, no risk of breaking something inside.

It will greatly speed up the computer, and so long as the user leaves a good amount of free space on the SSD, it will probably continue to perform well long enough to "outlive" the computer itself.

WRONG - I have an older Mac with SSD SATA Drives. TRIM did not help. Garbage collection did not help. It did not live long enough.

I also had a USB 3.0 drive as boot up for my iMac 5k. I bought it 2 years ago. Drive is failing.

I have a MacBook Air. It's 5 years old. Still works great but the drive has lost 12% of the original space. I always kept it at 50%. Trim was running. Drive is weighing in at 350MBs write 420MBs read.


It's not black or white. It's complex.

Here was my reply to your demand almost a year ago. If you like, I can find other posts with more examples.

I understand you have not had any issues running without TRIM, but some users do.

GREAT ANSWER

Relying on garbage collection instead of TRIM is viable to some degree but not ideal. I bought a Mac Mini in 2012. I didn't want to go through the trouble of switching the internal drive to an SSD so I bought a 2.5" SSD and put it in an external enclosure and used that as my boot drive. It worked well for about 2 years but then I started experiencing horrible slowdowns if I wrote more than ~1GB to the drive over the course of a few minutes. Clearly I was exceeding the ability of the drive to free up empty blocks. It didn't happen often enough but the slowdowns were bad enough that I had to do something about it. (The system would hang for several seconds if it tried to write ANYTHING to the drive past ~1GB and it would stay like that for 15-30 minutes.) I bit the bullet and installed the drive internally, enabled TRIM, and that eliminated the slowdowns.

ALSO WRONG


SIGH.... ok here's the deal folks. No one is talking about all the reasons SSDs fail:
1) Is it DRAM-less?
2) MLC? TLC? SLC?
3) How much are you using it? Are you writing video?
4) What are you using it for? Database r/w?
5) Heat (NVME runs hot. This risks the life of the drive if it's tucked under a pile of papers)

If you are doing video, get a RAID setup. seriously. You can get a dual bay laptop 2.5" RAID for a decent price. You can also get a 4-bay Desktop RAID for about $100. Why? SSD is not meant for frequent reading and writing!!! If you opted for a TLC based consumer SSD, make sure you got one with an 2, 4, 6, 8 at the end. In other words, 128, 256, 512, 1024. That extra space is actually DRAM. Some manufacturers use MLC cache. For the most part avoid the 120, 240, 480, and 960. They are TLC and TRIM will be an absolute MUST otherwise your drive will begin to fail in a year.

Also remember that NAND is faster in larger sizes. "Wha????? Now you're making stuff up" No. How do you think they get from 120+8 to 240+16? What about 480+32 or 960+64? Do the cards get bigger? Do they just make this super long card with more chips? No. The actually do add more chips but they run them in parallel. In other words, they are multiple 120+8 or 240+16 drives in tandem. It sounds like RAID but it's parallel I/O because it's on the same PCB.

This is why a drive can appear to be so fast. In actuality, it's just a byproduct of technology.

Ok what is SLC, MLC, TLC, or QLC? They stack the cells on top of each other to increase the amount of data that can be stored per chip. If you have a 240 GB drive with 1 chip and another with 4 then which one is likely to fail faster? what about a 256 with 2 chips and another with 6? The more chips there are, the less likely they are to be written on. The 6 chip is also odd because it is just 2 more chips but those are cache.

QLC solves the problem of adding fewer chips to the board. It creates a larger problem of adding more write cycles per chip. I don't care how much TRIM or garbage collection you have. Your drive is going to fail. Sure, TRIM will help and it will help the drive be fast but how fast is 240 or 256 compared to 960 or 1024? Remember - a 960 is just 240x4 run in parallel. A 1024 is 240+8 * 4 run in parallel.

All of this means you may think you're getting a performance boost with a large USB drive but it's only temporary if you got the 960. If you got a nice NVME then you have some cache so that will help. However, garbage collection will only help with that extra cache. TRIM helps even more and is more important without the cache.

So bottom line - UNMAP is great for USB 3.1. Apple DOES support UNMAP. It's in the code. If you know your way around the command line and can grep the drive with ioreg then you'll see. It's there. I am not going to give you the exact commands. I spent hours trying to figure it out - on my own. You can too.

The challenge is whether the UASP device is going to accept the commands and if the drive will accept those commands. They were written for SCSI. In Windows and Linux, you can get it to work but you also have to run the drive as a SCSI (or SAS) drive. The thing about Windows - it has more support for these types of DIY devices. Apple moved further and further towards an end-to-end solution. They claim to be moving back to the "modular" realm but They said that in Fall of 2017 - before the tariffs and before pressure to assemble more products over here. Once upon a time, SCSI was synonymous with Apple. Funny how times change. Apple has left third party hardware to fade into the wind.

I think I covered my bases here. Everything is possible and it's not a simple answer. Basically, spend the money and get a Thunderbolt PCI enclosure (WITH FAN) and use a driverless PCI express card with NVME SSD or wait for USB 4.0 which will include parts of the Thunderbolt specification. (The most important part - the PCI express part) It's better than buying a new SSD every 1-2 years because they keep failing.

Good luck.
 
Last edited:
WRONG - TRIM makes a difference.



SORTA WRONG - TRIM is not just about life cycles. It is relative to the type of NAND, Cache, and other "stuff"



WRONG - I have an older Mac with SSD SATA Drives. TRIM did not help. Garbage collection did not help. It did not live long enough.

I also had a USB 3.0 drive as boot up for my iMac 5k. I bought it 2 years ago. Drive is failing.

I have a MacBook Air. It's 5 years old. Still works great but the drive has lost 12% of the original space. I always kept it at 50%. Trim was running. Drive is weighing in at 350MBs write 420MBs read.


It's not black or white. It's complex.



GREAT ANSWER



ALSO WRONG


SIGH.... ok here's the deal folks. No one is talking about all the reasons SSDs fail:
1) Is it DRAM-less?
2) MLC? TLC? SLC?
3) How much are you using it? Are you writing video?
4) What are you using it for? Database r/w?
5) Heat (NVME runs hot. This risks the life of the drive if it's tucked under a pile of papers)

If you are doing video, get a RAID setup. seriously. You can get a dual bay laptop 2.5" RAID for a decent price. You can also get a 4-bay Desktop RAID for about $100. Why? SSD is not meant for frequent reading and writing!!! If you opted for a TLC based consumer SSD, make sure you got one with an 2, 4, 6, 8 at the end. In other words, 128, 256, 512, 1024. That extra space is actually DRAM. Some manufacturers use MLC cache. For the most part avoid the 120, 240, 480, and 960. They are TLC and TRIM will be an absolute MUST otherwise your drive will begin to fail in a year.

Also remember that NAND is faster in larger sizes. "Wha????? Now you're making stuff up" No. How do you think they get from 120+8 to 240+16? What about 480+32 or 960+64? Do the cards get bigger? Do they just make this super long card with more chips? No. The actually do add more chips but they run them in parallel. In other words, they are multiple 120+8 or 240+16 drives in tandem. It sounds like RAID but it's parallel I/O because it's on the same PCB.

This is why a drive can appear to be so fast. In actuality, it's just a byproduct of technology.

Ok what is SLC, MLC, TLC, or QLC? They stack the cells on top of each other to increase the amount of data that can be stored per chip. If you have a 240 GB drive with 1 chip and another with 4 then which one is likely to fail faster? what about a 256 with 2 chips and another with 6? The more chips there are, the less likely they are to be written on. The 6 chip is also odd because it is just 2 more chips but those are cache.

QLC solves the problem of adding fewer chips to the board. It creates a larger problem of adding more write cycles per chip. I don't care how much TRIM or garbage collection you have. Your drive is going to fail. Sure, TRIM will help and it will help the drive be fast but how fast is 240 or 256 compared to 960 or 1024? Remember - a 960 is just 240x4 run in parallel. A 1024 is 240+8 * 4 run in parallel.

All of this means you may think you're getting a performance boost with a large USB drive but it's only temporary if you got the 960. If you got a nice NVME then you have some cache so that will help. However, garbage collection will only help with that extra cache. TRIM helps even more and is more important without the cache.

So bottom line - UNMAP is great for USB 3.1. Apple DOES support UNMAP. It's in the code. If you know your way around the command line and can grep the drive with ioreg then you'll see. It's there. I am not going to give you the exact commands. I spent hours trying to figure it out - on my own. You can too.

The challenge is whether the UASP device is going to accept the commands and if the drive will accept those commands. They were written for SCSI. In Windows and Linux, you can get it to work but you also have to run the drive as a SCSI (or SAS) drive. The thing about Windows - it has more support for these types of DIY devices. Apple moved further and further towards an end-to-end solution. They claim to be moving back to the "modular" realm but They said that in Fall of 2017 - before the tariffs and before pressure to assemble more products over here. Once upon a time, SCSI was synonymous with Apple. Funny how times change. Apple has left third party hardware to fade into the wind.

I think I covered my bases here. Everything is possible and it's not a simple answer. Basically, spend the money and get a Thunderbolt PCI enclosure (WITH FAN) and use a driverless PCI express card with NVME SSD or wait for USB 4.0 which will include parts of the Thunderbolt specification. (The most important part - the PCI express part) It's better than buying a new SSD every 1-2 years because they keep failing.

Good luck.
So are you say I should not be using the 4 SSD's I have attached to my shiny new iMac? I am only using USB 3, so I should not be able to partake in the benefits of a SSD? 2 of them are 3 years old and still preform great. Everything is backed up on multiple drives and on the cloud. If I do loose a drive they are cheap to replace. I think for me it's a better option than spending a large amount of cash for a "Thunderbolt PCI enclosure (WITH FAN) and use a driverless PCI express card with NVME SSD" and also have to listen to a annoying fan.
 
So bottom line - UNMAP is great for USB 3.1. Apple DOES support UNMAP. It's in the code. If you know your way around the command line and can grep the drive with ioreg then you'll see. It's there. I am not going to give you the exact commands. I spent hours trying to figure it out - on my own. You can too.


Are you just saying that to be an absolute jerk? I've been searching for a long time to find those commands. Why should we all waste hours on something that's probably not even possible, just so you can feel superior?
 
  • Like
Reactions: Dru Nemeton
Are you just saying that to be an absolute jerk? I've been searching for a long time to find those commands. Why should we all waste hours on something that's probably not even possible, just so you can feel superior?

Everything he wrote is barely coherent and only vaguely correct. I suspect he won't share the commands he was talking about because he doesn't actually know any "commands."

There's every indication that MacOS supports UASP (SCSI protocol over a USB connection). There are ways to check to see if this is enabled for a particular drive. I guess sometimes it's not enabled for some drives, maybe depending on how they're formatted? You can do some googling about this.

SCSI has a command called UNMAP which is the same as TRIM (TRIM is a SATA command).

I can't figure out if MacOS sends the UNMAP command to drives or not. I guess there's no reason why it wouldn't but I can't tell for sure that it does.

So the last piece of this puzzle is that you need a USB drive enclosure that connects via UASP and translates the UNMAP command into a TRIM command for whatever SATA drive you put in the enclosure. I did a quick google and can't seem to find an enclosure that explicitly supports this translation but I know I've found some in the past.
 
The best I can figure is the JMicron 538 does not allow it but a different JMicron controller with a higher number does. I might just wait a few months and see if a TB3 NVMe adapter comes out that I can slip the drive into. I found a bunch of enclosures that claim to support SMART and TRIM but the fine print says "not on Macs" (or a comment from another customer does, for the poorly documented enclosures, e.g. Orico). I've found enclosures that support the Windows workaround but not for the Mac - so far. If I still had my Mac Pro (okay, I DO still have it, I just don't use it), I could slip the NVMe drive onto a card and that would work!

Thanks for your courteous reply, and I think you’re right. On the other hand... I've corresponded with kind, generous people and with people who seem to think that just because they had to work to find something out, everyone else should, too.
 
I can't figure out if MacOS sends the UNMAP command to drives or not. I guess there's no reason why it wouldn't but I can't tell for sure that it does.
I searched on this a LOT in the past and never could find any evidence macOS supports UNMAP. I think any UASP enclosure would do the trick for it under Windows though.
 
Agreed. Kinda useless though since I don't know of any way to read APFS under Windows... or does that not matter?
 
The best I can figure is the JMicron 538 does not allow it but a different JMicron controller with a higher number does. I might just wait a few months and see if a TB3 NVMe adapter comes out that I can slip the drive into. I found a bunch of enclosures that claim to support SMART and TRIM but the fine print says "not on Macs" (or a comment from another customer does, for the poorly documented enclosures, e.g. Orico). I've found enclosures that support the Windows workaround but not for the Mac - so far. ...

I'm pretty sure these enclosures and controller chips are OS-agnostic, i.e., they just support UASP and SCSI and SATA and so forth, they don't know if they're connected to a Mac or a Windows computer. So there's no way they could support TRIM for Windows and not a Mac, and there's no workaround that would allow TRIM via Windows but not the Mac. I suspect you might be looking at the description text for cheap Chinese-import enclosures on Amazon that have nonsense description text that may or may not be true, or even make any sense at all. :/

Looks like the JMicron 578 supports this translation for sure:
https://www.anandtech.com/show/10779/inateck-fe2010-and-ua1001-sata-usb-30-storage-bridges-review
 
The 538 definitely does not. It was explained earlier — they support various standards. Windows supports TRIM via a workaround. Apple did not write any support in because “we don’t sell it, so who cares?”

The 578 is the one I had mentioned as •possibly• supporting TRIM and SMART on Macs. The link you gave did not mention Macs as far as I could tell.
 
The 538 definitely does not. It was explained earlier — they support various standards. Windows supports TRIM via a workaround. Apple did not write any support in because “we don’t sell it, so who cares?”

The 578 is the one I had mentioned as •possibly• supporting TRIM and SMART on Macs. The link you gave did not mention Macs as far as I could tell.

I see, I found the discussion. The "Windows workaround" is not anything specific to Windows and it's not really a workaround, it's just that Windows definitely sends SCSI UNMAP commands to UASP drives.

And there's a Windows command that will explicitly send these UNMAP commands to clear out free space, and there doesn't seem to be a similar MacOS command. But just because MacOS doesn't have such a command doesn't mean it doesn't send UNMAP commands to drives through the course of normal use. In other words, the fact that Windows has this explicit command isn't really relevant to anything.

I still haven't been able to find a definitive answer anywhere re: whether or not MacOS sends UNMAP commands to UASP drives.
 
there's a Windows command that will explicitly send these UNMAP commands to clear out free space, and there doesn't seem to be a similar MacOS command. But just because MacOS doesn't have such a command doesn't mean it doesn't send UNMAP commands to drives through the course of normal use. In other words, the fact that Windows has this explicit command isn't really relevant to anything.


You aren't quite there yet. The Windows command means it can essentially do TRIM. The lack of a Mac command means it can't. That is very important—it is the difference between having it and not having it. What we need is for Apple to add a tiny bit of programming to allow this. The Mac Mini and iMac can't take more internal storage, any more than the laptops, so they tell us to add USB drives... yet, they make it harder to do so. Unless of course we want to spend $$$ for TB3 rather than USB 3.1!

I think, but I could be wrong here, that Windows also supports SMART reporting over USB, while Apple does not. Again, lazy programmers? cheep managers? Either way, the Windows people have a better system for dealing with external drives—and that shouldn’t be!
 
Last edited:
The DriveDX devs mention and link to that github page on their site.

If you go to acknowledgments under help in Drive DX you can also see where they disclose they are using the open source smartmontools for DriveDX and just putting a pretty GUI on it.
 
Yup. I tried versions 0.8 and 0.10 and neither worked, so I guess that's the end of my trying. It does seem to work for some other people on older versions of Mojave. The DriveDX version of the extension is signed, which is a bit of value-added in itself, and I understand it has some other mods... presumably more drive descriptions.
 
You aren't quite there yet. The Windows command means it can essentially do TRIM. The lack of a Mac command means it can't. That is very important—it is the difference between having it and not having it. What we need is for Apple to add a tiny bit of programming to allow this. The Mac Mini and iMac can't take more internal storage, any more than the laptops, so they tell us to add USB drives... yet, they make it harder to do so. Unless of course we want to spend $$$ for TB3 rather than USB 3.1!

No, UNMAP and TRIM are protocol-level commands that are issued to drives through the normal course of drive usage.

What's being discussed is a Windows command-line tool that will manually issue UNMAP/TRIM commands to clean up free space that was not already UNMAPed/TRIMed for some reason. This command should literally do nothing. The only reason why it would do anything at all is (1) if you were using the drive before with an OS that didn't issue UNMAP/TRIM commands correctly, or (2) if there's a bug in the OS that causes it to not issue UNMAP/TRIM commands correctly.

That's nice that the command-line tool exists in Windows (I guess) but it has absolutely literally nothing to do with whether or not the OS supports UNMAP/TRIM.

I think, but I could be wrong here, that Windows also supports SMART reporting over USB, while Apple does not. Again, lazy programmers? cheep managers? Either way, the Windows people have a better system for dealing with external drives—and that shouldn’t be!

I think that's the third time you've mentioned SMART. What are you hoping to get with SMART information? I remember it being somewhat interesting in the day and age of spinning platter hard drives because it could supposedly give you some clues about when the drive might break. I haven't heard anybody talk about it since everybody switched to SSDs though.
 
I think that's the third time you've mentioned SMART. What are you hoping to get with SMART information? I remember it being somewhat interesting in the day and age of spinning platter hard drives because it could supposedly give you some clues about when the drive might break. I haven't heard anybody talk about it since everybody switched to SSDs though.


SMART provides info on errors and use of the backup chips, it's apparently been helpful in predicting digital drive failure as well as spinny failure.
 
SMART provides info on errors and use of the backup chips, it's apparently been helpful in predicting digital drive failure as well as spinny failure.

I see. I wouldn't worry about that. Actually there's no such thing as backup chips in an SSD. SSDs are "overprovisioned" meaning they have more space than they're rated for, in case some sectors stop working, they can be remapped to different sectors. But they aren't on specific backup chips, they're all the same chips.

But moreover, the usual failure mode for an SSD is that you can no longer write to it. But you will usually be able to read from it. So if an SSD fails, you just buy a new one and copy your stuff from the old one to the new one, no problem. Assuming you don't do regular backups, which you should.
 
I test SSDs all the time for PCWorld. TRIM simply forces the drive to do it's house-cleaning. Erasing and deallocating blocks, etc. Depending on the drive, it may or may not perform every task at the time of TRIM. As so many SSDs allocate NAND as secondary (or even primary) cache by treating it as SLC, freeing up NAND and consolidating data can have an effect on performance. You'll likely never notice it under normal circumstance, but we definitely notice an improvement during long writes after a TRIM on a drive that's already seen heavy use.

While the drive may and should perform this housekeeping on its own eventually, for testing we like to give it a pat on the rump.
 
I test SSDs all the time for PCWorld. TRIM simply forces the drive to do it's house-cleaning. Erasing and deallocating blocks, etc. ...

This is not correct.

TRIM is a command issued to the drive to tell it that a specific block/sector is no longer being used (because its data belongs to a file that has been deleted).

Without TRIM, the drive has no way of knowing if the data in a block is still being used by the file system or not. If a block was ever used to store data, the drive has to preserve its contents for all of eternity.

So, with TRIM, all the unused blocks in a drive are marked as unused. The drive is free to erase them, and can thus quickly write large amounts of new data to all the empty blocks.

Without TRIM, the drive eventually fills up with data (some of which has been deleted, but the drive doesn't know this), and the only place the drive can quickly write new data is to the relatively small number of over-provisioned blocks. Moreover, the drive has to waste some time and effort to do wear-leveling and refreshing of blocks that contain deleted data, which is pretty silly.
 
Last edited:
I misspoke, but in effect, TRIM does inform the drive that house-cleaning is needed and most will take care of it in short order. However, I sometimes find that a reboot and/or a long pause is required for the drive to regain its former speed.
 
I misspoke, but in effect, TRIM does inform the drive that house-cleaning is needed and most will take care of it in short order. However, I sometimes find that a reboot and/or a long pause is required for the drive to regain its former speed.

If TRIM does trigger some kind of housekeeping on the drive, that was a decision made by the manufacturer. That's not the purpose of the command. The only purpose of TRIM is to indicate that a block is no longer being used by the file system. TRIM commands are issued whenever a file is deleted. You make it sound like TRIM commands are issued manually but that should never be necessary.

After you delete a file, it may take a while for the drive to actually erase the blocks that were occupied by the file. Drives will wait until they have been idle for a while to do this sort of housekeeping. This is expected.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.