MacInTouch Amazon link...

SSD, Fusion and flash drives

Channels
Security, Products

Ric Ford

MacInTouch
... I'm still on the fence about this.
Since the hard drive slowdown issue relates to fragmentation, it gets worse over time, as Bombich's testing showed. A fresh install shouldn't be as bad as it will become later, and different people will see different results, depending on their systems and work.
 



I wonder if it would be possible to create a fusion drive from the Apple blade SSD and your new SATA SSD. The logic here is that the PCIe SSD is probably faster than your SATA SSD, so you may see performance improvement when they're fused.
My guess is probably not. I learned from first-hand experience that under the CoreStorage Fusion model, if CoreStorage cannot detect a difference between the media types, it would not properly tier the data activity.

Specifically, when I first created my Fusion drive, I had my SSD behind a USB bridge that masked the fact that it was an SSD (as verified by System Profiler), and so CoreStorage thought they were two hard disk drives, and I could verify through iostat that the disk activity was not preferentially going to the SSD.

After connecting the SSD to an internal controller (thus detected as "SSD" type in System Profiler) and recreating the Fusion Drive, iostat showed that it was properly tiering the data activity.

All of which is to say that Apple probably has not put in the logic to distinguish between NVMe and SATA/AHCI SSDs, since they have never shipped a machine like that, and so Fusion would not tier the data access accordingly between the two types.

Who knows, something may have changed since Fusion Drive moved from CoreStorage to APFS, but I doubt it, and, of course, there is no documentation.
 


I just had a user's older iMac (2013 I think) upgraded to an internal 2TB SSD (Samsung EVO SATA). I did learn it was a pain to move data from one Catalina drive to another Catalina drive, as there were multiple users, and Apple treats the drive as two partitions (Macintosh HD and Macintosh HD - Data). Kudos to Bombich's Carbon Copy Cloner for making this possible.
A much simpler method, and the one that I followed, is to install the new drive, boot from the original, install the OS on the new drive, boot from the new drive, then use Migration Assistant before disabling the original drive.
 


Since the hard drive slowdown issue relates to fragmentation, it gets worse over time
Writing a Carbon Copy Clone from an hard disk drive has the side-effect of defragmenting the data as it is written. Years ago (12" G4 PowerBook) I did that and obtained noticeable throughput improvement when the clone was written back to the internal hard disk drive. I think I recall having an application that displayed the amount of fragmentation on my disk and finding CCC worked as a cure.

Still works, if there's a boot hard disk drive. Shouldn't matter if SSDs are both source and target.
Bombich Knowledge Base CCC5 said:
I want to defragment my hard drive
Defragmentation is a natural result of backing up your data to an empty backup volume. Simply prepare your backup volume for use with Carbon Copy Cloner, then use CCC to clone your source volume to your destination volume.
 


Ric Ford

MacInTouch
Writing a Carbon Copy Clone from an hard disk drive has the side-effect of defragmenting the data as it is written. Years ago (12" G4 PowerBook) I did that and obtained noticeable throughput improvement when the clone was written back to the internal hard disk drive. ...
Yes, that's very true for hard drives - clone-erase-restore is a good way to recover from performance degradation that accumulates over time, but be careful to get drive selection right and not wipe out the wrong one!

(SSDs are unaffected, due to their fundamental technical differences.)
 


Regarding a method to affix an external drive to an iMac: Sticky-backed Velcro is your alien technology friend. Place the drive on the iMac's foot, well back by the bend with the cable facing the ports, of course.

By the way, I'll mention that some clients have called when their iMac has failed to boot properly from the external drive; consistently, it's the cable that has been partially pulled out (yes, user error). Be prepared for these calls... or (horrors) use some gorilla snot [Amazon] to keep the cable ends in place.
 


Ric Ford

MacInTouch
... Place the drive on the iMac's foot, well back by the bend with the cable facing the ports, of course.
For what it's worth, different drives have lights on different ends. The Samsung T5, for instance, has a nice indicator light on the same end as the USB-C port, while other drives have lights on the front but USB ports on the back.
... or (horrors) use some gorilla snot [Amazon] to keep the cable ends in place.
I wonder if Sugru would also work? (N.B. I haven't tried it....)
 


Yes, that's very true for hard drives - clone-erase-restore is a good way to recover from performance degradation that accumulates over time, but be careful to get drive selection right and not wipe out the wrong one!
One nasty side effect of the clone-erase-restore is Time Machine will think the restored drive is a new drive and attempt to back up the entire drive to your Time Machine volume -- and potentially result in files being deleted from Time Machine. Been there -- experienced that.
 


One nasty side effect of the clone-erase-restore is Time Machine will think the restored drive is a new drive and attempt to back up the entire drive to your Time Machine volume -- and potentially result in files being deleted from Time Machine. Been there -- experienced that.
There are shenanigans you can pull to avoid this. They require some attention to detail and care. For me, the biggest gotcha is forgetting to disable Time Machine prior to replacing a disk with an existing Time Machine backup. I forget because Time Machine is always on and requires so little attention normally. Here are some pages that describe the shenanigans:
Caveat hax0r: these methods may be out of date. I used them successfully in 2018 with OS X 10.8.
 


I wonder if Sugru would also work? (N.B. I haven't tried it....)
Big fan of Sugru here (as well as Bondic). Fixed the end of the cable from my Thunderbolt Display a few years ago – it was suffering from bend fatigue, and the Sugru stabilized it before it failed completely. Also used it to fix a screw anchor that had pulled out of a plaster wall – saved me from the dreaded cycle of drill-larger-hole-use-larger-anchor. And that fix has held for over a year, and still looks solid.

These newer alternatives to glue are a fixer-upper's dream, in my view.
 


I would very much endorse adding a fast external SSD instead of messing around with a fusion drive. The fusion drive becomes a bootable backup. The T5 disk is so small that one could tape or otherwise affix the SSD to the base of an iMac.
I tried this before moving the 2.5" SSD inside of my mom's 2012 Mac Mini. Even though the SSD was insanely faster than the internal 5400 RPM rust bucket, the USB interface slowed it down.
 


Ric Ford

MacInTouch
I tried this before moving the 2.5" SSD inside of my mom's 2012 Mac Mini. Even though the SSD was insanely faster than the internal 5400 RPM rust bucket, the USB interface slowed it down.
The difference between SATA and USB 3 speeds is minimal, and I wouldn't think noticeable, but if you were using USB 2 (e.g. a USB 2 cable), the difference vs. 6Gbps SATA III would be dramatic.
 


Yes, that's very true for hard drives - clone-erase-restore is a good way to recover from performance degradation that accumulates over time
... (SSDs are unaffected, due to their fundamental technical differences.)
I assume you're referring to the nature of flash cells, wear-leveling and related technologies.

You can probably recover much of the performance if you are able to command the [SSD] to do a Secure Erase or simply TRIM all of the logical blocks, and then let the drive sit idle (but powered) as the on-board garbage collection resets every block. Wait until garbage collection completes (I don't know if there is any way to confirm this, but it maybe good enough to just wait a few hours) before restoring your files.

As a quick-and-simple approach, you might be able to simply have Disk Utility run a "repair" on the erased volumes before restoring files. If macOS supports TRIM on the [SSD], the repair process will TRIM all unused blocks after it finishes its work.
 


Ric Ford

MacInTouch
I assume you're referring to the nature of flash cells, wear-leveling and related technologies....
Actually, I was thinking about the performance penalty of seek and rotation delays on hard drives vs. instant random access for SSDs.

(As you note, though, there is a different sort of accumulative degradation with SSDs, but I believe this is mostly transparent and imperceptible from the user perspective except in extreme cases.)
 


Ric Ford

MacInTouch
FYI, from a solid source for SSD info:
Billy Tallis said:
AnandTech Year In Review 2019: Solid State Drives
In 2019, flash memory prices have leveled out and have even crept back upward a bit, and new technologies have been slow to roll out, although we are currently on the cusp of a PCIe 4.0 revolution. The pace of R&D is still keeping up, so as we move into 2020, we should start seeing plenty of interesting developments build on the backbone of 2019 designs. Here is our Year In Review 2019 for SSDs.
 


The difference between SATA and USB 3 speeds is minimal, and I wouldn't think noticeable, but if you were using USB 2 (e.g. a USB 2 cable), it would be dramatic.
I have an iMac with no USB 3 but FireWire 400. With FireWire 400 (or with FireWire 800 and an adapter), I have an SSD as startup disk, and it runs fairly well (faster than the internal hard drive).
 


Writing a Carbon Copy Clone from an hard disk drive has the side-effect of defragmenting the data as it is written. Years ago (12" G4 PowerBook) I did that and obtained noticeable throughput improvement when the clone was written back to the internal hard disk drive. I think I recall having an application that displayed the amount of fragmentation on my disk and finding CCC worked as a cure. Still works, if there's a boot hard disk drive. Shouldn't matter if SSDs are both source and target.
I have never used Carbon Copy Cloner but do use SuperDuper and it automatically uses the same steps described in the above link to prepare the SSD for backup.
 


Yes, that's very true for hard drives - clone-erase-restore is a good way to recover from performance degradation that accumulates over time, but be careful to get drive selection right and not wipe out the wrong one!
Make sure you turn off FileVault first [with APFS]. I had a problem with the Secure Token after I restored from a Carbon Copy Cloner backup.
 


Regarding a method to affix an external drive to an iMac: Sticky-backed Velcro is your alien technology friend. Place the drive on the iMac's foot, well back by the bend with the cable facing the ports, of course.
By the way, I'll mention that some clients have called when their iMac has failed to boot properly from the external drive; consistently, it's the cable that has been partially pulled out (yes, user error). Be prepared for these calls... or (horrors) use some gorilla snot [Amazon] to keep the cable ends in place.
Heh, I used 3M Clear Velcro (the same stuff for your EZpass), and the external never even budges. I thought of 3M VHB, but that is a bit too permanent.

I wish there were a locking-option for cables like internal SATA cables have for PCs. I agree that the connections are small (MiniDP, MicroUSB, MiniUSB, USB-C) and the cable ends are larger than the connectors. A bump or movement over time will have them slip out of some case designs.
 


Ric Ford

MacInTouch
I wrote about Samsung's new Portable SSD T7 Touch on the MacInTouch home page with its fingerprint security system and performance boost. As a fan of Samsung's current Portable SSD T5, I'm really looking forward to testing the new version.

Meanwhile, Samsung is previewing an ultra-fast update to its NVMe SSD product line:
AnandTech said:
CES 2020: Samsung 980 PRO PCIe 4.0 SSD Makes An Appearance
Samsung's booth at CES 2020 includes our first look at their next flagship consumer SSD, the 980 PRO M.2 NVMe SSD. This would appear to be Samsung's first client/consumer SSD to support PCIe 4.0, which has until now only been rolled out to their high-end enterprise drives. Since this just a low-key preview instead of a formal announcement with a press release, information is limited. The exhibit shows only sequential performance numbers: 6500 MB/s reads, 5000 MB/s writes.
 


I wrote about Samsung's new Portable SSD T7 Touch on the MacInTouch home page with its fingerprint security system and performance boost. As a fan of Samsung's current Portable SSD T5, I'm really looking forward to testing the new version. Meanwhile, Samsung is previewing an ultra-fast update to its NVMe SSD product line:
So, on the T7, it looks like they've updated the internal flash device and bridge to be NVMe, instead of the SATA device in the T5, thus breaking the 540 MB/s SATA speed barrier. On a USB 3.0 host, you won't see any difference between the T5 and T7, but on a USB 3.1/3.2 Gen 2 host, you will get almost twice the performance!

As for the 980 Pro, you'll need a PCIe 4.0 motherboard to see any improvement over current-gen PCIe 3.0 NVMe drives. It won't benefit Thunderbolt 3 drives because of Thunderbolt's 22 Gbps PCIe speed limit.

Interestingly, USB 4.0 drives should actually be faster than Thunderbolt 3. Unlike Thunderbolt, which reserves bandwidth for video signals and caps PCIe data at 22 Gbps, USB4 should be able to use the full 40 Gbps for data transfer if there are no video signals running. In that case, a PCIe 4.0 SSD would be necessary in order to fully saturate that link. We won't know for sure until USB4 is released.
 


Ric Ford

MacInTouch
So, on the T7, it looks like they've updated the internal flash device and bridge to be NVMe, instead of the SATA device in the T5, thus breaking the 540 MB/s SATA speed barrier.
Exactly - as shown in previous benchmarks: post-20751
Interestingly, USB 4.0 drives should actually be faster than Thunderbolt 3. Unlike Thunderbolt, which reserves bandwidth for video signals and caps PCIe data at 22 Gbps, USB4 should be able to use the full 40 Gbps for data transfer if there are no video signals running.
That's very interesting. Meanwhile, I've been wondering if it would ever be possible to connect an external storage device with dual Thunderbolt 3 connections to double performance, considering how Dell's 8K display uses dual DisplayPort connections to achieve its high resolution,
Dell said:
Dell UltraSharp 32 8K Monitor - UP3218K Troubleshooting Guide
The monitor is capable of displaying a max resolution of 8K x 4K @ 60Hz. To display 8K x 4K @ 60Hz:
  1. Check that your graphics card has two DP ports that are compliant to DP1.4 support; HBR3 & DisplayID 1.3 supporting Tile Display feature.
  2. Ensure your graphics card is updated with the latest graphics drivers and VBIOS. Please check your graphic card manufacturer's website for updates.
  3. Connect both DP ports on the monitor to the graphics card.
  4. Ensure graphics resolution is set to 8K x 4K (7680 x 4320) @ 60Hz.
 


how would I check its current firmware version? (I don't see anything about that in Disk Utility,
I have a Crucial MX100 SSD in an Akitio enclosure attached to an Akitio ThunderDock connected to a 2018 Mac Mini with an Apple Thunderbolt 2 to Thunderbolt 3 adapter. In the Catalina 10.15.2 System Report, some drive information is shown under Storage and other information under Thunderbolt, but neither includes the firmware version. Oddly (to me), the firmware version is included under SATA/SATA Express.
 


I have a Crucial MX100 SSD in an Akitio enclosure attached to an Akitio ThunderDock connected to a 2018 Mac Mini with an Apple Thunderbolt 2 to Thunderbolt 3 adapter. In the Catalina 10.15.2 System Report, some drive information is shown under Storage and other information under Thunderbolt, but neither includes the firmware version. Oddly (to me), the firmware version is included under SATA/SATA Express.
Makes sense to me. The MX100 is a SATA device. In order to provide Thunderbolt connectivity, Akitio is almost certainly using a PCIe-SATA bridge chip, taking advantage of Thunderbolt's ability to move PCIe data. As far as software is concerned, your SSD is attached to a SATA controller card installed in a PCIe slot.

They could have, theoretically, used a USB-SATA bridge chip (which is what a USB-C dock would have to do), and that would make your SSD appear as a USB device. But that wouldn't perform as well as a PCIe-SATA chip, so it would not be a good choice for a Thunderbolt dock.
 


OWC has switched some of their Mercury Extreme Pro SSD controllers from Silicon Motion to Phison with their newer black drives. Phison has only 11 smartmon attributes that are reported, compared to 30 with Silicon Motion. I've never heard of Phison. Opinions?

These big SSDs are primarily for updating data like documents, graphics, video, etc. Would APFS be more reliable or wear out the drive less in Mojave, than if formatted MacOS Extended? How about with High Sierra? I'm guessing that APFS is better with Mojave, but maybe MacOS Extended is better for High Sierra, assuming APFS is an older version there.

Of course, there's a current lack of utilities for repairing APFS as a factor, but it sounds like DiskWarrior is working on it, according to comments the author made on an Eclectic Light thread a while back. Any advice is appreciated, thanks.
 


Interestingly, USB 4.0 drives should actually be faster than Thunderbolt 3. Unlike Thunderbolt, which reserves bandwidth for video signals and caps PCIe data at 22 Gbps, USB4 should be able to use the full 40 Gbps for data transfer if there are no video signals running. In that case, a PCIe 4.0 SSD would be necessary in order to fully saturate that link. We won't know for sure until USB4 is released.
USB 4.0 adopts the base Thunderbolt networking protocol. That is a huge assumption that it is going to be better than the other Thunderbolt controller implementations on the first-generation implementations of USB 4.

USB 4 is going to try to route (encode and tunnel) three protocols, not just the two of Thunderbolt up to now. That isn't going to make low-latency tolerance for DisplayPort traffic get any better than what the Thunderbolt controllers have had to do.

The 22Gbps limitation is a limitation of the Thunderbolt controllers so far, not Thunderbolt protocol. There has only been one implementer company so far, [not] multiple implementers with perhaps differing design choices....
 


That's very interesting. Meanwhile, I've been wondering if it would ever be possible to connect an external storage device with dual Thunderbolt 3 connections to double performance, considering how Dell's 8K display uses dual DisplayPort connections to achieve its high resolution,
Transparently to the OS at the hardware level? I don't think so. One could build a "Rube Goldberg"-like peripheral where effectively two external drives shared the same enclosure and the same power supply. There would be two Thunderbolt controllers in the "box", and the storage drive in the box would present two connections back to the host system. Some host RAID driver/software could merge those back into one. To be effective, one would probably need to hook each port on the external drive to a separate Thunderbolt controller ('Thunderbolt Bus' in Apple parlance) on the host system. There wouldn't be much difference there between two boxes, each with their own power supply (either external or through the Thunderbolt connection).

[And] outside of Apple's Macs, there is very little number of implementations that have more than one Thunderbolt controller in the host system. So the market for a device would be very finite. That will change over the upcoming years, but for now, it is a limited space and market.
 


Ric Ford

MacInTouch
There would be two Thunderbolt controllers in the "box", and the storage drive in the box would present two connections back to the host system. Some host RAID driver/software could merge those back into one. To be effective, one would probably need to hook each port on the external drive to a separate Thunderbolt controller...
Yes, good points. A simple approach would be something like this: Connect two Samsung X5 Thunderbolt 3 SSDs to two separate Thunderbolt 3 ports on a modern Mac, and use SoftRAID to put them in a RAID0 configuration, then test performance.

As you note, it would be important to use specific, separate Thunderbolt 3 ports on the Mac, in order to get multiple controllers (and hence more bandwidth). I think this means using one on each side of a MacBook Pro, for example.

As you describe, it's probably impractical to use a multiple-cable/port approach in an attempt to fully utilize throughput for a single, ultrafast SSD (such as those that require PCIe 4).
 


As you describe, it's probably impractical to use a multiple-cable/port approach in an attempt to fully utilize throughput for a single, ultrafast SSD (such as those that require PCIe 4).
Even if Thunderbolt doesn't move in the next 2-3 years to "cover" PCIe v4, it is still faster than the commonly available alternatives. I expect that PCIe v4 (and, even more so, PCIe v5 ) devices will be internally focused ones and not external - and that the pricing for the PCIe v3 will drop and the pricing for v4-5 will be higher. The lower $/GB for v3 will make them more attractive for "sneaker net" data transfers and similar, external drive focused uses. If newer host and peripheral Thunderbolt controllers can max out x4 PCIe v3 throughput with a direct host to peripheral connection, that should be sufficient for a while.

Folks point to desktop AMD processors as to how PCIe v4 is going to rapidly flood the market. Well, AMD's Zen 2 mobile architecture APUs came, and they arrived with PCIe v3. Desktop chips with PCIe v4 arrived with substantial heat sinks and sometimes fans, so mobile skipping it wasn't much of a surprise. Laptops still dominate the general PC market, so there probably won't be a rapid, overwhelming herd of PCIe v4 devices thundering to market.
 



I would appreciate any thoughts or opinions on the viability of using the new Samsung T7 Touch as a MacBook Pro bootable startup disk to run a complete set of macOS, applications, and user account. The goal would be to have a self-contained secure work environment independent of the MacBook hardware. Thanks in advance...

P.S. Would it make any difference to use Mojave for the T7 Touch bootable OS vs. Catalina? (The MacBook Pro is currently running Sierra.)
 


I would appreciate any thoughts or opinions on the viability of using the new Samsung T7 Touch as a MacBook Pro bootable startup disk to run a complete set of macOS, applications, and user account. The goal would be to have a self-contained secure work environment independent of the MacBook hardware. Thanks in advance...
As long as the fingerprint sensor works, then bootable won't matter much, as long as the Mac is configured to allow external boot drives (e.g. custom settings for a Mac with T2). I suspect there is a corner case, though, if you need to unlock the drive with a password to boot. (You might need another Mac to turn off the security if the sensor failed. It wouldn't be completely locked out from data, just booting.)

The manual isn't on the Samsung site yet, but I suspect you can't enable the fingerprint sensor until you have added a password to the drive. The software to adjust security parameters is an external dependency relative to a boot drive function. You have to be already booted up for it to work (so if you can't boot. then it's an issue).

This looks more like a "sneaker net" data drive than a primary boot drive. It can work as a boot drive, but it has constraints.
Would it make any difference to use Mojave for the T7 Touch bootable OS vs. Catalina? (The MacBook Pro is currently running Sierra.)
After unlocking, it should simply be a standard USB drive at the firmware boot phase. The specs at Samsung site say it works with macOS, Windows, and Android, so it's probably not made difficult to work with at the initial boot phases.
 


Ric Ford

MacInTouch
FYI: Amazon now has the Samsung T7 Touch SSD available, and I just ordered one, but it's in very limited supply. Grab it fast if you want one soon.
I received the 1TB Samsung Portable SSD T7 Touch today that I had ordered from Amazon, and I'm starting to evaluate it. A few initial notes:
  • It's a bit larger and a bit thinner than the Samsung T5.
  • The T7 Touch has a blue, four-sided light around the fingerprint sensor. The light "rotates" around the square during data access. (The T5 has a small round light near the USB-C port that blinks during data access.)
  • The tiny "manual" with its microscopic type doesn't seem to talk about security/software setup but does warn that forgetting/losing the password renders the data/device unusable.
  • Both USB-C and USB Type-A cables are included (with USB-C plugs on the opposite end for the T7 Touch).
  • Ubuntu Linux saw the volume when I plugged in the T7 Touch but wouldn't mount it.
  • Connecting to a Mac running macOS 10.12 Sierra mounted a volume in exFAT format that contained installers for macOS and Windows, plus a text file with a link for getting Android software.
  • The macOS installer is signed by Samsung and contains 20.8 MB of files, including a kernel extension (kext) and SamsungPortableSSD_1.0.app. (The app lists OS X 10.7 as its minimum requirement.)
 


...The manual isn't on the Samsung site yet, but I suspect you can't enable the fingerprint sensor until you have added a password to the drive.
Found the manual on Samsung's semiconductor site (the US computer accessory site doesn't have it).
Samsung said:
Booting appears to be a problem because this looks like they aren't doing entirely self-contained encryption, so the password (and fingerprint also, it appears - they don't separate the two) needs the software to decrypt.
Samsung said:
Only Mac OS “Samsung Portable SSD” driver is required for the security functionality.
That could be that it can't manage the security function via keyboard/mouse without the driver. If that includes the touch sensor, then that is trouble for boot. The manual later makes it clear that if you enable the password mode and have something other than Windows or macOS, things may not work right.

If the touch sensor "unlock" is all self-contained on the drive (earlier I presumed it was) then this could work. It is a bus-powered drive so will need power to make the touch sensor work (so the Mac would have to be in Option key 'waiting to finish booting' mode).

For boot it isn't good to have an optional 3rd-party driver required [because] it isn't going to make the EFI boot driver list.
 





Booting appears to be a problem because this looks like they aren't doing entirely self-contained encryption, so the password (and fingerprint also, it appears - they don't separate the two) needs the software to decrypt.
That's not my reading of the manual. Although it's not made explicit, it seems pretty clear that the fingerprint reader is self-contained and fully unlocks the drive all by itself.

The drive can operate in 3 modes. "Security mode off" in which it acts like any other DAS [Direct-Attached Storage] device, "Security with Password", and "Security with Password and Fingerprint".

All the warnings about the drivers/software and forgetting passwords refer to the Security with Password mode. And make it clear that you can't use the Password-only feature with OSes other than Windows, macOS, and mobile, because of the requirement for the software to unlock the drive.

So, with Fingerprint enabled, it should be possible to use it as a boot drive, as you just have to scan your fingerprint whenever you start-up and the drive is locked (the lights will let you know if the drive is still locked).

Alternatively, you could operate with "Security mode off", making it a normal drive, and just rely on good-old FileVault to encrypt your data.

#security
 



Amazon disclaimer:
As an Amazon Associate I earn from qualifying purchases.

Latest posts