MacInTouch Amazon link...

USB, FireWire & Thunderbolt

Channels
Troubleshooting, Products
Switching to the Type A port on the front should solve the problem.
That appears to be true, but I am still confused about the capabilities of the various ports.

I used an AmazonBasics USB Type-C to USB-A Male 3.1 Gen2 Cable and connected to the USB Type A port on the front of the home CalDigit dock. System Information shows:
Current available 900 mA
Current requried 896 mA
Extra Operating Current 0 mA
The CalDigit kernel extension was not loaded. No "USB Accessories Disabled" warnings.

I did the same thing with the CalDigit dock I have at work:
Current Available (mA): 900
Current Required (mA): 896
Extra Operating Current (mA): 600
Sleep current (mA): 1500
The kernel extension is loaded. No warnings.

Connecting to the USB-C Thunderbolt port on the back of the dock with the cable that came with the iPad Pro, I see for both docks:
Current Available (mA): 900
Current Required (mA): 896
Extra Operating Current (mA): 600
Sleep current (mA): 1500
I see the extension loaded on the work dock, but it is inconsistently loaded on the home dock. No warnings (well, maybe one but certainly not a regular event).

Then I tried the USB Type-C port on the front with the iPad Pro cable and I see on either dock:
Current Available (mA): 500
Current Required (mA): 500
Extra Operating Current (mA): 1000
Sleep current (mA): 1500
The CalDigit kernel extension is not loaded in either case. This is the configuration that will generate OS warnings.

Surprisingly, to me, this Nonda Type-A to Type-C adapter and a USB 3 cable did not allow the iPad to be charged when connected to the USB Type-A port on the front of the dock, and the iPad did not show up in iTunes.

My conclusions are that Ric is correct in that using the Type A port on the front of the CalDigit dock seems to provide sufficient current to avoid the port being disabled by macOS but that the Thunderbolt port on the back is also adequate.

I am not sure about the utility, or need for, the CalDigit kernel extension. I also conclude that I don't know much about these different current categories as reported by System Information.

Thanks, Ric!
 


I have a Dell U2415 [display] connected to an OWC 14-port Thunderbolt 3 dock that is connected to the 2018 MacBook Pro 15" (model 15,3 with Radeon Pro Vega 20).
Prior to macOS 10.14.5, the U2415 came on as it is supposed to on a bootup and restart. Now it comes on with a normal boot, but on a 'restart' it does not, and the MacBook Pro will not recognize it, no matter what sort of futzing I do with it, including using the USB-C video adapter.
Oddly, on a hunch on a restart I pulled the Mini DisplayPort plug from the OWC dock and plugged it back in after the progress bar started, and the U2415 lights up. Weird.
I have had problems with the OWC dock from the get-go (I have the ThunderBay connected to the Thunderbolt 2 dock...). When the Mini would sleep, I had to unplug the Thunderbolt from the dock. Otherwise, my Dell display (1 of 2) would not wake, no matter what I did.

The USB port would not recognize the iPod or iPhone, even though it was there in iTunes. If I tried to do an update/backup for iOS update, I had to plug the device into the Mac's USB port and not the OWC dock ports. Also, it gets really hot.

I am considering losing the OWC dock for CalDigit.
 


I am considering losing the OWC dock for CalDigit.
Have you discussed your issues with OWC?

I no longer need to make all the laptop-to-peripheral connections I used to, when I traveled to a remote office, so my OWC Thunderbolt Dock has been in a closet for over a year. Back when I did use it, I'd occasionally have vexing problems when I'd first arrive at the remote office, but the monitor (an Apple Mini DisplayPort biggie) wouldn't balk at waking up from laptop sleep once the initial connection had succeeded.

Sometimes my issue would seem to stem from not having the laptop plugged into a/c power, sometimes not. Sometimes popping open the laptop's own display would help, sometimes not. I never did figure out the correct connection sequence and incantations to make it never happen, however.

My only support calls to OWC were when the "ship" dates slipped time after time when the whole "connectology" world was turned on its head at the dawn of the USB-C / Thunderbolt 3 / dongle-Medusa era.
 


Here's hoping for a dual-drive dock with Thunderbolt 3! (And to being much less expensive than a full-fledged Thunderbolt 3 dock, like the CalDigit TS3 Plus.)
 


Here's hoping for a dual-drive dock with Thunderbolt 3! (And to being much less expensive than a full-fledged Thunderbolt 3 dock, like the CalDigit TS3 Plus.)
The $595 BlackMagic MultiDock 10G looks like a great product, and I suspect, as USB-C not Thunderbolt, might prove more reliable.

Value per dollar does matter, and even at $595, the MultiDock seems to offer value compared to Thunderbolt alternatives.

For more value, the 10Gbps StarTech USB-C 3.1 to SATA adapter, currently $18.38 on Amazon, should deliver the same throughput to a single SATA SSD. (Per StarTech's website, it supports UASP.) Wonder how it would work connected to multiple disks through a high quality USB-C powered hub? The possible advantage of Thunderbolt is TRIM, if it's effectively passed over that connection.
 


Amazon sells the OWC "OWC Drive Dock" for the same price as OWC. Reviews are not encouraging. "Disconnects" seem to be the complaint....
I've a OWC Thunderbolt 2 dock and can confirm: it locks up the display, has wake issues if I put the Mac Mini to sleep more than 2 hours, is chained to a OWC ThunderBay, has issues with some USB devices mounting, gets extremely warm, and needs a driver (rather dated).

I may move to a CalDigit in the future, but for now, I turn the display off and disable sleep. Or I may just replace my CCFL Dell with an LED model (HDMI) and then leave Thunderbolt to the dock and ThunderBay, using HDMI for display.
 


Ric Ford

MacInTouch
Just did a quick test*, using dd to copy a single 3.6GB file from internal SSD to a Samsung T5 on USB 3.0 (2015 MacBook Pro, macOS 10.12):
Tried it with a Thunderbolt network and file-sharing

2017 iMac <-> Thunderbolt 3-Thunderbolt 2 adapter <-> 2015 MacBook Pro

58.6 MB/sec.

So, about half the speed of a USB 3 SSD.
 


*Here's the test:
Bash:
sudo time dd if=~/3.6GBsourcefile of=/Volumes/external/copyfile
If you're using dd with the default block size, then you probably also enjoy watching paint dry. If you add bs=2M to the dd command, you'll get much more realistic transfer performance. Or, even better, ditch dd entirely, and just use something like cp.
 


Per this "File Transfer Time Calculator", unloading 960 GB at the 20Gbps Thunderbolt 2 speed would require 6m 24s and just 3m 12s over Thunderbolt 3.
So, just like the 22 Gbps cap on PCIe speeds that Thunderbolt 3 has, I found that Thunderbolt 2 similarly has an 11 Gbps cap on PCIe transfer speed. I found this illustrated in this Intel slide:

https://www.aja.com/assets/support/files/3087/en/Thunderbolt2_BandwidthAllocation_2.pdf

What that means is that an external NVMe SSD drive connected to Thunderbolt 2 would only be about 10% faster than being connected to a USB 3.1 Gen 2 port (10 Gbps).

I was not able to find any comparable Intel information on Thunderbolt 1, but I assume it must have a similar data transfer cap. 5.5 Gbps?
 


Ric Ford

MacInTouch
Just did a quick test*, using dd to copy a single 3.6GB file from internal SSD to a Samsung T5 on USB 3.0 (2015 MacBook Pro, macOS 10.12):
116 MB/sec.​
If you're using dd with the default block size, then you probably also enjoy watching paint dry. If you add bs=2M to the dd command, you'll get much more realistic transfer performance. Or, even better, ditch dd entirely, and just use something like cp.
Who knew? Here's the same test using cp instead of dd:

450 MB/sec.​
Copying the file between two volumes on a 2015 MacBook Pro's internal SSD:

593 MB/sec.​
Copying the file from a 2017 iMac 5K’s internal SSD to a Samsung X5 Thunderbolt 3 SSD:

1166 MB/sec.​
 





Ric Ford

MacInTouch
Just did a quick test*, using dd to copy a single 3.6GB file from internal SSD to a Samsung T5 on USB 3.0 (2015 MacBook Pro, macOS 10.12):
116 MB/sec.​
If you're using dd with the default block size, then you probably also enjoy watching paint dry. If you add bs=2M to the dd command, you'll get much more realistic transfer performance.
Sorry, apparently, it wants a lower-case letter, so use bs=2m
Who knew? Here's the same test using cp instead of dd:
450 MB/sec.​
And the same test using dd bs=2m ...

444 MB/sec.​
Copying the file between two volumes on a 2015 MacBook Pro's internal SSD:
593 MB/sec.​
Switching to dd bs=2m gives:

1165 MB/sec.​
 


If you're using dd with the default block size, then you probably also enjoy watching paint dry. If you add bs=2M to the dd command, you'll get much more realistic transfer performance. Or, even better, ditch dd entirely, and just use something like cp.
Very good—thanks for the very worthwhile tip!
 


Here's a benchmark test of one of the first PCIe 4.0 NVMe SSDs. I found it interesting, because it is the only published benchmark of direct NVMe to NVMe copy I've found. It's peformed on the new ASUS X570 motherboard that supports PCIe 4.0 for AMD's recently recleased Ryzen processors. The motherboard is backward-compatible with earlier PCIe versions, which allowed the benchmarks to be performed copying to and from a Corsair PCIe 4.0 drve and a Samsing PCIe 3.0 SSD.
The Guru of 3D said:
Corsair MP600 PCIe 4.0 NVMe SSD review - Final Words & Conclusion
Copying many many gigabytes of movies and ISOs for example, well this SSD laughs at it, and while copying I actually laughed a bit nervously. I mean our 110 GB test file I had to copy towards this SSD from a Gigabit NAS, that took 18 minutes. Then copying the same file from an 870 EVO Plus towards the MP600 ... far less than a minute, bat-poop crazy.
So the results are a bit all over the place really, but it's also priced fairly competitive. I mean this super fast performing M.2. unit costs ~25 cents per GB in retail right now. The unit reveals speeds at 2GB/ to 4.5 GB/s reads sometimes, in writes things are varying more. IOPS performance is great as well, but it does need massive queues and preferably threads for it to be able to show.
Hardware Secrets said:
Everything You Need to Know About the PCI Express
... PCI Express is a point-to-point connection, i.e., it connects only two devices.... On a motherboard with PCI Express slots, each PCI Express slot is connected to the motherboard chipset using a dedicated lane, not sharing this lane (data path) with other PCI Express slots.
 


The new Raspberry Pi 4 announced last month looks like it will be a better desktop machine to learn Linux on, once they patch the USB-C power issues (they used non-standard firmware,so many USB-C cables don't supply power).
They tried to be "clever" (saving the cost of an additional resistor) and didn't follow the USB-C power spec exactly, but what they left out does make a difference - in certain circumstances. It only affects you if you use an e-marked cable (which apparently is only found on expensive USB-C chargers, like Apple laptop chargers) - the cable will not supply power to the Raspberry Pi, because the lack of the resistor in the Pi means that the cable thinks the Pi is an audio device. The Pi works fine with normal (most, I believe) cables. Obviously, the lack of a resistor will not be fixed by a firmware upgrade. They say they will fix it in a future revision of the board.

Full details here:
 


Ric Ford

MacInTouch
Here are a few of the benchmark results I've posted previously (you can probably search for "Blackmagic" to find many of them):
#benchmarks
I just put a 2017 iMac 5K in Target Disk Mode and connected it over Thunderbolt 3 to a 2018 MacBook Pro. Blackmagic Disk Speed Test says:

Write: 895 MB/s​
Read: 468 MB/s​

That's an awful lot slower than the 40 Gbps (or even 22 Gbps) that Thunderbolt 3 should provide (i.e. ~3000 MB/s).

#benchmarks
 


Ric Ford

MacInTouch
I just put a 2017 iMac 5K in Target Disk Mode and connected it over Thunderbolt 3 to a 2018 MacBook Pro. Blackmagic Disk Speed Test says:
Write: 895 MB/s​
Read: 468 MB/s​
That's an awful lot slower than the 40 Gbps (or even 22 Gbps) that Thunderbolt 3 should provide (i.e. ~3000 MB/s).
I flipped them around and tested the 2018 MacBook Pro in Target Disk Mode, connected via Thunderbolt 3 to the 2017 iMac 5K:

Write: 466 MB/s​
Read: 273 MB/s​

Wow, pretty pathetic performance...
#benchmarks
 


I just put a 2017 iMac 5K in Target Disk Mode and connected it over Thunderbolt 3 to a 2018 MacBook Pro.
I’d be interested to see your results if you connected them together by Thunderbolt networking and mount one as a network share (SMB and/or AFP) on the other. I believe you should be able to select a network volume for testing in Blackmagic Disk Speed Test.
 


Ric Ford

MacInTouch
I’d be interested to see your results if you connected them together by Thunderbolt networking and mount one as a network share (SMB and/or AFP) on the other.
I spent quite a while trying but was unable to get file-sharing to work, for unknown reasons. I had tried it previously between Thunderbolt 2 and Thunderbolt 3 systems, and it was much slower than a USB 3.0 SSD, which seemed like a better option for sync'ing files between systems.
 


They tried to be "clever" (saving the cost of an additional resistor) and didn't follow the USB-C power spec exactly, but what they left out does make a difference - in certain circumstances. It only affects you if you use an e-marked cable (which apparently is only found on expensive USB-C chargers, like Apple laptop chargers) - the cable will not supply power to the Raspberry Pi, because the lack of the resistor in the Pi means that the cable thinks the Pi is an audio device.
I may have read the article wrong, but it appears that what's going on is that with this (mistaken) resistor configuration, the Pi thinks the power brick is the audio device. As such, it is trying to supply current out the port (as it would have to if it was powering a pair of headphones, for example) instead of sinking current from the port.

If you don't sink any current, you don't power on.
 


it appears that what's going on is that with this (mistaken) resistor configuration, the Pi thinks the power brick is the audio device.
I've read that, too. When Google's Pixelbook was brand new, I bought one. Grrrr. Google's "state of the art" had only two USB-C ports, where my older Chromebook had a charge port (unfortunately, proprietary), two USB Type-A, MicroHDMI, and a MicroSD slot, in which I kept a 128GB card equal to to the $1,100 Pixelbook's internal eMMC storage. At least both devices had headphone jacks.

I purchased a USB-C dongle/dock that passed through the Pixelbook charging, provided USB Type-A, HDMI, and Ethernet. Did not work well.

I returned the Pixelbook within my Best Buy 14-day window. There may be a place for ChromeOS, and given the Pixelbook's self-constrained connectivity and the nature of ChromeOS, my Rockchip $200 Asus Chromebook Flip was a lot better match. After my experience with the Pixelbook, I've empathy for purchasers of similarly constrained and much more expensive Mac laptops.

That's a long detour from the Pi but does loop back. If the Pi's USB-C were properly designed, it should be possible to extend Pi connectivity with a dongle/dock like I tried with limited success on the Pixelbook. Perhaps that's in the Pi's future, perhaps it's just irrelevant to the Pi use case, but it's not going to happen if Pi's designers don't stick to published standards.
 


Ric Ford

MacInTouch
...I purchased a USB-C dongle/dock that passed through the Pixelbook charging, provided USB Type-A, HDMI, and Ethernet. Did not work well....
My sense from reading customer reviews (e.g. at Amazon or Apple Store) is that many of these USB-C mini-dock type devices are cheap, low-quality products with a relatively high risk of defects — probably more an issue of parts and manufacturing quality than a matter of meeting standards (though I suppose the two can blend). I've also had an unexpected amount of problems with simple USB 3 hubs from several different companies.
 


If the Pi's USB-C were properly designed, it should be possible to extend Pi connectivity with a dongle/dock like I tried with limited success on the Pixelbook.
If they were shipping a proper data+power USB-C port, you would be absolutely right, but that doesn't seem to be their goal. The USB-C port on the Pi 4 is just a power port. There is no data transfer at all on that port. All USB data is via four type-A ports (two USB 2.0 and two USB 3.0).
 


I just put a 2017 iMac 5K in Target Disk Mode and connected it over Thunderbolt 3 to a 2018 MacBook Pro. Blackmagic Disk Speed Test says:
Write: 895 MB/s​
Read: 468 MB/s​
That's an awful lot slower than the 40 Gbps (or even 22 Gbps) that Thunderbolt 3 should provide (i.e. ~3000 MB/s).
I think Apple would say this is "expected". Target Disk mode doesn't directly tie together the Thundebolt (or FireWire) bus and the drive controller with hardware or high-performance firmware, as a good external enclosure would. It's just a convenience for occasional use, implemented in the simplest possible way.

I share Todd's curiosity about what numbers one might see from a Thunderbolt 3-connected "point to point" networked drive, served by a machine with a very fast SSD. It would have to be SMB, I think, but if the protocol implementations are highly tuned, it wouldn't completely surprise me if it were faster than target disk mode.
 


Ric Ford

MacInTouch
I share Todd's curiosity about what numbers one might see from a Thunderbolt 3-connected "point to point" networked drive...
I spent more unsuccessful time yesterday trying to get this to work, and I have no idea what's wrong. Things I tried:
  • disabling firewalls
  • turning off WiFi (so the Thunderbolt connection was top on the Network list)
  • various SMB and AFP combinations/options
  • swapping Thunderbolt ports (though I didn't try every possible combination, and both computers showed "Active" Thunderbolt network connections)
  • different access methods (hostnames, IP addresses, protocols)
  • logging into different user (admin) accounts
It has been pretty frustrating.
 


many of these USB-C mini-dock type devices are cheap, low-quality products with a relatively high risk of defects
I've had those problems, too. Trying to avoid them, the "dongle/dock" I bought was from Anker, a brand which I've found reliable, and at $70, I don't consider it "cheap."

Hard to blame the gadget, though Amazon currently lists it as "upgraded." No way to know if what was "wrong" was the gadget itself, or ChromeOS, or Google's intentional implementation.

The USB-C port on the Pi 4 is just a power port.
Maybe we should consider the "defect" to be a "feature", as I'd suspect the 100 watts that USB Power Delivery could push could be dangerous?

It's the presence of that "universal" Type-C port that frustrates, because it suggests more options.
 


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

Latest posts