Thunderbolt: How devices affect each other on a daisy chain
The Macworld Lab has a small collection of Thunderbolt peripherals, and we thought we should put them to work to answer questions we had about connectivity and how devices affect performance.
After testing dozens of scenarios, we found that—for the most part—the two available bi-directional 10Gbps channels in the MacBook Pro (Late 2011) were more than able to keep up with the demand of multiple storage devices on a Thunderbolt daisy chain. However, if you add multiple displays to that chain, the throughput of some drives can be severely limited.
(For your reference: The iMac (Mid 2011) uses the same Light Ridge Thunderbolt controller as the MacBook Pro, but the iMac has two Thunderbolt ports and offers up to four bi-directional, 10Gbps channels. The MacBook Air (Mid 2011) uses an Eagle Peak Thunderbolt controller that allows for two bi-directional 10Gbps channels.)
Drives in a daisy chain
The Thunderbolt drives involved in our testing were the Promise Pegasus R6 RAID array, the LaCie Little Big Disk Thunderbolt Series SSD, and the LaCie Little Big Disk Thunderbolt Series 2TB hard drive.
To get a set of baseline speeds to gauge performance, we attached one drive at a time, with no other devices, to a 17-inch 2.4GHz Core i7 MacBook Pro. We ran the AJA System Test several times—specifically, the 2GB, 1920-by-1080, 10-bit RGB test—and calculated average performance speeds. The Pegasus R6 was the fastest device we tested, with a write speed of 538.1MBps and a read speed of 519.3MBps. The LaCie SSD posted a write speed of 253.0MBps and a read speed of 480.5MBps. The LaCie hard drive clocked in with a write speed of 184.7MBps and a read speed of 203.1MBps.
We then attached all three storage devices to the MacBook Pro, with the Pegasus R6 first in the chain connected directly to the MacBook Pro. The LaCie SSD was second in the chain, and the LaCie hard drive was last. When we ran the AJA System Test again, the Pegasus R6 saw a slight performance improvement, with a write speed of 543.0MBps, and a read speed of 520.3MBps.
The LaCie SSD, second in the chain (attached to the Pegasus R6), scored about the same as it did when connected alone, with a write speed of 246.9MBps and a read speed of 469.4MBps. The LaCie hard drive, the last device on the chain (attached to the LaCie SSD), posted scores that were a touch faster than when attached alone: a write speed of 190.1MBps, and a read speed of 205.4MBps.
We then changed the order of the devices on the chain, placing the Pegasus R6 last on the chain, and the LaCie SSD first, with the LaCie hard drive in-between. The Pegasus R6 scores were just a touch slower, with a write speed of 532.8MBps, and a read speed of 513.2MBps. With the LaCie SSD connected to the MacBook Pro, it scored about the same as it did in the second position. The same was true for the LaCie hard drive.
Drives and a display in a daisy chain
In our next series of tests, we added an Apple Thunderbolt Display to the end of the chain. All of the drives turned in speeds very similar to the speeds they posted without the display. Swapping the Thunderbolt Display with an Apple 27-inch LED Cinema Display didn't change performance much, either.
With the Thunderbolt Display as the first device (and the only display) in the chain, the LaCie SSD experienced a small performance hit. There was no real performance change on the LaCie hard drive.
Daisy chain with two external displays, three drives
Things got interesting when we connected two Thunderbolt Displays to the end of the chain. Our results were similar to the results we saw in a report by AnandTech. Having multiple displays in a chain affects performance.
The LaCie SSD went from 249.7MBps (write) and 472MBps (read) with one Thunderbolt Display attached, to 153.9MBps (write) and 438.8MBps (read) with two Thunderbolt Displays on the chain. The LaCie hard drive's read performance was unaffected, but the write speed decreased from 183.9MBps to 159.4MBps.
The Pegasus R6 saw the most dramatic change, with its write speed dropping from 553.7MBps to 231.0MBps, and its read speed dropped from 527.4MBps to 420.7MBps.
Granted, a setup with a pair of two Thunderbolt Displays, a Pegasus R6, a LaCie Little Big Disk SSD, and a LaCie Little Big Disk hard drive is pretty expensive ($8,593 from the Apple Store). And since most Thunderbolt peripherals available now tend to be for high-end use (RAID arrays, video capture boxes, etc.), the users involved in intricate Thunderbolt daisy chains tend to be professionals. But eventually, as Thunderbolt peripherals proliferate and come down in price, even casual Mac users will likely be creating chains of Thunderbolt-equipped devices
We also found that putting one of the Thunderbolt Displays at the beginning of the chain and putting the other Thunderbolt Display on the end didn't affect the performance of the Pegasus R6, but this configuration did slow down the write speed of the LaCie SSD from 153.9MBps to 139.3MBps, and the LaCie hard drive from 159.4MBps to 144.1MBps. Read speed was also a bit slower on the SSD in this chain, going from 438.8MBps to 412.8MBps.
With a Thunderbolt Display still at the beginning of the chain, we swapped the Thunderbolt Display at the end of the chain with a Cinema Display. The Pegasus R6 regained some of its speed, with a write speed of 433.4MBps and a read speed of 440.8MBps. The LaCie hard drive also benefitted, with speeds similar to when there was just a single Thunderbolt Display attached to the beginning of the chain. The LaCie SSD regained some ground, with a write speed of 208.5MBps and a read speed of 427.8MBps.
iMac, two displays, three drives
To see if the MacBook Pro was a factor in performance, we ran some tests using a build-to-order 27-inch 3.4GHz Core i7 iMac. We tested the Pegasus R6 by itself and saw a write speed of 533.9MBps and a read speed of 519.5MBps. With a Thunderbolt Display attached before the Pegasus R6 in the chain, we saw a write speed of 527.7MBps and a read speed of 526.8MBps. When we attached a Cinema Display to the Pegasus R6 (with the Thunderbolt Display still at the beginning of the chain), the Pegasus R6's write speed dropped to 430.1MBps, and its read speed dropped to 432.8MBps. Placing the Pegasus R6 between two Thunderbolt Displays dropped the write speed to 231.6MBps, while read speeds dipped, but at 396.9MBps, not by as much.
We also ran a few tests to see if the performance degradation would show up in everyday use. We copied a 6GB file from the iMac to the Pegasus R6, duplicated the file on the Pegasus R6, and then copied the duplicate file back to the internal SSD on the iMac. We saw a Pegasus R6 write speed of 225MBps in two configurations: with just a single Thunderbolt Display attached to the array, and then with the Thunderbolt and Cinema Display attached. With two Thunderbolt displays, the write speed slowed down a bit to 214.3MBps. Duplicating the file on the Pegasus R6 showed bigger differences, with 299.0MBps with no display attached, 297.5MBps with a single Thunderbolt Display attached, 247.1MBps with a Thunderbolt Display and a Cinema Display attached, and just 163.8MBps with two Thunderbolt Displays attached. When reading the files back to the internal SSD, we saw less of a performance difference, with speeds ranging from 184.0MBps with no display attached to 178.0MBps with two Thunderbolt Displays connected.
The reason for the performance decrease in a two-display setup is two-fold. First, the Pegasus is fast and can put plenty of bandwidth to use. Second, each Thunderbolt Display gobbles up about 7Gbps of an outbound channel's available 10Gbps bandwidth. Display traffic is one-way, so that's why the read speeds from the Pegasus aren't as affected. To test this out, we attached the second Thunderbolt Display to the iMac's second Thunderbolt port and found performance to come back to the level of a setup with just one Thunderbolt Display attached.
Thunderbolt Software Update 1.1
During our testing, we found that the Pegasus R6 was faster before applying the Thunderbolt Software Update 1.1. The file transfer tests we ran on the iMac were very close to the tests runs with the update applied, but our file duplication tests saw speeds as fast as 341.7MBps without the update, versus a peak of 299.0MBps with the update.
Using the AJA System Test with the iMac, we saw a peak write speed of 644.3MBps and a peak read speed of 521.3MBps from the Pegasus R6. Interestingly, if we restarted the Mac with the Pegasus R6 connected, but did not restart the array, write speeds would drop to about 559.9MBps. Read speeds weren't affected much, with speeds averaging 525.4MBps. Restarting the Pegasus R6 would bring the write speeds back up to the 644MBps range.
Once we installed Thunderbolt Software Update 1.1, the scores stayed the same whether we restarted or not. The Pegasus R6's write speed of 534MBps, however, was a drop from the 644MBps we saw before installing the update. The Pegasus R6 had a read speed of 520MBps. Neither of the LaCie drives were affected by the update.
One other thing we found with the Pegasus R6: If you want to play audio through the built-in speakers on a Thunderbolt Display, attach the array to the display. Don't attach the display to the Pegasus R6—the audio played through the Thunderbolt Display's speakers would distort, stutter and stop when running the AJA System Test on the Pegasus R6. Audio playback wasn't affected when the display was ahead of the Pegasus in the chain. With the LaCie drives, the audio played fine during the AJA tests, regardless of the drives' position on the chain.
Have your own burning Thunderbolt questions? Let us know in the comments.
James Galbraith is Macworld's lab director.