Looks like no one’s replied in a while. To start the conversation again, simply ask a new question.

Dual Thunderbolt Displays - Big Performance Hit on RAID and CPU

Short and broad version:


I have two Thunderbolt Displays connected to my MBP in addition to a Thunderbolt RAID and a Thunderbolt Matrox video I/O device. I've found that the performance of the RAID (read/write) is drastically affected by the presence of two Thunderbolt Displays - even though I have only filled 4 of the 6 devices which are supported by a single Thunderbolt chain. Though it's harder to see the performance drop as clearly as you can in a disk speed test, it appears the performance of the Matrox device as well as my CPU are also hit in a major way by the presence of the displays (CPU utilization is up, etc).

I've done a lot of reading and I'm not sure there's a solution hidden anywhere in these forums. As far as I can tell, these Thunderbolt displays suck up so much bandwidth that my laptop's single bus can't even come close to supporting the 6 devices it's supposed to. At this point am I just going to have to wait for the new Mac Pro (and it's multiple Thunderbolt busses) to get releif? Thanks for any help/info you can provide!



Long and detailed version:


I'm a video editor, and I have a 2011 17" 2.2 GHz i7 MacBook Pro with 16 GB RAM which drives two Thunderbolt displays and a Matrox Mini MXO2 video breakout box, also connected via Thunderbolt so I can run an HDMI monitor for clients. Up until a month ago, I've been running FCP 7, After Effects, Avid, etc on my system with no problems; both screens looked and performed amazingly and my machine, although getting a little older now, worked well. The problem is, with new video projects having increasingly demanding specs (increased resolutions, higher bit rate compressions, etc) I noticed my eSata RAID enclosure was starting to bottleneck me. So I decided to look into upgrading myself to a Thunderbolt storage solution for hopes of taking advantage of the breakneck speeds, and also to future proof myself for a while. Instead of going for a Pegasus RAID as many did in these forums, I eventually decided to go with the Areca ARC-8050, an 8 bay Thunderbolt RAID enclosure that would allow me to bring my own hard drives to the party.


So I filled the thing with 3TB drives, set it up for RAID 6, leaving me with 18TB of useable space and set off to speed testing. I ran the Blackbagic Disk Speed Test and was not met with the amazing results I'd expected.



DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 4

Chain order: MacBook Pro > TB Display 1 > TB Display 2 > ARC-8050 Enclosure > Matrox MX02 Mini

Results: 240 MBs Write & 300 MBs Read

User uploaded file


While this was still a significant boost from my old eSata setup, I wasn't getting the near-SSD speeds I was seeing in benchmarks I had found online. So I scratched my head a bit, did some searching and stumbled upon an article claiming dual thunderbolt displays can cause a significant performance hit to a single Thunderbolt chain. GULP. Needless to say I was pretty disappointed with what I was reading, I'd just put a good $4,000 into building this Thunderbolt-heavy system, not counting my MacBook Pro itself. Here I am with only 4 devices on this Thunderbolt chain, which is spec'd out for 6, yet I'm maxing it out!? Here's the article in case you wanted to take a look: http://www.fcp.co/hardware-and-software/pro/715-does-thunderbolt-slow-down-when- daisy-chained-with-an-extra-display


So I went ahead and took one of my Thunderbolt displays out of the equation and tested the whole thing again. Sure enough my drive's performance nearly doubled! These were the kind of numbers I had expected.


SINGLE THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain order: MacBook Pro > TB Display 1 > ARC-8050 Enclosure > Matrox MX02 Mini

Results: 594 MBs Write & 467 MBs Read

User uploaded file


What's most puzzling is that while the original dual display setup only yields 240 Read / 300 Write (which is still considerably faster than my old eSata drive could manage) real world performance in FCP and AVID is drastically worse than it had been with my eSata solution. I can't edit at all with both Thunderbolt displays in use, even with very low bitrate content - the footage just sputters chopily as if I'm taxing a 2006 machine with a 5200rpm drive. I've essentially bought and built a state of the art system that functions worse than any old one I've ever used!


So in closing, I'm not sure if this is a problem specific to you all with Pegasus drives, but instead a problem with the Thunderbolt technology / Apple's displays / or the way Apple is marketing Thunderbolt as an all-in-one solution for professionals. These Thunderbolt displays apparently suck up so much bandwidth to drive their 2560x1440 resolutions that my laptop's single bus can't even come close to supporting the 6 devices it's supposed to. At this point I've pretty much resigned to hoping that the new Mac Pro, with it's multiple busses of Thunderbolt 2 ports, will be my saving grace. While I wouldn't trust the claim that it can handle 36 devices via Thunderbolt, I'm hopefull it can handle the measely 4 I need!


I'm crossing my fingers that someone will come up with a firmware/software fix, but at this point I'm guessing I spent $4,000 only to learn that I need to drop another $4 or $5 grand when the Mac Pro comes out!


Thanks again for reading this long winded thing!

MacBook Pro (17-inch Early 2011), Mac OS X (10.7.5), Dual Thunderbolt Displays

Posted on Aug 26, 2013 10:38 AM

Reply
Question marked as Best reply

Posted on Aug 27, 2013 1:06 AM

Have you tried changing the order of the Thunderbolt devices? e.g. connecting the disk drives between the computer and the displays, or between the two dsplays.

7 replies

Aug 27, 2013 10:45 PM in response to spence3eb

OK, I did some tests on how changing the Thunderbolt chain order affects disk speed and overal system performance. First I tried putting the RAID in the front of the chain:


DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain Order: MacBook Pro > ARC-8050 Enclosure > TB Display 1 > TB Display 2

Results: N/A


Unfortunately, if I connect to the ARC-8050 RAID enclosure before daisy chaining to the monitors, my setup doesn't work. The enclosure itself powers down when I close the lid on my laptop and the screens behind it on the chain do not remain powered, so it seems I can't make the Thunderbolt Displays my primary displays without one of them being the first in the chain. So I think this configuration is out.


Next I tried moving the RAID enclosure between the two Thunderbolt Displays in the chain. Here I actually saw a moderate increase in the READ speed compared to having the RAID come after both displays in the chain (increase from 300 MB/s to 364 MB/s) but nothing so definitive that I can call this a cut and dry winner.


DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain order: MacBook Pro > TB Display 1 > ARC-8050 Enclosure > TB Display 2

Results: 237 MB/s Write & 364 MB/s Read


User uploaded file



I also bought a license for Geekbench 3 today in order to test the impact of the Thunderbolt Displays on more than just my RAID's disk speed - namely overall system performance. The numbers are pretty shocking, here are my results:


KEY: (Single-Core Score | Multi-Core Score) (All 64-bit Benchmarks)


Test 1 - Laptop Alone (Early 2011 17" MBP 2.2 GHz i7 16MB RAM)

2775 | 9641

Test 2 - Laptop with Areca ARC-8050 RAID attached

2775 | 9398

Test 3 - Laptop with One Thunderbolt Display

2699 | 8894

Test 4 - Laptop > Thunderbolt Display > ARC-8050

1614 | 6051

Test 5 - Laptop > Thunderbolt Display > Thunderbolt Display 2

523 | 2031

Test 6 - Laptop > Thunderbolt Display > Thunderbolt Display 2 > ARC-8050

1809 | 6010

Test 7 - Laptop > Thunderbolt Display > ARC-8050 > Thunderbolt Display 2

1471 | 4625

Test 8 - Laptop > Thunderbolt Display > Thunderbolt Display 2 > ARC-8050 > Matrox MXO2

901 | 3236


So it seems to be pretty clear that overall system performance is affected by the inclusion of Thunderbolt devices. That said, Thunderbolt Displays seem to affect performance considerably more than my RAID enclosure did - but determining the exact effect they have is a bit more elusive (if not just down right bonkers). The absolute worse score came from Test 5, where I only had the two TB displays hooked up to my laptop. In that configuration, my 2011 i7 MBP scored more like a PowerMac G5 circa 2005 according to GeekBench's online benchmarks. The crazy thing is, I actually saw a 3-fold increase in performance by simply adding the ARC-8050 to that configuration (2031 to 6010)!


I don't know what to make of all of this data, other than that this machine can't handle all these Thunderbolt devices - a fact that unfortunately seems to remain regardless of the chain order.


Thanks again for your help, Malcolm! Anyone else experiencing these huge performance hits with their TB displays?

Aug 27, 2013 11:55 PM in response to spence3eb

Unfortunately, if I connect to the ARC-8050 RAID enclosure before daisy chaining to the monitors, my setup doesn't work. The enclosure itself powers down when I close the lid on my laptop and the screens behind it on the chain do not remain powered, so it seems I can't make the Thunderbolt Displays my primary displays without one of them being the first in the chain.

Are you sure the laptop was connected to external power when you did this test? If the power connector had come off things would not run with the lid closed.

Aug 28, 2013 12:27 AM in response to Malcolm J. Rayfield

Ah, what a dummy! You're exactly right, I've been using the Thunderbolt Display to power my laptop, and I couldn't connect it's data cable to the RAID while still using the connected MagSafe on my MBP, so I tried it on battery power.


I'll pull out my separate power cord and test that configuration tomorrow morning. Judging from all my other tests, I think its unlikely it'll reduce the overall system performance degradation, but it could affect the disk speed of the RAID. We'll find out!


Thanks, Malcolm!

Aug 28, 2013 12:01 PM in response to Malcolm J. Rayfield

The good news is you were right, as long as I had external power connected, putting the ARC-8050 first did work - though the monitors only power up after the RAID has done its 30 second boot up process - which is inconvenient but not a deal breaker.


The bad news is that unfortunately, I'm not seeing any benefits from the RAID being first in line.


In fact, with the RAID placed before the TB displays in the food chain, I'm now experiencing the relatively common crackling distorted audio I've read so much about. All I have to do is try and use either display's built in speakers while stressing the RAID drive with read/writes and the audio goes terribly awry. As far as I can tell, this audio degredation is merely an obvious manifestation of the dwindling throughput of TB. The only difference is that when I have the RAID at the far end of the chain, the problem manifests itself in drastically reduced read/write speeds rather than something as blatantly noticeable as distorted audio playback.


------------------------------------------


Disk Speed Comparisons:


ORIGINAL SINGLE THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 2

Chain order: MacBook Pro > TB Display 1 > ARC-8050 Enclosure

Results: 599 MB/s Write & 624 MB/s Read


NEW SINGLE THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 2

Chain order: MacBook Pro > ARC-8050 Enclosure > TB Display 1

Results: 603 MB/s Write & 465 MB/s Read



RAID LAST vs RAID FIRST

599 MB/s vs 603 MB/s - Pretty much the same read performance

624 MB/s vs 465 MB/s - Fairly noticeable drop in write performance


So the read performance doesn't seem drastically affected by the chain order, but write performance is noticeably worse when the RAID is first in line.


------------------------------------------


ORIGINAL DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain order: MacBook Pro > TB Display 1 > TB Display 2 > ARC-8050 Enclosure

Results: 239 MB/s Write & 299 MB/s Read


SECOND DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain order: MacBook Pro > TB Display 1 > ARC-8050 Enclosure > TB Display 2

Results: 237 MB/s Write & 364 MB/s Read


NEW DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain order: MacBook Pro > ARC-8050 Enclosure > TB Display 1 > TB Display 2

Results: 238 MB/s Write & 330 MB/s Read



RAID LAST vs RAID BETWEEN vs RAID FIRST

239 MB/s vs 237 MB/s vs 238 MB/s - Pretty much the same read performance across the board

299 MB/s vs 364 MB/s vs 330 MB/s - Marginal boost in write speeds with the RAID between the monitors


Here again, the read performance doesn't seem drastically affected by the chain order, but write performance is better when placed first in line, and best when the RAID is between monitors. This doesn't fall in line with degrading write performance of the single monitor test above.


------------------------------------------


Geekbench Comparisons:

KEY: (Single-Core Score | Multi-Core Score) (All 64-bit Benchmarks)

ORIGINAL SINGLE THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 2

Chain order: MacBook Pro > TB Display 1 > ARC-8050 Enclosure

Results: 1614 | 6051


NEW SINGLE THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 2

Chain order: MacBook Pro > ARC-8050 Enclosure > TB Display 1

Results: 2714 | 8788



RAID LAST vs RAID FIRST

1614 vs 2714 - Noticeable increase when RAID is first in line

6051 vs 8788 - Noticeable increase when RAID is first in line


Whereas I noticed some degredation in write speeds when the RAID was placed first in line, there was actually a noticable increase in system performance in that configuration.


------------------------------------------


ORIGINAL DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain order: MacBook Pro > TB Display 1 > TB Display 2 > ARC-8050 Enclosure

Results: 1809 | 6010


SECOND DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain order: MacBook Pro > TB Display 1 > ARC-8050 Enclosure > TB Display 2

Results: 1471 | 4625


NEW DUAL THUNDERBOLT DISPLAY TEST SETUP:

Total Thunderbolt Devices: 3

Chain order: MacBook Pro > ARC-8050 Enclosure > TB Display 1 > TB Display 2

Results: 1779 | 5550



RAID LAST vs RAID BETWEEN vs RAID FIRST

1809 vs 1471 vs 1779 - Single-Core performance drops when RAID is between monitors, otherwise scores are similar.

6010 vs 4625 vs 5550 - Multi-Core performance drops when RAID is between monitors, otherwise scores are similar, if not a little worse when the RAID is first in line.


In direct dischord with the RAID write speed test (where the RAID performed best when between monitors), there was a noticable decrease in system performance in that same configuration.


------------------------------------------


Here's an overlay of all the comparisons, where we can start to see a trend showing itself:


RAID LAST vs RAID FIRST

599 MB/s vs 603 MB/s - Pretty much the same read performance

624 MB/s vs 465 MB/s - Fairly noticeable drop in write performance when RAID is first in line

1614 vs 2714 - Noticeable increase when RAID is first in line

6051 vs 8788 - Noticeable increase when RAID is first in line


RAID LAST vs RAID BETWEEN vs RAID FIRST

239 MB/s vs 237 MB/s vs 238 MB/s - Pretty much the same read performance across the board

299 MB/s vs 364 MB/s vs 330 MB/s - Marginal boost in write speeds with the RAID between the monitors

1809 vs 1471 vs 1779 - Single-Core performance drops when RAID is between monitors, otherwise scores are similar.

6010 vs 4625 vs 5550 - Multi-Core performance drops when RAID is between monitors, otherwise scores are similar, if not a little worse when the RAID is first in line.


It seems to me that the numbers are bearing out an interesting fact: the performance of my RAID drive seems to be inversly proportional to the performance of my CPU/overall system. When my RAID is fastest, the system is slowest, and when the system is fastest, the RAID isn't performing well.

Aug 28, 2013 9:13 PM in response to spence3eb

Looks like you have reached the bandwidth limit of your MBP's chipset. Yours is one of the earlier Thunderbolt versions, an in laptops, they have to make tradeoffs between performance and power consumption. You may get better results with the new Retina MacBook pros, with newer chipsets and two Thunderbolt ports. Even better, a 27" iMac. It has two Thunderbot ports, and since it include one 27" display, you would need on;ly one Thunderbolt display

Dual Thunderbolt Displays - Big Performance Hit on RAID and CPU

Welcome to Apple Support Community
A forum where Apple customers help each other with their products. Get started with your Apple ID.