USB 3.x PCIe Cards for Classic Mac Pro

Discussion in 'Mac Pro' started by ActionableMango, Dec 3, 2012.

Thread Status:
The first post in this thread is a WikiPost, and can be edited by anyone with the appropriate permissions.
  1. h9826790 macrumors G4

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #2001
    Using it right now, good all the way up to 10.13.4 17E202
     
  2. kohlson macrumors 68000

    Joined:
    Apr 23, 2010
    #2002
    I use this card on a 4/5.1. Works as expected on 10.13.4.
     
  3. Caesar_091 macrumors regular

    Caesar_091

    Joined:
    Jan 18, 2005
    Location:
    Italy
  4. pixelatedscraps macrumors 6502

    pixelatedscraps

    Joined:
    Jul 11, 2017
    Location:
    Hong Kong
    #2004
    I have this on my cMP 5,1 and also a 4,1-5,1 but I have to say without any scientific reasoning that it feels very slow in comparison to the USB 3.0 ports on my 2014 MBP, let alone my iMP. As soon as I get a bit of spare cash I’m going to pick up two of the Highpoint 1344 cards.
     
  5. flowrider macrumors 603

    flowrider

    Joined:
    Nov 23, 2012
    #2005
    ^^^^Yep, that's why I got rid of mine a couple of years ago. I'm running the HighPoint RocketU 1144C. Much Better!

    Lou
     
  6. jscipione, May 14, 2018
    Last edited: May 16, 2018

    jscipione macrumors member

    jscipione

    Joined:
    Mar 27, 2017
    #2006
    The KT4004 will not give maximum performance but I wouldnt waste your money on two 1344 cards what are you possibly doing that requires that kind of bandwidth? Unless you absolutely need 10gb/s over 5gb/s and why 2 of them? KT4004 should perform similarly to USB on the late 2013 Mac Pro since it uses the same FL1100 chipset.

    From: http://vrzone.com/articles/usb-3-0-speed-tests-7-way-host-controllers-roundup/13358.html/11

    "It's really quite hard to sum things up. We spent a week testing all the host controllers and we spent even more time sourcing the hardware we've used for this test, yet we've come out disappointed after weeks of hard work. Why you ask? Well, for one none of the host controllers reached the kind of performance numbers we expected, especially as the host controller makers have been showing benchmarks in excess of 300MB/s, reaching as high as 400MB/s in the case of Fresco Logic. We did our best to bypass system bottlenecks and we made sure we used the latest drivers available to us."

    It sounds like the Fresco Logic was the best of a bad bunch back in 2011 which is probably one reason why Apple chose FL for the Late 2013 Mac Pro (no integrated USB on Xeon chip, also FL and Mac Pro are both made in US so may have been a supply chain consideration). However none of the 7 controllers tested including the FL1009 and ASM1042 seem to match the Intel USB chipset used in your 2014 MBP. We are largely stuck in the same predicament today.

    My educated guess is that the High Point RocketU 1344A is using a pair of ASMedia ASM2142 chips which is a generation newer than the ASM1042 in the benchmarks above and is capable of using a x2 PCIe 3.0 link. However that still may not be as fast as the Intel chip in your 2014 MBP or the one in the in the iMP.
     
  7. pixelatedscraps, May 15, 2018
    Last edited: May 16, 2018

    pixelatedscraps macrumors 6502

    pixelatedscraps

    Joined:
    Jul 11, 2017
    Location:
    Hong Kong
    #2007
    I'm going for two 1344s (possibly the 2-port Sonnet Allegro USB 3.1) as I have 2x cMPs to fit out - both are currently running Inateck cards connected to OWC USB-C docks (the cMPs are slung under each desk and we need multiple ports next to the monitor, hence maximising bandwidth where possible: multiple hard drives, card readers, etc.) with very poor performance. I'm aware of the weaknesses of the OWC docks and how that could / most likely does impact transfer speeds but at this moment, we don't have a better solution, especially as we already have the docks ;)

    I actually only need a 2-port card as a lot of our backups are done over LAN so one port would be direct to the dock and the other would be used for an external hard drive or direct to a dedicated card reader.

    Update: That said, the OWC dock only connects to the cMP / supports USB 3.1 Gen I input and output but this dock is so underwhelming, I know I'll need to look at other, faster solutions if I'm to keep these cMPs performing for us. Perhaps ditching the docks altogether and getting the 4-port HighPoint cards with long USB cables are the way to go after all...
     
  8. thepawn, Jun 4, 2018
    Last edited: Jun 4, 2018

    thepawn macrumors 6502

    thepawn

    Joined:
    May 27, 2009
    #2008
    Doh, I feel sheepish, my fault! Nothing to see here. :)
     

    Attached Files:

  9. h9826790 macrumors G4

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #2009
    From what I can see, you are getting USB 3.0 connection correctly. Otherwise, you won't get that 112.7MB/s read speed (is that the USB 3.0 thumb drive?).

    For USB connected SSD. There is no TRIM available. It may affect the write speed, what's the read speed on that device?

    Also, how you measure the performance on the PC? Using different benchmarking software can make a huge difference.
     
  10. thepawn macrumors 6502

    thepawn

    Joined:
    May 27, 2009
    #2010
    Hey thanks, actually for an experienced tech pro I can hope I can only attribute this to lack of sleep, but the drive I was using I thought as an SSD was NOT an SSD but a 5400rpm. -_-

    Doh.

    I hooked up my SSD and it got 300+MB/s ... it's all good!
     
  11. thepawn macrumors 6502

    thepawn

    Joined:
    May 27, 2009
    #2011

    MyPassport SSD + Sonnet Pro USB3.0 card.
     

    Attached Files:

  12. binba macrumors newbie

    Joined:
    Jul 23, 2010
    #2012
    So for the latest in this 6-year, 2,011-chapter saga: What's the latest verdict on sleep? In High Sierra in particular...

    1. I dug through the thread, there seemed to be optimism around post #710, but things changed and 'died': the note on MacSales was revised from "sleep supported with driver" to "Sleep should be disabled when using devices with this card as you may otherwise experience volume ejections when waking from sleep. [..] NewerTech has reported a potential xHCI bug to Apple regarding sleep and we hope that Sleep support will exist in the future." And no word on custom drivers.

    2. Conversely, some people reported sleep working fine for them... and yet post #1 as well as Sonnet's site state that the problem is on the OS side and no card would support sleep.

    3. At first glance, needing external (Molex) power seems like bad implementation - and they did get rare - but it does seem like the most promising workaround for the sleep issues. Yet no one ever talked about it. So do any of the externally-powered cards fix the sleep issue, or not?

    4. I'm trying Mountain, which is like Jettison. My concern is that I'm using Premiere, and if the drive is offline the moment Premiere wakes up, it would freak out and decide all the media has gone offline, even if it comes back 2 seconds later. Will investigate, but I don't need extra corruption in my projects.
     
  13. h9826790, Jun 11, 2018
    Last edited: Jun 11, 2018

    h9826790 macrumors G4

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #2013
    I have an Orico PME-4U which need Molex power. I further replaced it by a KT4004 to free up a SATA port.

    I never really use sleep, but very occasionally will test it (e.g. after hardware change, or OS upgrade). From memory, when I use the PME-4U with the Orico 9558RU3 HDD enclosure. This setup doesn't suffer from any strange disconnection during the sleep test. (This test ran at about 2 years ago. So, most likely with El Capitan)
     
  14. crjackson2134 macrumors 68040

    crjackson2134

    Joined:
    Mar 6, 2013
    Location:
    Charlotte, NC
    #2014
    1. AFAIK thinks haven’t changed at all. Appearenty when entering the sleep mode, the USB drive MUST be ejected. Anytime this happens to be done for you (automatically by the OS) it generates a notification upon waking. It’s annoying, but I’ve never had, nor do I know of, a single instance of data loss as a result. I’m not saying it hasn’t happened to someone else, but I’m not aware of any such cases.

    2. Sonnet is correct I believe. It has something to do with the way OS X / macOS handles power management.

    3. I have no expierenc with Molex powered implemention. However, if that does solve the issue, it means that the drive never enters sleep at all. The whole problem (AFAIK) stems from the PCIe Buss power reduction when sleep is invoked.

    4. Mountain & Jettison work sporadically. I have them both and at times felt each worked great. Invariably however, each would fail to eject at random times. My only reliable solution was to simply eject the drive and click sleep when done.

    So, in short, nothing has really changed that I can see...
     
  15. firedownunder macrumors regular

    firedownunder

    Joined:
    May 5, 2011
    #2015
    FWIW, the Highpoint card in my sig in conjuction with Jettison works flawlessly (so far). El Cap if it matters.
     
  16. crjackson2134 macrumors 68040

    crjackson2134

    Joined:
    Mar 6, 2013
    Location:
    Charlotte, NC
    #2016
    Cool... I have the same card, and had reduced occurrences with Jettison/Cappy. What enclosure are you using?

    I've tried countless enclosures and they've all given the same results for me.
     
  17. firedownunder macrumors regular

    firedownunder

    Joined:
    May 5, 2011
    #2017
    Have Intel/Samsung ssds in a couple of OWC portable usb3 cases (use for backups/downloads/etc.). Improper eject notifications disappeared soon as I added Jettison. Have 2 Orico 5 drive bays (usb3/esata) for music/video but those are formatted zfs raidz2 boxes, they're exported and powered down when not in use.

    Tried the Inateck KT4004, wasn't seamless with El Cap. Never had the improper ejections issues when on Mavericks with a CalDigit Fasta-6GU3 card, but that card doesn't support El Cap+. The KT4004 also had issues recognizing all 5 drives in the Orico box. The Highpoint card solved those issues, Jettison took care of the sleep problem.
    --- Post Merged, Jun 11, 2018 ---
    I haven't loaded anything onto the 2nd Orico, when I get some time I'll throw a few drives in and do a test.
     
  18. jscipione macrumors member

    jscipione

    Joined:
    Mar 27, 2017
    #2018
    I have the Sonnet Pro which uses same FL1100 as KT4004 and I have improper eject errors when and USB disk or thumb drive plugged during sleep. In addition to this the Mac will wake from sleep periodically when USB drives are plugged in. Log says hidd is responsible. This does not seem to happen with hard drives plugged into the Macs USB2 ports. I wonder if Jettison would fix this problem... Also I wonder if the Highpoint card works any better.
     
  19. MIKX, Jun 11, 2018
    Last edited: Jun 11, 2018

    MIKX macrumors 6502a

    MIKX

    Joined:
    Dec 16, 2004
    Location:
    Japan
    #2019
  20. firedownunder macrumors regular

    firedownunder

    Joined:
    May 5, 2011
    #2020
    From Sonnets support page:
    817 [​IMG] USB storage connected to a PCIe USB 2.0, 3.0, 3.1 card (or combo card) gives an macOS error message upon wake from sleep. Dec-02-17
    To minimize power usage during sleep, power is removed from PCIe cards. Because of this power-saving specification, a USB 2.0, 3.0, or 3.1 PCIe card is unable to maintain USB port power during sleep. In macOS, this results in a storage device disconnect upon wake from sleep. The storage device will automatically remount, but the system reports a disconnect message. Because macOS flushes all caches before sleeping, this disconnect should never result in any loss of data.


    I like the Highpoint card because it specifies support for El Cap and later, unlike the Inateck card. Dedicated controllers for each port is a plus. That said, any PCE-e card will lose power when the machine is in sleep mode resulting in improper eject notifications. Jettison solves this and is configurable. And cheaper than replacing an otherwise working card.
     
  21. Kaenath macrumors newbie

    Joined:
    Oct 7, 2009
    #2021
    Has anyone had any issues with the Allegro Pro USB 3.0 PCIe card? I have one i my Mid 2010 Mac pro, and use it extensively with a pair of Drobo 5Ds.

    Recently, the 5D stopped mounting on this card. I can plug the fine into the USB 2 ports on the mac, but no-go in the sonnet. Other drives work completely fine with the card, just not the Drobo units themselves. any suggestions?
     
  22. h9826790 macrumors G4

    h9826790

    Joined:
    Apr 3, 2014
    Location:
    Hong Kong
    #2022
    Same situation with another cable or via a hub?
     
  23. crjackson2134, Jun 13, 2018
    Last edited: Jun 13, 2018

    crjackson2134 macrumors 68040

    crjackson2134

    Joined:
    Mar 6, 2013
    Location:
    Charlotte, NC
    #2023
    Not sure if you have the Pro version w/Independent controllers on each port, or the standard version with one controller for all 4 ports.

    I have the Pro version, and I've never had a problem with it at all, but I don't have any Drobos.
     
  24. ActionableMango thread starter macrumors G3

    ActionableMango

    Joined:
    Sep 21, 2010
    #2024
    Some troubleshooting is in order.

    First try different cable and different USB port on the Sonnet.

    Then check to see if 5D will show up on other computers. If not, the 5D is the problem.

    If it will, then will the 5D show up on your Mac Pro in a different OS installation? (Windows, Linux, or another MacOS drive?) If so, then all of your hardware is fine and there's a software issue.
     
  25. joevt macrumors regular

    Joined:
    Jun 21, 2012
    #2025
    Look it up on Wikipedia. USB 3.0 uses 8b/10b encoding (10 bits per byte) so 5 Gbps on the cable = 500 MB/s of data. That doesn't account for protocol overhead which will reduce that to something in the 400 MB/s range. So any benchmark that shows over 500 MB/s indicates a USB 3.1 gen 2 connection of 10 Gbps (which uses a more efficient 128b/130b encoding).

    Any ASM1142 card should be able to do 10 Gbps with the correct firmware and a PCIe 2.0 slot that supports x2 (except for cards that use the ASM1142 in PCIe 3.0x1 mode which requires a PCIe 3.0 slot for full performance). Some PCIe slot controllers do not support x2 and will then force a x2 PCIe device to use x1. Examples are: the PCIe 1.0 slots of the south bridge of the MacPro3,1 Mac Pro 2008, the PCIe 2.0 slots connected directly to a Sandy bridge CPU - 2nd Generation Core Processor Family Desktop CPU. Intel's cpu and chipset documentation will show the supported PCIe lane configurations.

    PCIe cards with more than one controller (two or four USB controllers or a mix of USB and SATA controllers will use a PCIe switch chip. The slot negotiates a link with the switch chip. The switch chip negotiates links with the USB and SATA controllers.

    If the switch chip has a 4 lane upstream port then it will eliminate the issue of x2 devices being negotiated down to x1. A PCIe switch can provide a benefit of converting fast/narrow links to equal speed slow/wide links (for example: the PCIe 2.0x2 of the ASM1142 to PCIe 1.0x4 of slot 3 or 4 of a Mac Pro 2008).

    PCIe slot bifurcation is a cheaper option than a PCIe switch but bifurcation doesn't have the link conversion benefits of a PCIe switch. Also, many slots don't support bifurcation (definitely there is no Mac that supports bifurcation).

    The picture of the Highpoint RocketU 1344a shows a PLX (PEX) switch chip. The two PCIe 3.0 x2 controllers are connected to that. The switch chip will be able to translate the PCIe 3.0 x2 links of the two ASM2142 controllers to PCIe 2.0 x4 so you should be able to get full performance from each controller separately (up to 1969 MB/s), but the max from using both controllers simultaneously will be PCIe 2.0 x4 in a Mac Pro's PCIe 2.0 slot (up to 2000 MB/s).

    Link speed is the speed per lane.
    PCIe 4.0 = 16 GT/s (128b/130b)
    PCIe 3.0 = 8 GT/s (128b/130b)
    PCIe 2.0 = 5 GT/s (8b/10b)
    PCIe 1.0 = 2.5 GT/s (8b/10b)

    A link may use multiple lanes. This is the link width (x1, x2, x4, x8, x16, x32).

    System Information.app does not show information about the bridge chip. For that you need to use ioreg or IORegistryExplorer.app to get device information, link speed and link width information. None of those will show the current actual link speed and link width. For that you need to use pciutils to examine the PCIe configuration space registers.

    The switch chip is a PCIe 3.0 switch chip. In a Mac Pro, such chips might negotiate a link speed of 2.5 GT/s instead of the expected 5.0 GT/s. PCIe 1.0 x4 is still fast enough to get full performance from a single USB 3.1 gen 2 port (10 Gbps). You can't notice a problem unless you try a USB raid of 3 or 4 SSD's (> 1000 MB/s).

    Read my posts in the "Amfeltec x16 PCIe with 4 SSDs: 5900+ MB/s" thread for information about PCIe 3.0 switch chips, the Mac Pro, and pciutils. #207
     
Thread Status:
The first post in this thread is a WikiPost, and can be edited by anyone with the appropriate permissions.

Share This Page