testing TB3 AIC with MP 5,1

Discussion in 'Mac Pro' started by Flint Ironstag, Sep 25, 2018.

  1. Flint Ironstag, Sep 25, 2018
    Last edited: Sep 26, 2018

    Flint Ironstag macrumors 6502a

    Flint Ironstag

    Joined:
    Dec 1, 2013
    Location:
    Houston, TX USA
    #1
    The good folks at egpu.io have had some encouraging results with the latest Gigabyte TB3 add-in card and will be testing in MP 5,1. Will be very interesting to see the results.

    Hat tip to @itsage.

    [edit] to correct link
     
  2. handheldgames, Sep 25, 2018
    Last edited: Sep 25, 2018

    handheldgames macrumors 65816

    handheldgames

    Joined:
    Apr 4, 2009
    Location:
    Pacific NW, USA
    #2
    Thanks for sharing..

    This looks promising enough to almost order one from Amazon. I wouldn't mind converting one of the PCI 3.0 M.2 ports on the Highpoint SSD7101-a to an x4 PCIe 3.0 slot for the GC-titan ridge thunderbolt controller.

    Coming up with 2 additional 6-pin power connections is a challenge to make this work.
     
  3. joevt macrumors regular

    Joined:
    Jun 21, 2012
    #3
    The card being tested is the GC-TITAN RIDGE, not the GC-ALPINE RIDGE.

    I've tried the GC-TITAN RIDGE in slot 2 of a Mac Pro (Early 2008) MacPro3,1 running Windows 10. Read speed from a Samsung 960 Pro 1TB in an OWC Mercury Helios 3 was only 706 MB/s meaning the slot was probably running at PCIe gen 1 speed instead of gen 2. Do you have a link or instructions for installing pciutils for Windows? I could use that to try and force gen 2 speed. The Amfeltec Squid Gen3 PCIe Carrier Board for 4 M.2 SSDs is similar to the Highpoint SSD7101-a. I could try getting gen 3 speed with that.

    The 6 pin power connections are not required unless you plan on charging a laptop from the Thunderbolt ports. The USB 2.0 connection is not required unless you want to connect USB 2.0 devices to the Thunderbolt ports (such as in the LG UltraFine 4K). The Thunderbolt AIC header connection is not required (maybe it's for hot plug support which I don't think works in my case).
     
  4. XNorth macrumors regular

    XNorth

    Joined:
    Feb 23, 2018
    Location:
    United States
  5. Flint Ironstag thread starter macrumors 6502a

    Flint Ironstag

    Joined:
    Dec 1, 2013
    Location:
    Houston, TX USA
    #5
    @joevt, thanks for the correction - I linked the incorrect product. Editing the main post. I'm not familiar with pciutils. Switched to HP Z series, but may Hackintosh one of them.
     
  6. Pressure macrumors 68040

    Pressure

    Joined:
    May 30, 2006
    Location:
    Denmark
    #6
    I can't see how you could route a display through this solution? I only know of the SUNIX UPD2018 that allow you to pass through a signal from your graphic card to an USB-C/Thunderbolt monitor, like the LG Ultrafine 4K and 5K.
     
  7. joevt macrumors regular

    Joined:
    Jun 21, 2012
    #7
    pciutils is discussed at #206 and later. It can be used to get information about a PCIe card, including the current link speed and width, using the PCIe registers. In some cases, a 8.0 GT/s card may have it's speed incorrectly set to 2.5 GT/s in a 5.0 GT/s slot and some pciutils commands can adjust the speed back up to 5.0 GT/s. This problem exists with the amfeltec and GC-TITAN RIDGE when used in the MacPro3,1. Newer Mac Pros get a new firmware update with Mojave that may fix this.

    If you go to the support page for the GC-TITAN RIDGE which was linked in the first post, you can see there are two DisplayPort 1.4 inputs.

    The Sunix UPD2018 has only one DisplayPort 1.2 input and can output only USB-C with DisplayPort alt mode (like used by the LG UltraFine 4K). It cannot output two DisplayPort signals plus PCIe over Thunderbolt (like used by the LG UltraFine 5K).

    With two DisplayPort 1.4 inputs, the GC-TITAN RIDGE can support two USB-C displays such as the LG UltraFine 4K (but you should also connect the USB 2.0 header to support audio, brightness control, and USB ports of the display). You can transmit two 4K DisplayPort signals using a single Thunderbolt 3 port, or one display (up to 8K 30 Hz) per Thunderbolt 3 port.

    The GC-TITAN RIDGE also supports Thunderbolt displays.The LG UltraFine 5K uses two DisplayPort 1.2 signals. Lower resolution Thunderbolt displays only use one DisplayPort signal. Newer Thunderbolt displays (don't exist yet) may use DisplayPort 1.4 which can do 5K with one DisplayPort signal so you could connect a second to the other Thunderbolt port.
     
  8. Pressure macrumors 68040

    Pressure

    Joined:
    May 30, 2006
    Location:
    Denmark
    #8
    Ah great, wasn't sure it had inputs, which makes totally sense. Thanks.
     
  9. joevt macrumors regular

    Joined:
    Jun 21, 2012
    #9
    I've redone my Samsung 960 Pro 1TB test (connected to a MacPro3,1 via Thunderbolt 3 using a GC-TITAN RIDGE). This time, after getting the connection to work in Windows (may take up to two reboots - first is to detect the Thunderbolt device, then second is to enumerate it), I warm boot into macOS Mojave (warm boot keeps the previous enumeration I guess).

    At 2.5 GT/s, write is 724 MB/s and read is 742 MB/s. After using pciutils to renegotiate the PCIe link speed up to 5.0 GT/s, write speed goes up to 889 MB/s and Read is 1490 MB/s (using AJA System Test).

    To obtain full Thunderbolt 3 speed (up to 2750 MB/s but probably less than that for the old MacPro3,1), I supposed it would be possible to connect the GC-TITAN RIDGE to an Amfeltec or Highpoint via M.2 to PCIe 3.0 x4 adapter. The PEX converts the fast but narrow PCIe 3.0 x4 of the GC-TITAN RIDGE to the slow but wide PCIe 2.0 x16 of the Mac Pro.
     
  10. kings79 macrumors regular

    kings79

    Joined:
    Sep 16, 2015
    #10
    Thunderbolt on the cMP?
     
  11. joevt macrumors regular

    Joined:
    Jun 21, 2012
    #11
    Yes. I'll try getting > 2000 MB/s later this week. Hotplug, sleep/wake probably don't work (except USB 3.1 gen 2 seems to work with hotplug). And it's annoying to have to boot into Windows to get it working after a cold boot because I have to use BIOS mode for Windows on my MacPro3,1 which means I can't have other hard drives connected in the PCIe slots (such as OWC Mercury Accelsior or Sonnet Tempo SSD Pro Plus). I moved my SSDs from the Sonnet Tempo to the four drive bays of my Mac Pro using two SATA 22 Pin Male to SATA 7 Pin and 2 X 15 Pin SATA Female Connectors - 6". I used a small piece of adhesive on one end of each SSD to hold them up. The SATA cable gives me access to SATA power (includes 3.3V which is not included by molex power from the DVD drive bay).
     
  12. kings79 macrumors regular

    kings79

    Joined:
    Sep 16, 2015
    #12
    Wow, interesting and impossibility thought post. Keen to follow your findings! And if at all possible to get around the warm boot problem. Cause that would suck to have to do each time.
     
  13. equals macrumors newbie

    equals

    Joined:
    Jun 24, 2018
    #13
    Thanks for testing and for providing such encouraging information.
    As for booting into Windows for controller detection and initialization, I guess we have to figure out either how to keep it initialised once and forever (a couple of years back some Hackintosh folks with Alpine Ridge controllers told that they had to boot into Windows only once, and controller kept initialized even on cold boots) or how to initialize it on every boot without Windows. The second option looks pretty interesting to me, since the latest MacBook Pro has a Titan Ridge controller on board. Maybe (just maybe) it is reasonable to extract its DXE drivers, inject them into a cMP firmware and give it a try. I checked the MBP 2017 firmware a couple of minutes ago, and there are three DXE drivers mentioning Thunderbolt: ThunderboltNhi, ThunderboltXDomainDevice and PciThunderbolt. Of course, since MBP 2017 uses a different controller, these particular drivers should be useless in our case, but I guess we could find something pretty similar inside the MBP 2018 firmware.
     
  14. tsialex macrumors 68020

    tsialex

    Joined:
    Jun 13, 2016
    Location:
    Brazil
    #14
    Don't inject anything from recent Macs into MP5,1 firmware. I injected the rMBP'18 NVMe EFI module into the BootROM and bricked my MP5,1. Apple is compiling EFI modules with optimizations and instructions that do not work in a MP51. If you want to check, use rEFInd.
     
  15. LightBulbFun macrumors 68000

    LightBulbFun

    Joined:
    Nov 17, 2013
    Location:
    London UK
    #15
    further adding onto this post

    by check,

    we mean first extract the DXE drivers as .efi files and load em from an EFI shell, this way you can test if they will lock up or mess with the system in anyway before you make a change to your BR and brick something. (although I dont know how something like thunderbolt might react to being activated after POST like that)
     
  16. equals macrumors newbie

    equals

    Joined:
    Jun 24, 2018
    #16
    It was just a quick assumption about a possible path, not a detailed action plan I'm going to follow off the bat :)
    Anyway, thanks for your kind advice.
    At the moment I have neither GC-Alpine Ridge nor MBP 2018 BootROM ready at hand, but I'm going to order an AIC for testing in... a few minutes I suppose. Unfortunately, since Gigabyte doesn't import these boards to my country yet, I'll have to wait a bit for the order to be delivered internationally.
     
  17. equals macrumors newbie

    equals

    Joined:
    Jun 24, 2018
    #17
    Okay, my GC-Titan Ridge card is on its way from Amazon to my intermediary, from where it will be sent to me par avion. It usually takes about 10-14 days total.

    BTW, can anybody give me a hint where in the world I can find a rMBP'18 BootROM except dumping it from an actual machine which I don't have? I've inspected the 10.13.6 Supplemental Update for MacBook Pro with Touch Bar (2018), and its follower Supplemental Update 2, but I could find all supported models firmware updates inside, excluding exactly any signs of rMBP'18 firmware. I suppose there are only software and SMC/T2 patches. Or I'm just looking in the wrong place.
     
  18. tsialex macrumors 68020

    tsialex

    Joined:
    Jun 13, 2016
    Location:
    Brazil
    #18
    Open the Seed SUCatalog, find the 2018-10-02 FirmwareUpdate.pkg. Extract it, all firmwares are inside Tools folders.
    --- Post Merged, Oct 3, 2018 ---
    Screen Shot 2018-10-03 at 10.30.01.png
     
  19. equals macrumors newbie

    equals

    Joined:
    Jun 24, 2018
    #19
    Well, just like in case with Supplemental Updates (from where I took the FirmwareUpdate.pkg earlier), the 'latest' MBP firmware I was able to find inside is intended for rMBP'17 (MBP143...).
    1014Seed-2018-10-02-FirmwareUpdate-01.png
    1014Seed-2018-10-02-FirmwareUpdate-02.png
    The same with the 10.13 Seed SUCatalog.
    Maybe I'm just a little blind because of working on a few projects at the same time.
     
  20. LightBulbFun macrumors 68000

    LightBulbFun

    Joined:
    Nov 17, 2013
    Location:
    London UK
    #20
    you wont find the BootROMs for T2 based macs in there

    the only place I know where you can find em is inside the IPSWs used to restore the T2 chip (they end in .im4p but they are .fd files for the most part)

    upload_2018-10-3_16-24-47.png

    upload_2018-10-3_16-25-50.png
     
  21. equals macrumors newbie

    equals

    Joined:
    Jun 24, 2018
    #21
    Thank you, found it!
    The same driver names as I supposed. So, now we have to test it with GC-Titan Ridge and rEFInd or Next Loader on cMP I guess. I understand that it's not so likely that it will work so straightforward, considering what @tsialex said about the new optimizations and instructions, but we must at least try. If no one has time and/or hardware to test, I'll do it when my AIC arrives. Снимок экрана 2018-10-03 в 18.53.08.png
     
  22. tsialex macrumors 68020

    tsialex

    Joined:
    Jun 13, 2016
    Location:
    Brazil
    #22
    Recent Macs firmware are compiled by LLVM 9.x, older ones by 3.x, and Apple is using new instructions too.
     
  23. equals macrumors newbie

    equals

    Joined:
    Jun 24, 2018
    #23
    Okay, further research gave more information.
    Some hackintosh GC-Titan Ridge users say that if they try to boot with 'GPIO3 Force Pwr' setting in their UEFI BIOS turned off, they also have to boot into Windows first and then warm reboot into macOS to make TB3 work under macOS. But if they turn this setting on, GC-Titan Ridge will work in macOS on cold boot. And 'GPIO3 Force Pwr' appears to be a setting which forces motherboard to set the TB GPIO header's pin 3 to high (supposedly 3.3v or 5v DC) thereby giving the controller a command to fully power on. Seems like Windows can detect and power it on using its own drivers. So, the second supposed workaround here could be just to apply DC to the pin 3 of the card's GPIO connector. It would be interesting to test some Titan Ridge compatible Gigabyte B360/H370 motherboard's TB header to figure out if there really is just a current and to measure its actual voltage.
    Of course, this might be a more complex problem, and then we'll have to resolve both the card power on issue and its initialization issue.
    As for hotplug, it appears to be the ACPI issue which can be fixed on hackintosh computers by DSDT edits, so it has to be done in any other special way I guess if even possible.
     
  24. kings79 macrumors regular

    kings79

    Joined:
    Sep 16, 2015
    #24
    So are you following a software based fix or a hardware?
     
  25. equals macrumors newbie

    equals

    Joined:
    Jun 24, 2018
    #25
    Here it is:
    I mean, it may happen that both fixes are needed. As I mentioned earlier, my card is on its way, and I'll dive into testing in about 12 days or so. In the meantime, I'm just looking for as much potentially useful information as I can find.
     

Share This Page