I've ordered a ST-NGFF2013-C so will see how I get on, just need to wait forever for it to come...Zzzzz
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
NVMe with ST-NGFF2013-C; Vega Internal GPU; Mac Pro 2013 (6,1)
- Thread starter CodeJingle
- Start date
- Sort by reaction score
First Mac Pro 2013 GPU pin mapped (in the video; haven't updated docs yet)! I was running out of time but still squeezed in to double-cross-reference the first Mac Pro CPU <=> Mac GPU <=> PC GPU pin combination (for PCIe lane pin, we already have some ground pins mapped)! In the video I just woke up after a nap but am still quite tired so don't mind the incomprehensibility.
https://www.dropbox.com/s/nmvy7rupk178asw/20180416_063316.mp4?dl=0
[doublepost=1523902265][/doublepost]Some screenshots from dwdrummer959
flex cable connector (GPU side); also called 'meg array' connector
which pins of flex cable connect to ground (GPU side); also called 'meg array'
W7000 die
W7000 pads
[doublepost=1523903751][/doublepost]
I started soldering leads on the GPU
https://www.dropbox.com/s/nmvy7rupk178asw/20180416_063316.mp4?dl=0
[doublepost=1523902265][/doublepost]Some screenshots from dwdrummer959
flex cable connector (GPU side); also called 'meg array' connector
which pins of flex cable connect to ground (GPU side); also called 'meg array'
W7000 die
W7000 pads
[doublepost=1523903751][/doublepost]
I started soldering leads on the GPU
Last edited:
Can anyone confirm on a Mac Pro 2013, if you uplug it from AC, press and hold the power button while disconnected from AC, keep the power button held while re-connecting AC, wait about 5 seconds after reconnecting AC (button still held), then release the power button, does their Mac Pro enter a special mode, including the fan spinning up to its max speed?
Update: reproduced by mikeboss
https://forums.macrumors.com/thread...ac-pro-2013-6-1.2085886/page-19#post-25987103
I found a pair of pins off the IO Wall connector that trigger this and I'm not sure if it is the actual power button or not. If it is then I've successfully mapped the power button I think.
Update: power button pins mapped, see this post
https://forums.macrumors.com/thread...ac-pro-2013-6-1.2085886/page-19#post-25987381
Update: reproduced by mikeboss
https://forums.macrumors.com/thread...ac-pro-2013-6-1.2085886/page-19#post-25987103
I found a pair of pins off the IO Wall connector that trigger this and I'm not sure if it is the actual power button or not. If it is then I've successfully mapped the power button I think.
Update: power button pins mapped, see this post
https://forums.macrumors.com/thread...ac-pro-2013-6-1.2085886/page-19#post-25987381
Last edited:
held powerbutton for a few seconds while connecting power: nothing happened. then pressed powerbutton again to turn on the machine and it started with fan running at 100%. shutdown, start: fan still at 100% 
EDIT:
disconnected mains power, pressed button. connected power and all seems to be back to normal.
EDIT:
disconnected mains power, pressed button. connected power and all seems to be back to normal.
No more Reboot issues, ZERO. So nice.
This is the taller, green adapter I was recommended.
Did you let it boot all the way? Did it boot as normal other than the fan at 100%? I don't see this mode documented in the service manual (really there is nothing useful in the service manual).held powerbutton for a few seconds while connecting power: nothing happened. then pressed powerbutton again to turn on the machine and it started with fan running at 100%. shutdown, start: fan still at 100%
EDIT:
disconnected mains power, pressed button. connected power and all seems to be back to normal.
I thought you only had to let go after 5 seconds and wait a moment and then it boots without having to press again, but that is nitpicking I guess. You did generally repro so now I know the pinout for the power button.
Pin 3 and pin 6 on the IO Wall connector act as the power button (connect the pins together as the equivalent of when the button is pushed down). Pretty sure. So I don't need an IO Wall anymore.
[doublepost=1523910331][/doublepost]
What was the model number or do you have a web link for purchasing?
Last edited:
@CodeJingle yes, I let it do a full boot twice, logged into the system. apart from the fan running at full blast, everything else seemed to work as usual.
At 3am with only a few hours of sleep the night prior I get a bit lazy. I may try a live stream instead of recording video footage, during further probing and documenting the mac pro 2013 gpu pinout. Sounds easier than recording and uploading. I previously heard Periscope was good for live streaming straight from a phone with minimal setup and teardown. But that recommendation is not recent. Anyone familiar with live streaming what app or service do you prefer? Something trending that people actually use.
[doublepost=1523990751][/doublepost]I prepped the rest of the wires and test probes for mounting to the S9000. It gets a bit painful after shoving 164 pogo pin test probes into receptacles. I labeled each pin. Though I'm not sure, the wires may be too long.
[doublepost=1523990751][/doublepost]I prepped the rest of the wires and test probes for mounting to the S9000. It gets a bit painful after shoving 164 pogo pin test probes into receptacles. I labeled each pin. Though I'm not sure, the wires may be too long.
Last edited:
I got the new* taller Sintech adapter in today and installed and had zero issues.
In detail what I had.
2013 Mac Pro
I had previously purchased the Sintech adapter and was sent the v2 black adapter.
Pairing this with a 500GB Samsung 960EVO.
Taped adapter/pins/practically entire adapter in two layers of tape.
The only issue I experienced was a restart caused [?] folder. Cold boot, no problem.
Tried specifically to find the green adapter, was sent another black, tired of the waiting on shipping game.
Came across the newer taller Sintech adapter on Amazon, 2-3day shipping for a few bucks more.
https://www.amazon.com/gp/product/B01CWWAENG/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1
I ordered the adapter through amazon on Sunday evening (4-15) received today at noon (4-17).
Taped the new adapter in the same fashion, I just layered two times completely around the adapter. Added the 960 EVO, installed, booted. Restarted.
Everything working perfectly.
1300+ Reads and Writes.
Remotely rebooting machine, since this one is becoming a headless system, worked perfectly.
Hoping this helps others get an adapter quicker and takes the guess work out of which one they may receive.
I ordered a 2nd one tonight just to hold on to for future.
As of when I ordered they were reporting 19 left in stock.
Good luck guys.
Next will be upgrading to a 1TB 960.
In detail what I had.
2013 Mac Pro
I had previously purchased the Sintech adapter and was sent the v2 black adapter.
Pairing this with a 500GB Samsung 960EVO.
Taped adapter/pins/practically entire adapter in two layers of tape.
The only issue I experienced was a restart caused [?] folder. Cold boot, no problem.
Tried specifically to find the green adapter, was sent another black, tired of the waiting on shipping game.
Came across the newer taller Sintech adapter on Amazon, 2-3day shipping for a few bucks more.
https://www.amazon.com/gp/product/B01CWWAENG/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1
I ordered the adapter through amazon on Sunday evening (4-15) received today at noon (4-17).
Taped the new adapter in the same fashion, I just layered two times completely around the adapter. Added the 960 EVO, installed, booted. Restarted.
Everything working perfectly.
1300+ Reads and Writes.
Remotely rebooting machine, since this one is becoming a headless system, worked perfectly.
Hoping this helps others get an adapter quicker and takes the guess work out of which one they may receive.
I ordered a 2nd one tonight just to hold on to for future.
As of when I ordered they were reporting 19 left in stock.
Good luck guys.
Next will be upgrading to a 1TB 960.
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
Just to confirm the new Sintech adaptor that is reportedly working is the ST-NGFF2013-C as per Sintech link below...
http://eshop.sintech.cn/ngff-m2-pcie-ssd-card-as-2013-2014-2015-macbook-ssd-p-1229.html
I ordered mine over a week ago, still waiting, in the meantime the other shorter one I ordered from eBay also had the cold boot success but flashing question mark on subsequent reboots. Will report how this longer ST-NGFF2013-C adapter goes in this 2013 nMP with the Samsung EVO 960 1TB NVMe SSD as soon as it arrives.
As an Amazon Associate, MacRumors earns a commission from qualifying purchases made through links in this post.
thanks for detailed feedback!
[doublepost=1524033579][/doublepost]
Yes so I've been told. I may get one just to be able to say I can personally vouch for it (the adapter I am using is no longer on the market).
Looks so promising! Hope you succeed! Great work so far. I really enjoy this thread!
Correct, this is their newer offering.
The pinouts seem to be along the lines of the original large green adapter.
The added length helps a bit with stability and holding the nvme firmly in place with little to no chance of 'wiggling" loose of breaking connection. More so for laptop users than us with desktops.
Last night from home, remoting in to this computer, I did probably 3 restarts and it came back up every time, no issues.
The only thing I have no tried is a CCC backup to an external where some others have been reporting failures with Samsung drives on the small black adapter. I'll check that out later today.
I can edit numbers documents using the web app at icloud.com which makes it easier since I don't have a Mac until I am done probing.
I shared these documents as read-only here just because it is easier for now while I only have a PC. It wouldn't let me upload until I upgraded my icloud plan kind of lame had to shell out $10 to be able to share these. Haven't updated them in a while (they still don't have cross references except the last two). I am probing tonight so I may update in the morning.
https://www.icloud.com/numbers/0wDiQsALfc4MvCQfsRsWp9f4g#LogicBoard1
https://www.icloud.com/numbers/0KRG7tliZ4NyNzl3FfJ1-jpXA#LogicBoard2
https://www.icloud.com/numbers/0uN128HcM1gojM1t1t8AgN38A#LogicBoard3
https://www.icloud.com/numbers/0BIMDqwFcFeTqu44T9TXuKBZA#cpuRiserCardFront
https://www.icloud.com/numbers/0QGBtViJ6E-CBGROgP_r6-r6A#cpuRiserCardBack
https://www.icloud.com/numbers/0gdC1u3FLFGEEg1LtozrR7hnQ#BlackAndBluePrintable
https://www.icloud.com/numbers/0QQRHBd3TI8QlGuL3iW412iPQ#XeonE5v2NoImageColorCoded
I shared these documents as read-only here just because it is easier for now while I only have a PC. It wouldn't let me upload until I upgraded my icloud plan kind of lame had to shell out $10 to be able to share these. Haven't updated them in a while (they still don't have cross references except the last two). I am probing tonight so I may update in the morning.
https://www.icloud.com/numbers/0wDiQsALfc4MvCQfsRsWp9f4g#LogicBoard1
https://www.icloud.com/numbers/0KRG7tliZ4NyNzl3FfJ1-jpXA#LogicBoard2
https://www.icloud.com/numbers/0uN128HcM1gojM1t1t8AgN38A#LogicBoard3
https://www.icloud.com/numbers/0BIMDqwFcFeTqu44T9TXuKBZA#cpuRiserCardFront
https://www.icloud.com/numbers/0QGBtViJ6E-CBGROgP_r6-r6A#cpuRiserCardBack
https://www.icloud.com/numbers/0gdC1u3FLFGEEg1LtozrR7hnQ#BlackAndBluePrintable
https://www.icloud.com/numbers/0QQRHBd3TI8QlGuL3iW412iPQ#XeonE5v2NoImageColorCoded
My one mapping in the previous video might be wrong. Another lol. Good to double check these things.
I confirmed dwdrummer959's ground pin mappings for the flex cable connector (he mapped the gpu side of the connector, I mapped the logic board side). I have most of the ground pins mapped now for all internal connectors.
Under certain conditions, perhaps under all conditions except fully energized and with interlock defeated, there is a physically-based switch, maybe electromechanical or magnetic-based relay in nature, that may affect continuity, to cutoff or short circuit the system otherwise. So there may be some ground or power pins whose mapping is missing, or ground and power may be swapped in some cases. It hopefully only affects the mapping of ground and power/vcc/vdd and not signal pins. FYI. Unless I want to physically permanently defeat the relay (safety interlock) by bypassing it electrically. I don't want to do that, at least not yet anyway. It possibly doesn't affect the mapping, I am saying it might.
Even though it sounds like more work to map another 324 pin connector, it is actually less effort overall to map the CPU Riser Card pins first rather than jump straight into mapping the gpu connector pins.
You can tell right away from looking at the top layer of the CPU Riser Card and Logic Board, even with the masking intact, that a bunch of pcie lanes are fed directly to one of the flex connectors (I think for the gpu without the hard drive. Which is good since that is the thing we are testing first).
I'll update the icloud shared documents tonight with the ground pin mappings and try to highlight the pcie traces coming in on the top layer. If I have time tonight I might also map the cpu pins for those pcie lanes.
I confirmed dwdrummer959's ground pin mappings for the flex cable connector (he mapped the gpu side of the connector, I mapped the logic board side). I have most of the ground pins mapped now for all internal connectors.
Under certain conditions, perhaps under all conditions except fully energized and with interlock defeated, there is a physically-based switch, maybe electromechanical or magnetic-based relay in nature, that may affect continuity, to cutoff or short circuit the system otherwise. So there may be some ground or power pins whose mapping is missing, or ground and power may be swapped in some cases. It hopefully only affects the mapping of ground and power/vcc/vdd and not signal pins. FYI. Unless I want to physically permanently defeat the relay (safety interlock) by bypassing it electrically. I don't want to do that, at least not yet anyway. It possibly doesn't affect the mapping, I am saying it might.
Even though it sounds like more work to map another 324 pin connector, it is actually less effort overall to map the CPU Riser Card pins first rather than jump straight into mapping the gpu connector pins.
You can tell right away from looking at the top layer of the CPU Riser Card and Logic Board, even with the masking intact, that a bunch of pcie lanes are fed directly to one of the flex connectors (I think for the gpu without the hard drive. Which is good since that is the thing we are testing first).
I'll update the icloud shared documents tonight with the ground pin mappings and try to highlight the pcie traces coming in on the top layer. If I have time tonight I might also map the cpu pins for those pcie lanes.
See how I overlaid the labeled cpu pads on top of the documented layer. In this case the pins are mapped by look alone. If I could post higher resolution images to macrumors the overlay would have legible text. That quarter of the cpu contains all of the pcie pins so that is definitely pcie signals being traced out to the socket pinout. 20 pairs.
[doublepost=1524189972][/doublepost]I added labels for the ground pins on the CPU Riser Card socket (on the front side)
https://www.icloud.com/numbers/0BIMDqwFcFeTqu44T9TXuKBZA#cpuRiserCardFront
[doublepost=1524189972][/doublepost]I added labels for the ground pins on the CPU Riser Card socket (on the front side)
https://www.icloud.com/numbers/0BIMDqwFcFeTqu44T9TXuKBZA#cpuRiserCardFront
I said in my last post the junk built-in apps that come with Windows can't handle the high resolution scans of the CPU Riser Card. I absolutely need a very high resolution scan to add a readable text overlay for the CPU pins. It is blocking me. I had everything I needed on the Mac Pro, now I have this junk Windows laptop until I finish the probing and quasi-permanently put the mac pro back together (the laptop is not junk because it is Windows though that sucks too, it is junk because it is under-powered). I'll see if I can get photoshop working on this laptop then I can continue progress. Hopefully there is enough memory otherwise even in photoshop it will start paging memory to disk and then it will be impossible to do the work. Though I don't know how long it will take I've been going full blast and need a short break.
I did buy a vega frontier card today which is nearly half of my disposable income for the year (especially now i just enrolled my kids in before-school daycare, with their mom moving soon i need my day-job schedule to start and end earlier in the day). So I have even more incentive to get this working on top of everything else.
I did buy a vega frontier card today which is nearly half of my disposable income for the year (especially now i just enrolled my kids in before-school daycare, with their mom moving soon i need my day-job schedule to start and end earlier in the day). So I have even more incentive to get this working on top of everything else.
Last edited:
If anyone accidentally breaks their iMac Pro screen and logic board when trying to do their own teardown or upgrade *cough* I would love for them to donate their totalled iMac Pro for the cause and I would publicly document the board layers and in general a more in-depth teardown of the hardware. I'm kind of completely broke at the moment but it is for a good cause. DM me anyone who has completely destroyed their iMac Pro and Apple refuses to fix it even if you paid the entire cost for a voided warranty repair because it would be cheaper to buy a new iMac Pro instead of repairing lol .
[doublepost=1524274803][/doublepost]also i just checked cpu requirement for VR and the min seems to be 3.3 ghz. so even after i finish probing i am limited to only boot testing until i either have custom brackets made for the PC heatsink or if i acquire the correct cpu i can permanently mount the heatsink using potting epoxy.
.[doublepost=1524274803][/doublepost]also i just checked cpu requirement for VR and the min seems to be 3.3 ghz. so even after i finish probing i am limited to only boot testing until i either have custom brackets made for the PC heatsink or if i acquire the correct cpu i can permanently mount the heatsink using potting epoxy.
@CodeJingle Please start a GoFundMe-Compaign.
I really enjoy watching your progress and would very much like to help you out with some money so you can continue.
I really enjoy watching your progress and would very much like to help you out with some money so you can continue.
Besides the NVMe stuff, I haven't really accomplished anything yet other than I can boot my Mac Pro 2013 after escaping its prison enclosure. It wouldn't be fair to start a campaign yet. If I get one external GPU working I can start an official campaign to get dual external GPU working which also requires mapping out the NVMe lanes. And then I can't test VR without a headset and a faster CPU. Some people have contacted me to help and ended up not helping so nobility is in spirit, in words and not in action. But also because what I am doing is legally questionable, such an official campaign might get shut down by Apple.
The community helping out with shared discovery of the NVMe stuff has been great but for some reason the GPU stuff seems one step removed and that is where it has been difficult getting any direct or indirect support except for one other person and the effort needs more support that two people to succeed. Hopefully I am able to cross the threshold where getting support becomes easier before I give up. I've spent about 2,500 USD on the project so far not counting the Mac Pro and the NVMe.
Last edited:
Looks like I'll be getting an influx of [donated] boards soon. I'll be starting to travel through a D300/D500/D700 board, which I'll need to do with as much care as I've been doing with the CPU Riser Card. Also, I totally botched the job I did for the 3rd layer of the I/O Board, which I knew was probably going to happen because I was using only sandpaper. It is bad that is the only recorded video footage I have of layer documentation. That means I should show you how to properly use abrasives to travel between layers, which involves a whole lot of fiberglass, which is both dangerous and disgusting. Also a lot of isopropyl alcohol which I should have used last time. Although that process can take up to 16 hours for a single layer, so it requires mild time lapsing. At least 2x. I got 180, 320, and 400 grit sandpaper yesterday, in addition to the 60, 80, 100, 120, 150, and 220 that I already had. Having the full set should come in handy. I'm running low on fiberglass which is going to be a problem with the total number of boards to sand down coming in at 4+ and each fiberglass refill is $15. I also should acquire another Logic Board and do a better job of sanding it down because I missed a whole bunch of pcie lanes in the deeper layers of the board. With what I have now I am probably only able to extract the pinout for one of the GPU connectors, not both.
Last edited:
Thanks! I can keep a running tab of donations and what I spend the money on, to make it a similar experience to crowdfunding. The thread itself and how I post progress is already very similar to how a crowdfunding campaign posts progress. Though [at least right now] what I offer isn't a physical good but rather organically discovered and somewhat unencumbered intellectual property, which is also difficult to get approval for in crowdfunding.
There may eventually be a physical good involved, some kind of adapter or something similar to eGPU box but for direct internal connection not pass-thru over Thunderbolt. Depending on circumstances also giving away the IP for schematic and layout for such a device and having others worry about making it is also possible but I need to be farther along and have a few conversations first before I can guarantee that. And it all hinges on any of this working at all which we don't know yet.
The Vega GPU I bought Friday off of eBay already has a tracking number and should be here Monday wow that was fast shipping.
Last edited: