NVMe with ST-NGFF2013-C; Vega Internal GPU; Mac Pro 2013 (6,1)

[adonis3k]

I bought another OEM drive incase of this issue as I had the older rom versions which would not upgrade with the OS.

My 960 with adapter in title...is on the below to after updating to OS 10.13.5 (17F77)

Boot ROM Version: MP61.0123.B00
SMC Version (system): 2.20f18
Illumination Version: 1.4a6

 

[CodeJingle]

if i get the mac pro booting with vega that is not a full featured test. the resolution is on the low end but still htc vive gen 1 is a good follow-up test once vega boots.

it was lame i had to call to reattempt delivery and it took 90 minutes more for them to come. turns out they went to the wrong address initially. too big to ship to amazon locker.





[doublepost=1527967053][/doublepost]getting nvme booting on the gen 3 pcie instead of gen 2 is also a high priority for me to at least try. and amazingly enough is relevant to the thread!

[doublepost=1527968246][/doublepost]picked up thermal potting compound from amazon locker.





update: next day (sunday 11am). picked up cpu cooler (heatsink) from amazon locker. lol eye roll the self-branded screwdriver.

 

[crhendo]

I bought another OEM drive incase of this issue as I had the older rom versions which would not upgrade with the OS.

My 960 with adapter in title...is on the below to after updating to OS 10.13.5 (17F77)

Boot ROM Version: MP61.0123.B00
SMC Version (system): 2.20f18
Illumination Version: 1.4a6

I also kept my old Apple drive just in case this happened.

Thank you all for making the effort to come back to me on what I hope WON'T be an ongoing perplexing issue.

 

[CodeJingle]

trying out adobe indesign for tile-printing large images. seems to work better than numbers for this purpose but numbers is free so no clear winner here.

 

[CodeJingle]

I am potting the heatsink to the cpu now. then i can test 3rd party internal graphics hardware.

 

[CodeJingle]

because thermal potting epoxy takes a week to dry at room temperature you have to heat it up. but that is a problem. the epoxy is between heat sink and cpu. heatsink's job is to pull heat away. time for tryhard.

 

[CodeJingle]

heatsink attached to front of cpu riser card. attaching duplicate heatsink on back for ballast and leverage.

 

[CodeJingle]

both heat sinks are mounted i had to run so no time to test. i checked before i started. after 4 hours of baking good idea to check again. i'll continue progress tonight.

 

[CodeJingle]

the computer was working with the heatsinks but i got a little careless and shorted the circuit board for wifi/bluetooth/fan with the ports emi shield on the io board. now the 11v and 12v rails are both gone but i am not clear if it is because of component failure on the power supply board or if i fried some portion of the interlock circuitry on the io board and now with a broken interlock the rails refuse to turn on. placing the magnet definitely no longer triggers the relay(s) on the power supply. i would be risking death modifying the ac-dc power supply circuitry directly with all of those huge caps no thank you.

it was always the plan for the first prototype to use a generic pc power supply not the original. so if i did fry the mac pro power supply it isn't a big deal as long as i can figure out the work around (though figuring out the workaround is easier with a working original). as long as the cpu riser card and cpu are undamaged.

for now i have gotten back the 12v rails with a windows pc power supply. but 11v rails is non-standard not sure the best way to do that yet.

ac-dc adapters are usually not that precise going 11->12 or 12->11 either direction less than 10% change. if i start with 12 can just use an LDO rated for high capacity to reduce the voltage by less than 10%.

pain in my lumping buttocks that apple has a custom 11v rail alongside a 12v rail. i bet for now i can back-power the 11v rail with 12v and it will work fine. i may have to delayer the power supply to know what the pinout is for the 11v rails.

at this point it is still worth it to replace the boards i fried because everything was working fine before stupidity intervened. i hadn't yet tried a 3rd party gpu.

 

[RyanXM]

The latest firmware for the MacPro6,1 is MP61.0124.B00 as seen in the 10.13.5 Update.

 

[Flint Ironstag]

Epic thread. I have nothing technical to add

 

[MIKX]

Your pictures are terrifying.

. .. but . . .. you must have learned so much more than Apple allows us to know.

'twould probably make a good book for real Mac enthusiasts.

I wish you success.

 

[CodeJingle]

beginning of documentation for input and output pins for mac pro 2013 original power supply. the ones i know could be wrong. and some i don't know.

12v rails (bus bar three pairs)


12v rails (5 line pairs)


i have no idea about this connector or its six pins. noteworthy that only two of the six pins are twisted pair, the rest are coming in straight.


i have no idea what this lone line is for.

 

[CodeJingle]

this is bullocks. not enough bloody sanity left in the budget for depopulating and delayering another board. if it wasn't for me buggering up and causing this mess myself.

 

[CodeJingle]

finished depopulating next delayering

 

[CodeJingle]

i never thought a scan of a delayered board would look like a gas cloud nebula (which coalesce slowly into new suns and solar systems).

literally beautiful. i bet even non-engineers will find beauty in it. also, another win for fractals.

full-resolution version
https://www.dropbox.com/s/rfme0ryrji9668m/img_logic036_full.jpg?dl=0



lol so proud i hung it up on the fridge at work


coworker didn't know what it was says it reminded him of Mars. it does have a planetness to it.

 

[CodeJingle]

lost a few traces but that is what happens when you delayer another board with lack of sleep and having just finished a solid week of already working through boards. hopefully enough information here for what i need to know. i'll take a look tonight.

 

[CodeJingle]

For all the pins coming off the power supply only seven are unknown. Here they are, the seven unknown, mapped to the trace.

update: the board may have several inner layers. i may have to go through all the layers on this board as well. as-is with just the top layers documented i don't have enough information to tell what these 7 pins are doing. no pair pass the continuity test so each of the 7 pins serve a distinct purpose. like i said if i wanted to replace the function of these 7 pins with a standard power supply it would have been more effective to do so while the custom power supply was still working and i would have hooked up the seven pins to an oscilloscope or a logic analyzer to more easily detect their purpose.

 

[CodeJingle]

looks like for now i'll have to wait till i can replace the custom power supply to continue progress. once i have the replacement power supply i can hook up those seven pins to a logic probe and document their function, then afterward i can finish replacing the custom power supply with a standard power supply. i'll also know once i replace the power supply if anything else was damaged.

 

[CodeJingle]

i am blocked on a broken machine and have to replace parts until the machine works again (there were visible sparks between the io board, wifi+bluetooth board, and the board for the exhaust fan assembly, and obviously all those sparks were supplied by the power supply, so that is at least 4 parts involved). i could replace all four parts at the same time for about 500 USD i’m not sure i want to do that, i might just start with the power supply.

but i can still make progress. i have three more signal layers of the logic board that i have printed out and need to stitch together. i have the second set of signal layers of the cpu riser card to finish uncovering and document before passing through to the third set of signal layers behind the second set of ground planes (then print out and stitch together). i have more layers to uncover and document for the io board, mac gpu, custom power supply, and pc motherboard, pc gpu.

once i am booting again, logic probe on the 7 unknown pins of the power supply, and logic probe on the unknown pins of the gpu flex cables.

 

[CodeJingle]

three more logic board signal layer posters. i've printed four total. there are six signal layers to print for this board but i've run out of ink on one printer and the other printer is broken.

once i finish taping these up i'll rebuild the missing traces. doing it by hand this way is my best option right now.

 

[CodeJingle]

video update. my feedback is not necessarily validated, some of it may be incorrect, take it with some salt.

https://www.dropbox.com/s/qzfueabaaddy1ln/20180609_223204.mp4?dl=0

 

[CodeJingle]

color printer still broken. using a new backup non-color printer, settled for grayscale printout of top two signal layers of logic board.

 

[CodeJingle]

the same two layers are now slightly more documented (after three more hours with the cpu riser card)

 

[CodeJingle]

i bought io panel, io board, two power supplies (just in case i blow it again later), wifi bt card, interposer card, exhaust fan, and fan cable. i also got a broken apple ssd for delayering. everything should arrive next week then i'll test if the machine boots up.