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
pro-tip if you end up with microscopic glass shards resting on your face, maybe even near your eyes, instead of touching it, blow it off with a dab from a can of compressed air. (because fiberglass shards if small enough can float in the air and come to rest on your face).
if you have industrial dryvac you can use that too, no attachment to the hose especially no brush, wipe off the end of the hose inside and out, and literally vacuum over your face, even over your eyes. try not to seal your skin with the hose keep it a short distance away and it will pick up microscopic fiberglass shards. be careful not to have the vacuum suck an eye out. i just vacuumed my clothes and my face hair etc, i am not joking. make sure the dryvac has a fresh bag and fresh hepa filter and everything is properly attached and sealed otherwise the glass goes right back into the air. when the bag is full don't reuse it please get rid of it.
Another 6 hours to bore through to the second layer on the front of the cpu riser card (following partial progress photo in previous post). I'm realizing for most purposes that scattered nicks on a ground plane is ok. Which means going from signal plane to ground plane I can half-ass it and it takes maybe 8 hours instead of 16 hours boring through from ground plane to signal plane (it takes that long for boring through layers with each layer being the surface area of the cpu riser card). Comparing the effort boring through layers on the cpu riser card versus the logic board, logic board layers being about half the surface area take about half the time.
if you have industrial dryvac you can use that too, no attachment to the hose especially no brush, wipe off the end of the hose inside and out, and literally vacuum over your face, even over your eyes. try not to seal your skin with the hose keep it a short distance away and it will pick up microscopic fiberglass shards. be careful not to have the vacuum suck an eye out. i just vacuumed my clothes and my face hair etc, i am not joking. make sure the dryvac has a fresh bag and fresh hepa filter and everything is properly attached and sealed otherwise the glass goes right back into the air. when the bag is full don't reuse it please get rid of it.
Another 6 hours to bore through to the second layer on the front of the cpu riser card (following partial progress photo in previous post). I'm realizing for most purposes that scattered nicks on a ground plane is ok. Which means going from signal plane to ground plane I can half-ass it and it takes maybe 8 hours instead of 16 hours boring through from ground plane to signal plane (it takes that long for boring through layers with each layer being the surface area of the cpu riser card). Comparing the effort boring through layers on the cpu riser card versus the logic board, logic board layers being about half the surface area take about half the time.
Last edited:
@CodeJingle
the 12-core CPU has arrived already, thank you. dropped it into my MacPro6,1 and ran the Intel Processor Diagnostic Tool (MS-Windows only): TEST PASSED!
the 12-core CPU has arrived already, thank you. dropped it into my MacPro6,1 and ran the Intel Processor Diagnostic Tool (MS-Windows only): TEST PASSED!
one good thing about the iMac Pro, is that quite possibly there are 16 unused lanes of PCIe, as pass-thru over the 4 thunderbolt ports. at least i am guessing that based on 48 pcie lanes total, with 16 lanes of one internal gpu, 4 for the M.2, 16 pass-thru to thunderbolt, and the various other peripherals can make due switching out the remaining 12 lanes.
This means we should be able to connect a 16x pcie slot to the imac pro, though it would use up all thunderbolt bandwidth, and requires some rights negotiation with the thunderbolt standard, and would require multiple simultaneous thunderbolt connections. To even be able to make and sell such an adapter would require membership in Apple's thunderbolt club (myFi?) and maybe the PCI-sig too. and when the product development is finished certification is required which is another fee. it is a possible future product i guess depending on how the current project turns out. at least the dream is alive if nothing else.
such a product isn't possible for the Mac Pro 2013 because there are not 16 lanes available for thunderbolt (because of the dual internal gpu uses up 32 lanes).
[doublepost=1526191242][/doublepost]It occurs to me that a version of the Mac Pro 2013 with dual cpu, or single cpu with hbm architecture, or single cpu with more ram slots, might only require swapping out the cpu riser card and not buying a whole new computer.
though it wouldn't work because of thermals, well and also the metal enclosure is hard to mod it basically requires welding.
This means we should be able to connect a 16x pcie slot to the imac pro, though it would use up all thunderbolt bandwidth, and requires some rights negotiation with the thunderbolt standard, and would require multiple simultaneous thunderbolt connections. To even be able to make and sell such an adapter would require membership in Apple's thunderbolt club (myFi?) and maybe the PCI-sig too. and when the product development is finished certification is required which is another fee. it is a possible future product i guess depending on how the current project turns out. at least the dream is alive if nothing else.
such a product isn't possible for the Mac Pro 2013 because there are not 16 lanes available for thunderbolt (because of the dual internal gpu uses up 32 lanes).
[doublepost=1526191242][/doublepost]It occurs to me that a version of the Mac Pro 2013 with dual cpu, or single cpu with hbm architecture, or single cpu with more ram slots, might only require swapping out the cpu riser card and not buying a whole new computer.
though it wouldn't work because of thermals, well and also the metal enclosure is hard to mod it basically requires welding.
Last edited:
top layers front and back W7000. 400 grit sandpaper has been spot on this weekend.
Last edited:
depends on which SSD your machine is equipped with. SSUAXs are slower than the 960 Pro. SSUBXs are more or less the same as a 960 Pro speedwise. but I doubt that anyone could tell the three apart from boot-times only.
the extra boot time is for cold boot only not reboot. even warm boot should be ok, so full shut down and boot from full shut down might still avoid the delay if the AC adapter stays connected the whole time. this is for a non-Apple drive using any adapter. the extra boot time is based on Apple boot process doing extra testing on the 'alien' drive because it is not expecting it (or something like that).
This is like the small cleaning robot mo from wall-e who freaks out with the 'foreign contaminant' alert. It forgets every cold boot.
Last edited:
Thank you! Doesn't sound so bad. Looking forward to seeing if there is a noticeable speed increase with an 960 Evo Pro.
Keep up the great work CodeJingle!
Keep up the great work CodeJingle!
I bought another liquid-damaged Mac Pro 2013 Logic Board at a good price so I can resurface the layers again and recover the information I previously lost. Flatbed scanning each layer will be awesome. This is the smallest of the boards so it doesn't take much effort.
Last edited:
Awseom tip on using nvme drives in this thread. I snagged a 1TB OEM PM961 and the long black sintech adapter mentioned here. Works great on my 6 core mac pro 2013.
Would like the verify that in Startup Manager (boot with Option key) the drive should display as an external (yellow icon). Also had some panic when setting up Bootcamp and Windows 10 as the Bootcamp partition shows up as "EFI Boot" rather than Bootcamp but it seems to boot just fine.
I "think" is normal when a drive is identified in the boot rom as external (by virture of it being on an adapter).
Can anyone confirm this is normal behavior?
Thanks
Would like the verify that in Startup Manager (boot with Option key) the drive should display as an external (yellow icon). Also had some panic when setting up Bootcamp and Windows 10 as the Bootcamp partition shows up as "EFI Boot" rather than Bootcamp but it seems to boot just fine.
I "think" is normal when a drive is identified in the boot rom as external (by virture of it being on an adapter).
Can anyone confirm this is normal behavior?
Thanks
Yes - Mine does the same. I’ve tried with 960pro and XG3 and no difference.
I may be wrong but I’m sure on an earlier version of High Sierra that it didn’t do this.
The adapter is passive it should have no effect on internal vs external. The drive should be showing up as internal. Though I can't double check at the moment, if anyone else wants to chime in here.
[doublepost=1526641809][/doublepost]Instead of getting a 2nd vega gpu or a vr headset, at least for now, I bought an amscope (the lab had one already), and am trying to convince the boss to go back to dedicated spaces in the lab. With two microscopes in the lab and only two people doing microscopic soldering it shouldn't be an issue anymore.
I still have plenty to do trying to hook up the Vega GPU to Mac Pro 2013, and currently delayering 5+ boards.
Last edited:
Where will future pcie expertise come from if apple is purposefully obfuscating the pcie standard behind thunderbolt (rhetorical)? There probably isn't even a standard way to develop a pcie driver for macos unless you are an apple employee. It is not fair. The new apple is the old microsoft, fueling the death of expertise. Why apple wants a huge ramp up for every new employee they hire to work on pcie / thunderbolt i have no idea. In a way this is the end of macos.
For me (using Sintech ST-NGFF2013-B or Amfeltec Angelshark) when booting from:
Apple stock 256GB AHCI SSD: 15s
Samsung 960 PRO 2TB: 2min
That's for both cold boot and reboot under macOS 10.13.4 and 10.13.5 Beta 4.
Last edited:
Again just a reminder the difference in boot time is not because of the adapter, it is because there is a slight difference in architecture between the previous and current generation of M.2 drives, and the boot sequence on Mac Pro 2013 notes this difference and runs extra testing once every cold boot because of it. If Samsung made a NVMe drive with the Apple native proprietary M.2 connector there would still be the extra boot time every cold boot, unless Apple decides to change the boot sequence to passively accept an NVMe drive without extra testing.
That is surprising I normally don't get the longer boot time for reboot. Unless something has changed in a recent os update. Does anyone get a shorter reboot time or shorter warm boot from full shutdown (the AC adapter stayed plugged in between shutdown and boot) with a Samsung NVMe running the latest macOS on Mac Pro 2013?
Last edited:
OMG the latest generation NVMe by Samsung is 33% cheaper than the previous generation, and both sustained transfer and random access are faster. Why is the newer one cheaper?!
can you answer my original question again, why the drive with better specs is 33% cheaper than the one with worse specs? please answer the question without regard to the superficial name difference.
the specs of the mentioned 970 EVO are better than the 960 PRO, for both sustained transfer throughput and latency for random access (and they are both 2TB). in this case i fail to see the difference a name makes, EVO vs PRO, in this case it seems clearly EVO is better than PRO, regardless of what the branding by Samsung would tell you.
maybe they are coming out with a 970 PRO, and the 970 PRO is better than the 970 EVO. that STILL would not explain the price difference i am talking about here. imagine as if we are comparing two random drives here, why is the better one 33% cheaper?
Last edited:
I read about this today....
Reports say Samsung is using 64 layer memory vs older 48 layer memory chips plus better power efficiency so less cooling required.
I guess new stuff gets cheaper as technology improves, and Samsung is finally getting competition from other manufacturers such as WD, which released its Black label NMVe ssd recently.
yes the EVO has worse cooling characteristics than the PRO, that is the only difference that i know of between the brands. So I can pay a few extra dollars for an NVMe heatsink and thermal paste/tape (if there is room in the computer case for a heatsinked NVMe), and theoretically the thermal performance at that point is the same, the cheaper one should perform better, and i've saved myself ~400 USD.
The PRO versions are using MLC (2 bits per cell) while the EVO uses TLC (3 bits per cell). In theory, MLC is more reliable and has a longer life than TLC (10000 cycles vs 5000) . It is also more expensive to produce.
In practice, the difference doesn't matter for most end-user use-cases.