Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

sentential

macrumors regular
May 27, 2015
169
72
Hi everyone,

I own a Mid-2015 rMBP with dGPU, too. And I have the same glitch at startup (after the bar reaches around 25% there is a short interruption, and after that the black screen with Apple logo appears and boot up is finished). And sometimes, I hear a clicking sound at startup, too.

BUT ... I own the rMBP for six months now, and I hadn't had a single kernel panic. May be luck ... or the glitches mentioned above may also appear if the dGPU (and the other hardware) is fine.

I am really curious to know the result of the hardware test at the Apple store...

Best regards,
Stephan

I wouldn't put much faith in the hardware test, mine passed with flying colors and as believe yours would as well. Mine also does the flashing you're referencing but I'm not worried about that. Some display modules on some GPUs do this when changing resolutions, others don't.

I'm not trying to defend Apple mind you as I believe this issue is not fixable and in our case first guys willing to go the extra mile are going to get the best outcomes.

I would much rather get $2500 in store credit to buy a non dGPU MacBook with Thunderbolt 3 than a warranty extension. TB3 is the real solution we power users need and won't be prone to these kinds of issues as a eGPU can be cooked and powered properly with non-Apple OEM drivers

That said for your sake get this on record now and work to either troubleshoot or fix it because I strongly suspect this thing is going to blow up, so please everyone reading this COVER YOUR ASS!
 

ThatSomeone

macrumors 6502
Original poster
Jan 9, 2016
323
73
Here, where else?
Yea you'd be surprised how little it takes to pop the dGPU, it's not that it can't handle it its just the threshold on switchable systems is very very low. Just opening chrome is enough to do it which is why I use Safari.

I'm not sure how familiar you are with Windows and overclocking in general but if you were to use something like CPUZ you can see how violently the CPU frequency thrashes during very basic use the same thing happens for the GPU both for enabling and disabling but also what frequency it operates on.

Hell this could be some kind of weird voltage sag that's causing it now that I think about it. The old Nelhalem i7s and current Visheras from AMD are notorious for it. When you push a chip right up to the edge with overclocking the margins go away and just because it's stable at 1.45v doesn't mean it'll stay that way during normal use.

Those chips would be ok on heavy loads but during light usage when power saving features kicked on the voltage would sag either a lot or just enough to crash the system forcing you to pad the voltage above minimums to keep the sagged voltage at 1.45 or recalibrate the PWM and MOSFET tolerances to compensate. It could just as easily by that than a bug or something related to the tin whisker effect the 2011s saw


I haven't personaly OCed any hardware in my life yet, but I've heard of the aggressive throttling on the GPU... I really hope this is an software bug and it'll be fixed in 10.11.3, but one can only hope... Hm, I've launched Chrome around 2 times since the crash and I haven't experienced anything. The chassi is way to cool at the moment to be running off the dGPU, been watching this stream for 1:30 hours.
It doesn't look like the dGPU has given up hope YET, the error message in the panic log doesn't has 'gpu panic' as the main reason which the other MacBook's throughout the years that had defective GPUs. So again, if we're lucky, this is all just an software bug...

Hi everyone,

I own a Mid-2015 rMBP with dGPU, too. And I have the same glitch at startup (after the bar reaches around 25% there is a short interruption, and after that the black screen with Apple logo appears and boot up is finished). And sometimes, I hear a clicking sound at startup, too.

BUT ... I own the rMBP for six months now, and I hadn't had a single kernel panic. May be luck ... or the glitches mentioned above may also appear if the dGPU (and the other hardware) is fine.

I am really curious to know the result of the hardware test at the Apple store...

Best regards,
Stephan

Intresting... I'm also very curious to see what the test says aswell. I'll be sure to mention the thread to the technician and hopefully they can seriously look into it!

I wouldn't put much faith in the hardware test, mine passed with flying colors and as believe yours would as well. Mine also does the flashing you're referencing but I'm not worried about that. Some display modules on some GPUs do this when changing resolutions, others don't.

I'm not trying to defend Apple mind you as I believe this issue is not fixable and in our case first guys willing to go the extra mile are going to get the best outcomes.

I would much rather get $2500 in store credit to buy a non dGPU MacBook with Thunderbolt 3 than a warranty extension. TB3 is the real solution we power users need and won't be prone to these kinds of issues as a eGPU can be cooked and powered properly with non-Apple OEM drivers

That said for your sake get this on record now and work to either troubleshoot or fix it because I strongly suspect this thing is going to blow up, so please everyone reading this COVER YOUR ASS!

Hm, did Apple run their test aswell or was it the one built in? If it comes back clean, I'll be sure to tell the Genius that this is a serious issue and the software engineers must be notified URGENTLY.
 
Last edited:

sentential

macrumors regular
May 27, 2015
169
72
I haven't personaly OCed any hardware in my life yet, but I've heard of the aggressive throttling on the GPU... I really hope this is an software bug and it'll be fixed in 10.11.3, but one can only hope... Hm, I've launched Chrome around 2 times since the crash and I haven't experienced anything. The chassi is way to cool at the moment to be running off the dGPU, been watching this stream for 1:30 hours.
It doesn't look like the dGPU has given up hope YET, the error message in the panic log doesn't has 'gpu panic' as the main reason which the other MacBook's throughout the years that had defective GPUs. So again, if we're lucky, this is all just an software bug...

Intresting... I'm also very curious to see what the test says aswell. I'll be sure to mention the thread to the technician and hopefully they can seriously look into it!

Overclocking is quite rewarding and gives a glimpse into the people who designed and made the chip as they all have their individual quirkyness particular to certain designers and manufacturers. The throttling you mention is completely normal btw.

Make no mistake about it the KP I linked IS GPU related, specifically the High-Definition-Audio decoder/encoder on the m370X GPU. I'm not an expert however Apple appears to use similar warnings. So...

ACG HDA Register Restore Failed:
ACG - accelerated graphics card
HDA - high Def audio
Register Restore Fail- either a GPU recovery failure or switch to integrated failed

Back on topic, I can only describe what I suspect is happening makes me think it is one of the following:

Software bug:

quite possible, my KP appears to occur when switching from the discrete to integrated. While I do not have my laptop it would be interesting to try and duplicate the issue by disabling graphics switching for observation purposes. However I'm lazy and have no patience for that, for what I paid that's apples problem since they make a point of keeping people out of their equipment

Voltage sag:

This is an interesting one especially now that I've had a chance to think about it. First this is the highest wattage GPU ever used in this chassis and most of my KPs occur while attached to AC power while charging. During that time the AC adapter gets hot enough that both the MagSafe connector and adapter are painfully hot to the touch.

Tin Whisker (aka Radeongate 2):

The fans are way to aggressive to maintain silence which causes wild spikes in temperature which I found to be unacceptable enough to disable turboboost in Bootcamp
 

David58117

macrumors 65816
Jan 24, 2013
1,237
523
Do the issues still occur if you disable the dGPU?

I haven't had a dGPU MBPr in awhile, but remember a program that let you control which GPU to use.
 

sentential

macrumors regular
May 27, 2015
169
72
Do the issues still occur if you disable the dGPU?

I haven't had a dGPU MBPr in awhile, but remember a program that let you control which GPU to use.

I don't have my laptop with me but it never crashed or artifacted in bootcamp which is dGPU only if that helps, so it seems to be a OSX issue only

Either way my MacBook is on its way to depot so we'll find out next week. Pics attached for proof
 

Attachments

  • image.jpeg
    image.jpeg
    645.9 KB · Views: 264
Last edited:

ThatSomeone

macrumors 6502
Original poster
Jan 9, 2016
323
73
Here, where else?
Overclocking is quite rewarding and gives a glimpse into the people who designed and made the chip as they all have their individual quirkyness particular to certain designers and manufacturers. The throttling you mention is completely normal btw.

Make no mistake about it the KP I linked IS GPU related, specifically the High-Definition-Audio decoder/encoder on the m370X GPU. I'm not an expert however Apple appears to use similar warnings. So...

ACG HDA Register Restore Failed:
ACG - accelerated graphics card
HDA - high Def audio
Register Restore Fail- either a GPU recovery failure or switch to integrated failed

Back on topic, I can only describe what I suspect is happening makes me think it is one of the following:

Software bug:

quite possible, my KP appears to occur when switching from the discrete to integrated. While I do not have my laptop it would be interesting to try and duplicate the issue by disabling graphics switching for observation purposes. However I'm lazy and have no patience for that, for what I paid that's apples problem since they make a point of keeping people out of their equipment

Voltage sag:

This is an interesting one especially now that I've had a chance to think about it. First this is the highest wattage GPU ever used in this chassis and most of my KPs occur while attached to AC power while charging. During that time the AC adapter gets hot enough that both the MagSafe connector and adapter are painfully hot to the touch.

Tin Whisker (aka Radeongate 2):

The fans are way to aggressive to maintain silence which causes wild spikes in temperature which I found to be unacceptable enough to disable turboboost in Bootcamp

There was a time roughly 2 years ago I thought of installing Windows to get better performance in games. Although, my MBA had a corrupt OS X installation and refused to install it and I didn't feel like reinstalling the OS(Never had any problems even though it was corrupt, how much? Absolutely no idea) and I scratched the idea.

I've got mine plugged in all the time.. I don't what Apple was thinking when they accepted AMDs contract. 25% higher TDP and 25-30% worse performance against the GTX 950M and no need to put in an more powerful battery without increasing the watt on the power brick... Heck, I think they could even put in an 960M AND implement Nvdia's power saving technology into OS X and yet stay under AMDs high TDPs. I hope 2016 is the year Nvidia makes a return to the Macs...

Unfortunately gfxcardstatus isn't no longer maintained(If you look at it's GitHub page, you can see it gets some updates, but that's just localization things), so I don't know how we'll be able to switch GPU as we like...

Stream just ended and I can happily report that I'm writing on the MacBook again!:D

I don't have my laptop with me but it never crashed or artifacted in bootcamp which is dGPU only if that helps, so it seems to be a OSX issue only

Either way my MacBook is on its way to depot so we'll find out next week. Pics attached for proof

Crossing my fingers it goes well!
Seems like it is an OS X bug then...
 

stanft

macrumors member
Jun 18, 2015
82
14
Hamburg, Germany
Oh well ... I just disabled the "automatic graphics switching" in the system preferences for some testing ... and boom ... the first kernel panic screen :-(

I had a look into "/Library/Logs/DiagnosticReports" ... and it seems the my machine generates a "com.apple.AmbientDisplayAgent" crash report at each startup.
 

sentential

macrumors regular
May 27, 2015
169
72
There was a time roughly 2 years ago I thought of installing Windows to get better performance in games. Although, my MBA had a corrupt OS X installation and refused to install it and I didn't feel like reinstalling the OS(Never had any problems even though it was corrupt, how much? Absolutely no idea) and I scratched the idea.

I've got mine plugged in all the time.. I don't what Apple was thinking when they accepted AMDs contract. 25% higher TDP and 25-30% worse performance against the GTX 950M and no need to put in an more powerful battery without increasing the watt on the power brick... Heck, I think they could even put in an 960M AND implement Nvdia's power saving technology into OS X and yet stay under AMDs high TDPs. I hope 2016 is the year Nvidia makes a return to the Macs...

Unfortunately gfxcardstatus isn't no longer maintained(If you look at it's GitHub page, you can see it gets some updates, but that's just localization things), so I don't know how we'll be able to switch GPU as we like...

Stream just ended and I can happily report that I'm writing on the MacBook again!:D



Crossing my fingers it goes well!
Seems like it is an OS X bug then...

Well I can tell you exactly why they want AMD and it is for the right reasons. Ever since Fermi nvidia has abandoned all things compute related and have focused entirely on games. Apple wants/needs compute and OpenCL support and unless anything changes I'd expect another m370x or something Polaris based for the next MBP. This is why I want actual cash or a credit and not next year's MBP as it'll have the same exact problems. The only way you'll see nvidia and Apple together again is if you use an eGPU with thunderbolt 3 which I plan on doing if given the opportunity as I don't see this situation being resolved anytime soon.
[doublepost=1452467691][/doublepost]
Oh well ... I just disabled the "automatic graphics switching" in the system preferences for some testing ... and boom ... the first kernel panic screen :-(

I had a look into "/Library/Logs/DiagnosticReports" ... and it seems the my machine generates a "com.apple.AmbientDisplayAgent" crash report at each startup.

Interesting... Is it reproducible? If so take it to an Apple Store asap
 

stanft

macrumors member
Jun 18, 2015
82
14
Hamburg, Germany
Interesting... Is it reproducible? If so take it to an Apple Store asap

Well, I am not able to reproduce the kernel panic. But the kernel panic report looks very identical to yours:

Code:
*** Panic Report ***
panic(cpu 0 caller 0xffffff7f949565ba): "AGC HDA REGISTER RESTORE FAILED : rdar://7254528, hda BAR0 invalid"@/Library/Caches/com.apple.xbs/Sources/AppleGraphicsControl/AppleGraphicsControl-3.12.6/src/AppleMuxControl/kext/AGCPowerManagement.cpp:1196
Backtrace (CPU 0), Frame : Return Address
0xffffff81fc493db0 : 0xffffff80120de792
0xffffff81fc493e30 : 0xffffff7f949565ba
0xffffff81fc493e70 : 0xffffff7f94959324
0xffffff81fc493ea0 : 0xffffff80126bac13
0xffffff81fc493f00 : 0xffffff8012112afa
0xffffff81fc493fb0 : 0xffffff80121cbf77
      Kernel Extensions in backtrace:
         com.apple.driver.AppleMuxControl(3.12.6)[44D361A1-4938-3AA5-9F73-9C909B66214B]@0xffffff7f9494b000->0xffffff7f9495efff
            dependency: com.apple.driver.AppleGraphicsControl(3.12.6)[1654475C-9A4B-386C-AFA8-0A530194A2F9]@0xffffff7f94943000
            dependency: com.apple.iokit.IOACPIFamily(1.4)[CBAE26D8-0ACB-3C1F-8347-FDCA67EC40B3]@0xffffff7f93007000
            dependency: com.apple.iokit.IOPCIFamily(2.9)[6ADFE830-2986-35D2-8C2F-6ABEDBC94F6A]@0xffffff7f928e9000
            dependency: com.apple.iokit.IOGraphicsFamily(2.4.1)[48AC8EA9-BD3C-3FDC-908D-09850215AA32]@0xffffff7f933bf000
            dependency: com.apple.driver.AppleBacklightExpert(1.1.0)[EBCEDD19-7230-3FB6-94B7-075ABA534875]@0xffffff7f94946000
 

ThatSomeone

macrumors 6502
Original poster
Jan 9, 2016
323
73
Here, where else?
Well, I am not able to reproduce the kernel panic. But the kernel panic report looks very identical to yours:

Code:
*** Panic Report ***
panic(cpu 0 caller 0xffffff7f949565ba): "AGC HDA REGISTER RESTORE FAILED : rdar://7254528, hda BAR0 invalid"@/Library/Caches/com.apple.xbs/Sources/AppleGraphicsControl/AppleGraphicsControl-3.12.6/src/AppleMuxControl/kext/AGCPowerManagement.cpp:1196
Backtrace (CPU 0), Frame : Return Address
0xffffff81fc493db0 : 0xffffff80120de792
0xffffff81fc493e30 : 0xffffff7f949565ba
0xffffff81fc493e70 : 0xffffff7f94959324
0xffffff81fc493ea0 : 0xffffff80126bac13
0xffffff81fc493f00 : 0xffffff8012112afa
0xffffff81fc493fb0 : 0xffffff80121cbf77
      Kernel Extensions in backtrace:
         com.apple.driver.AppleMuxControl(3.12.6)[44D361A1-4938-3AA5-9F73-9C909B66214B]@0xffffff7f9494b000->0xffffff7f9495efff
            dependency: com.apple.driver.AppleGraphicsControl(3.12.6)[1654475C-9A4B-386C-AFA8-0A530194A2F9]@0xffffff7f94943000
            dependency: com.apple.iokit.IOACPIFamily(1.4)[CBAE26D8-0ACB-3C1F-8347-FDCA67EC40B3]@0xffffff7f93007000
            dependency: com.apple.iokit.IOPCIFamily(2.9)[6ADFE830-2986-35D2-8C2F-6ABEDBC94F6A]@0xffffff7f928e9000
            dependency: com.apple.iokit.IOGraphicsFamily(2.4.1)[48AC8EA9-BD3C-3FDC-908D-09850215AA32]@0xffffff7f933bf000
            dependency: com.apple.driver.AppleBacklightExpert(1.1.0)[EBCEDD19-7230-3FB6-94B7-075ABA534875]@0xffffff7f94946000

Now this is very interesting.. It does indeed look just the same as ours!
This is so weird, I had mine crashed for the first time on saturday morning, sentential had his crash on saturday too and now you get the very first one today... Is there some sort of message? :eek:
Jokes aside, this is perfect to show Apple, we might be able to get an end to this!
I'll be sure to tell the technician that more people has this problem and demand that the software engineers are notified!
 

sentential

macrumors regular
May 27, 2015
169
72
Well, I am not able to reproduce the kernel panic. But the kernel panic report looks very identical to yours:

Code:
*** Panic Report ***
panic(cpu 0 caller 0xffffff7f949565ba): "AGC HDA REGISTER RESTORE FAILED : rdar://7254528, hda BAR0 invalid"@/Library/Caches/com.apple.xbs/Sources/AppleGraphicsControl/AppleGraphicsControl-3.12.6/src/AppleMuxControl/kext/AGCPowerManagement.cpp:1196
Backtrace (CPU 0), Frame : Return Address
0xffffff81fc493db0 : 0xffffff80120de792
0xffffff81fc493e30 : 0xffffff7f949565ba
0xffffff81fc493e70 : 0xffffff7f94959324
0xffffff81fc493ea0 : 0xffffff80126bac13
0xffffff81fc493f00 : 0xffffff8012112afa
0xffffff81fc493fb0 : 0xffffff80121cbf77
      Kernel Extensions in backtrace:
         com.apple.driver.AppleMuxControl(3.12.6)[44D361A1-4938-3AA5-9F73-9C909B66214B]@0xffffff7f9494b000->0xffffff7f9495efff
            dependency: com.apple.driver.AppleGraphicsControl(3.12.6)[1654475C-9A4B-386C-AFA8-0A530194A2F9]@0xffffff7f94943000
            dependency: com.apple.iokit.IOACPIFamily(1.4)[CBAE26D8-0ACB-3C1F-8347-FDCA67EC40B3]@0xffffff7f93007000
            dependency: com.apple.iokit.IOPCIFamily(2.9)[6ADFE830-2986-35D2-8C2F-6ABEDBC94F6A]@0xffffff7f928e9000
            dependency: com.apple.iokit.IOGraphicsFamily(2.4.1)[48AC8EA9-BD3C-3FDC-908D-09850215AA32]@0xffffff7f933bf000
            dependency: com.apple.driver.AppleBacklightExpert(1.1.0)[EBCEDD19-7230-3FB6-94B7-075ABA534875]@0xffffff7f94946000

That's a shame, I'd still take it in because the only way this is going to get fixed is if people bitch about it. Mine started crashing shortly after updating to 10.11.2 in early December
 

ThatSomeone

macrumors 6502
Original poster
Jan 9, 2016
323
73
Here, where else?
Definitely couldn't hurt

I'll start right away!
[doublepost=1452535620][/doublepost]To sum up:

On friday night, my brand new rMBP had an kernel panic. I contacted Apple and got it booked in at my local Apple Store, and I then decided to open this thread asking some questions if my logic board had to be replaced. However, it turned out that I wasen't alone.. After some replies we concluded that this is an bug in OS X, so no matter how many times you change your logic board, your whole rMBP or even re-install OS X, you're still vulnerable. So all we can do is to tell Apple this is a highly serious bug which may crash any rMBP with an dGPU and it needs to be fixed ASAP!

I'm unsure if this affects users with an Nvidia card, but it definitley seems to affect the AMD M370X. And it's also unkown if gfxcardstatus works with the newer models as development has been abandoned by its developer.
 
Last edited:

noodle654

macrumors 68020
Jun 2, 2005
2,070
22
Never Ender
Hi everyone,

I own a Mid-2015 rMBP with dGPU, too. And I have the same glitch at startup (after the bar reaches around 25% there is a short interruption, and after that the black screen with Apple logo appears and boot up is finished). And sometimes, I hear a clicking sound at startup, too.

BUT ... I own the rMBP for six months now, and I hadn't had a single kernel panic. May be luck ... or the glitches mentioned above may also appear if the dGPU (and the other hardware) is fine.

I am really curious to know the result of the hardware test at the Apple store...

Best regards,
Stephan

Wait wait wait...I have this happen too. Computer boots almost immediately to my login with desktop image, once I enter my password it loads 25%, screen goes black, black screen with Apple logo and loading bar comes up, and then eventually logs in. This is related to the dGPU? I thought maybe my OS is corrupt (seen a few upgrades, probably should start fresh soon).
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.