Stability issues with 4930k, Sabertooth X79 -- ONLY when PM SSDT loaded

[mr. west]

First, a word of thanks to Omni, rampagedev, and others who have worked to get Power Management working on LGA 2011 boards.

 

My current rig is a Sabertooth X79 (latest BIOS), 4930k PCU, 32gb 1866 memory (Corsair), EVGA GTX 650 graphics.

 

I ran solid for about a year using a i7-3820.  No stability issues at all.  Everything just works.  I upgraded to a 4930k a few months ago without much incident.  No stability issues at all and again everything just works.

 

I was excited to see that Apple finally supported Power Management for LGA2011 CPUs in 10.9.2.   Following guides by rampagedev (http://rampagedev.wordpress.com/) I was able to get PM up and running, both under Chameleon and Clover.  It works using the vanilla AICPUPM kext and an SSDT generated using Pike's script.

 

However, what's tripped me up has been some stability issues.  When I load SSDT I get random restarts.  I've tried to pin down the cause using system log but it doesn't seem much help.  I've chased a few red herrings (certain applications, sound drivers, etc.) but no matter what I do the crashes persist. Sometimes the restarts are instant, but typically they come after things start to crash -- especially Chrome and Finder.

 

When I remove the PM SSDT things once again become stable.  

 

I've run memtest and my memory is fine.

 

I've tried various BIOS settings and as long as I have the PM SSDT loaded, I get the random crashes.  It happens on stock settings, overclock settings, etc. 

 

Shilohh suggested I do a CMOS reset, load optimized defaults, and try things out.  I'm going to try that now, leaving my 1866-certified RAM at 1600.

 

 

Any ideas?

[mr. west]

Hi mr. west.

 

reading some others users experience with your CPU, it seems not properly supported at this stage by any PM SSDT.

Some users have good results with a SSDT that you can find in omni thread released by Balamut

 

If I understand well your system is pretty stable without using any SSDT

In this case is your CPU working well?

Do you have states? Or system is stuck to one frequency?

Usually, without using a SSDT system uses more watts to feed your CPU a due to this maybe you have a sort of good stability.

 

I don't know if you have ever test your setup with others OS with a stability test, maybe you have to try

 

Hi fabio.  

 

Yes, system is stable without SSDT, including with a strap-based Overclock.  No issues at all once I get the voltage right for OCs.  No issues at all on defaults with stock.

 

Only with the SSDT.  With the SSDT it works beautifully UNTIL the crash.  I get states 12, 13, 23, 29, 34, 36, 37, 39.  Power consumption is super low at idle.  Temps are great.  But then comes the crash, sometimes as long as 2 days later, but it is consistent.

So perhaps it is related to power nevertheless.

 

My next attempt is this.  I've reset my CMOS and defaulted everything to optimized defaults.  No XMP even.  I've got a working SSDT in, running Prime 95 torture test in the background while I work.  If I can stay stable for quite some time, I might try bumping memory to 1600MHZ.  Then 1866MHz (my memory is rated for that).  I'm happy there.  Maybe try OC then, but the 4930k doesn't OC well as it is, and I think I got a weak chip.  I need 1.4Vcore just to run 4.25 stable -- hardly seems worth it for an extra 300 turbo MHz on top of stock.

 

 

As noted, stable in OS X, Windows, and occasionally Linux for over a year with the 3820 on same MB, and another 3 months with 4930k.  Only issues when loading an SSDT for PM.

[jasjeet]

Quite odd, im not having any issus with my 4960x at 4.3Ghz currently. I have run about 2 consecutive hours at most though.

[Rampage Dev]

Are you OCing the CPU? 

[mr. west]

Are you OCing the CPU? 

No, not beyond the Turbo setting which is stock.

 

Just crashed again on optimized defaults (PC1333 Memory and everything).  I had Prime95 torture test running for hours today while doing a lot of work -- no crashes.  Then I walked away to greet my wife for 10 minutes and I came back to a restarted machine.

 

I made a gist of the system log right before the crash:

 

https://gist.github.com/anonymous/9630931

 

As you can see, some Google Chrome activity.  I had suspected that may have something to do with it before, so I ran Safari all day and it still randomly restarted later. Frustrating.

 

One clue -- I had a vimeo video on when it crashed.  I seem to recall other times videos were on when it crashed.  Not sure though.

 

Maybe a video card thing?  Is that possible?  I have noticed in HWMonitor the GPU2 memory gets stuck at 5 GHz sometimes.  Not sure if that happens when there is no SSDT.  Right now with no SSDT it looks normal.

 

Shilohh suggested trying StepContextDict.  I may give that a go. 

 

Ideas?

[Pike R. Alpha]

This looks like a hardware issue triggered by power management. First untick the checkbox: System Preferences / Energy Saver / Start up automatically after a power failure and see if it still restarts with the power management SSDT or just switches off. In that case it is a software error and you will need to start checking drivers (kexts) by unloading all video, network and other drivers that are not absolutely necessary for a stable system. Add them back one after the other until it starts to reboot.

 

I would also like to suggest that you stop using prime 95 as torture test. Thing is it is hugely overrated and not designed for modern processors. And don't OC with a processor that requires a high voltage because that usually is a sign that it starting to give in. What you can try is to use a higher factory wattage (TDP) to see if that helps – your CPU may not like low voltages anymore.

[mr. west]

This looks like a hardware issue triggered by power management. First untick the checkbox: System Preferences / Energy Saver / Start up automatically after a power failure and see if it still restarts with the power management SSDT or just switches off. In that case it is a software error and you will need to start checking drivers (kexts) by unloading all video, network and other drivers that are not absolutely necessary for a stable system. Add them back one after the other until it starts to reboot.

 

I would also like to suggest that you stop using prime 95 as torture test. Thing is it is hugely overrated and not designed for modern processors. And don't OC with a processor that requires a high voltage because that usually is a sign that it starting to give in. What you can try is to use a higher factory wattage (TDP) to see if that helps – your CPU may not like low voltages anymore.

 

Pike, Thanks for the suggestions.  Interesting.

 

I haven't really OCed this chip much at all.  Most of the time I've run stock.  Periodically I've run with strap overclock or turbo with PM, but quickly go back to stock.  Confident the chip is fine.  

 

 

Anyhow, I already "automatic restart" is unchecked, so I know for sure it restarts automatically.

 

Perhaps you or Andrew could give me advice.  The only kexts I use are:

 

FakeSMC

GenericUSBXHCI

AppleIntelE1000

 

and Toleda's audio patch.

 

I also use Andrew's SSDT for Sabertooth board + nvidia injection.

 

I suppose I'll remove the USB and LAN kexts (neither are crucial) and see how we do.  If I crash I'll go without sound and see how that goes.  Maybe it's the USB?  That seems plausible. 

 

 

Any other ideas?  Could it possibly be my video card?

[mr. west]

Well, I ran two days stable without the GenericUSB driver kext but I crashed over night.  

 

On the software front, that leaves audio.

 

I also wonder if it could be the SSDT I'm using (Andrew's for the Sabertooth X79).  Without it I can't get audio to work using Toleda's.  Maybe I should make my own SSDT. 

[Rampage Dev]

Don't change the ssdt. Not the issue. Could be a physical hardware issue.

[mr. west]

Don't change the ssdt. Not the issue. Could be a physical hardware issue.

 

Any guesses on where to go from here? It's just strange it only happens with a PM SSDT and never otherwise.  What hardware would be affected by that?

 

GPU?

[Rampage Dev]

Voltage drops and or lack of current when changing steppings.

[mr. west]

Voltage drops and or lack of current when changing steppings.

 

Makes sense.  What BIOS settings would I adjust to avoid those issues?  Manual VCORE + higher load line calibration?

 

I just set a manual voltage, upped the LLC, and increased the current capability.  We'll see if that makes a difference.  I'd done this before but no luck.  

 

I also turned off Time Machine because there was come correlation between that in the logs and crashing. Also it crashes overnight when nothing is happening but time machine.  So maybe that's it...

 

Guess I keep trying.

[rotoyouoio]

Well, I ran two days stable without the GenericUSB driver kext but I crashed over night.  

 

On the software front, that leaves audio.

 

I also wonder if it could be the SSDT I'm using (Andrew's for the Sabertooth X79).  Without it I can't get audio to work using Toleda's.  Maybe I should make my own SSDT. 

 

 

Voltage drops and or lack of current when changing steppings.

how did this resolve? i have similar behaviour on an msi board with the same processor. for an example when launching Davinci (not when it renders). though the computer stayed on for 2 weeks for distant access without a problem. also it has the states generated by clover without ssdt (btw clover 2682+ doesn't generate those cause they were "fixed" in 2682 so i have to convert to ssdt for future OSX, which i don't wanna do cause the wrongly generated states consume much less power than the ssdt)