Sudden, regular reboots started tonight.. reboots stopped, but now multiple SigSegv code 11 app crashes

[TheBloke]

Hi all

 

My 10.13.3 system (which was upgraded from 10.11.6 a week ago) has suddenly started rebooting frequently tonight.  Since it first happened about 90 minutes ago, it has rebooted at least six times.   There is no warning for each reboot - the screens go black, and moments later I see my BIOS screen.

 

I am using Clover 4411, booting in Legacy mode on my Gigabyte GA-X58A-UD3R with Xeon X5670.

 

In Console I see the following error repeated thousands of times:

kernel  Sandbox: mdworker(555) deny(1) system-fsctl 4a0e

 

I don't know if this is a new error or not as I was not watching the Console much before tonight.  Google shows this error coming up related to SIP issues in High Sierra, but I have not added or changed any kexts in the last few days.  I am booting with kext-dev-mode=1 and have been for the last week.

 

I have checked /var/log/system.log to see if I can see any error just before reboot, but I cannot see any 'smoking gun'.  I do have these errors before the most recent reboot:

Feb 20 20:29:30 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[228]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]

Feb 20 20:29:36 Marvin com.apple.xpc.launchd[1] (com.apple.mdworker.single.0E000000-0000-0000-0000-000000000000): Service only ran for 9 seconds. Pushing respawn out by 1 seconds.

 

Before the previous reboot, these were the last /var/log/system.log errors:

Feb 20 20:24:46 Marvin com.apple.xpc.launchd[1] (com.apple.mdworker.single.0C000000-0000-0000-0000-000000000000): Service only ran for 5 seconds. Pushing respawn out by 5 seconds.

Feb 20 20:24:46 Marvin com.apple.xpc.launchd[1] (com.apple.mdworker.single.09000000-0000-0000-0000-000000000000): Service only ran for 5 seconds. Pushing respawn out by 5 seconds.

Feb 20 20:24:47 Marvin com.apple.xpc.launchd[1] (com.apple.mdworker.single.08000000-0000-0000-0000-000000000000): Service only ran for 5 seconds. Pushing respawn out by 5 seconds.

Feb 20 20:24:47 Marvin com.apple.xpc.launchd[1] (com.apple.mdworker.single.05000000-0000-0000-0000-000000000000): Service only ran for 5 seconds. Pushing respawn out by 5 seconds.

Feb 20 20:24:47 Marvin com.apple.xpc.launchd[1] (com.apple.mdworker.single.07000000-0000-0000-0000-000000000000): Service only ran for 6 seconds. Pushing respawn out by 4 seconds.

Feb 20 20:24:49 Marvin com.apple.xpc.launchd[1] (com.apple.preference.displays.MirrorDisplays): Service only ran for 0 seconds. Pushing respawn out by 10 seconds.

 

So it is not always the same errors.  But the mdworker deny(1) error in Console is always there.   I had to close Console eventually because it had 500,000 messages in it since boot 20 mins ago!

 

I can also see in Activity Monitor that mds, coreservicesd and mds_stores are all taking up a lot of CPU - right now for example, mds is taking 130% CPU and mds_stores is at 112%! I also see multiple instances of mdworker in the Activity Monitor list.

 

Here's a screenshot of Activity Monitor showing an example of the CPU usage I see:

 

Attached is:

•  
• a zip of the EFI folder I am currently booting from.  But note that I have not changed any config in the last few days; the issue suddenly started tonight without warning or changes.
• my current /var/log/system.log file and the previous one (Gzip'd)

 

I don't know how I can save/copy the information I see scrolling in Console, as I think it's stored in a binary file?

 

However on this latest boot, I so far have not rebooted after 30 minutes.  This is the longest time I've been up since the problems started - before it was usually crashing within 5 minutes, or 10 minutes tops.  The difference is that this time I closed some programs I had open before, especially Firefox with many tabs.  So maybe certain software is able to trigger the problem?

 

But something is clearly still not right, based on all the CPU usage being used by mds, mds_stores, diagnosticd and others.  I have also seen ReportCrash and trustd taking up some CPU.

 

My full spec is in my sig. Note that my sig says I'm using an LSI card for boot, but since the problems started I have moved my boot SSD back to onboard SATA3 to rule the LSI card out as a problem.  Also I am currently using an NVidia 760, not the 980Ti listed in the sig (I have been testing difference between Native vs NVidiaWeb drivers.)

 

As the system has still not crashed after 30 minutes now, I will now wait another ~30 minutes to be sure it still doesn't crash, and then I will open Firefox to see if that triggers it again.  But first I wanted to make this post!

 

Thanks very much in advance.

EFI.zip

system.log.200218-2043.txt

system.log.previous.200218-2045.txt.gz

[PPCnostalgic]

Try this

config.plist.zip

[TheBloke]

OK thanks a lot.

 

I see you've made quite a few changes to the ACPI section.   When I was on 10.11.6 and I upgraded Clover to 4411 I had to add FixUSB and FixDarwin or it wouldn't even boot.  But then I wasn't sure if I still needed them in 10.13.  I see you've removed them, so that gives me an answer.

 

One thing to report before I reboot with your config.plist:  after my post I went away, leaving the OS running for 15 minutes, and A. it didn't crash (still haven't opened Firefox though) and B. all the high CPU on mds, mdworker and mds_store has gone.

 

So maybe these processes were doing some long-running task and that has now finished?  Maybe I had file system problems?

 

Anyway I will try your config.plist and report back in 15 minutes or so.

 

Thanks again!

 

ADDED:

OK I'm booted successfully with your config.plist.   After boot I noticed I did have a couple of mdworker processes using some CPU - not as high as before, but 30-50%.

 

I know mdworker is Spotlight.  I thought I had disabled Spotlight indexing by unticking everything in System Preferences->Spotlight->"only selected categories will appear.."  But I think I need to instead go to Privacy and add my entire SSD to "Prevent Spotlight from searching these locations".  I don't use Spotlight as I have Alfred 3 which does a similar but better job.

 

Anyway, so far so good..  booted for 20 minutes now, with Firefox running, and no sign of a crash.  And those mdworker processes have dropped out of Activity Monitor again, and I haven't seen mds or mds_stores appear at all.  Also, Console is not full of thousands of errors any more.

 

I don't know if any of this is related to the new config.plist, given the high CPU had already stopped before I rebooted.   But if I had bad ACPI settings then it has to be good to fix them as well.

 

Do you find anything in my config.plist that would definitely be a problem?

 

Thanks again for your help!

[PPCnostalgic]

Your motherboard is quite similar to mine (see signature) and we both have Mac Pro 5,1 as SMBIOS but yours had a mistake in the firmware date (it was 2010) and it's not the right version to run Sierra 10.13.x So I copied my firmware date release (you can see it's 2017 now) and pasted onto your config.plist, this was the main culprit IMHO. The other (little) changes I made it's just because of our Mobos similarity (X58-GA) and I thought what's working for me it could work for you as well...

[TheBloke]

OK wonderful, thanks very much for all your fast help!

[TheBloke]

OK no it's not fixed.. I have serious problems here

 

No further reboots or high CPU visible.  But apps are crashing all over the place, always with SigSegv code 11.  

 

Activity Monitor, iTerm2, clang processes when I was trying to compile something.   iTerm2 and clang crashed three or four times each before I got them to run OK again.  iTerm 2 first crashed after I had been running it for hours, just randomly in the middle of a session.  Then it crashed again immediately twice more when I hit restart, and on the third time it survived and is currently still running.  EDIT:  looks like uBar crashed at some point as well.

 

I don't know what's changed to break all this   I have been installing software recently, like some Photoshop plugins from Topaz Labs - which I later found had put some libraries in /usr/local/lib.  Maybe it's put a bad dynlib somewhere?  But then why would programs crash sometimes and not always.   And I haven't done any system updates recently since I upgraded to 10.13.3.   All I've updated  from Apple is XCode and iTunes, but that was a week or so ago.

 

Maybe it's bad hardware?  Maybe first thing I should do is roll back my OC settings a bit?  They've been stable for the last week at least (when I came back to macOS) but maybe they're not fully stable or something..

 

Actually I did make a hardware change fairly recently - I replaced my NVidia 980Ti with an NVidia 760 so I could run Native drivers.  But that was two days ago, and problems only started tonight.  And also seems really odd that GPU would cause these issues.

 

Of course I did change my ACPI settings tonight, moving to the settings working for PPCnostalgic.   But again, that seems like it would cause problems on boot, not SigSegv crashes of random apps at random times?

 

Attached is a recent iTerm2 crash log, and here's what's in a search of system.log for "segmentation":

 

Feb 21 01:44:15 Marvin com.apple.xpc.launchd[1] (com.apple.xpc.launchd.oneshot.0x10000005.Activity Monitor[303]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 02:15:50 Marvin com.apple.xpc.launchd[1] (com.apple.xpc.launchd.oneshot.0x10000002.iTerm2[300]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 02:16:47 Marvin com.apple.xpc.launchd[1] (com.googlecode.iterm2.5004[34266]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 02:16:47 Marvin com.apple.xpc.launchd[1] (ca.brawer.uBar.5276[533]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]

iterm2.crash.log.txt

[chris1111]

what is the web drivers revision you have installed?

[TheBloke]

what is the web drivers revision you have installed?

 

I currently have 387.10.10.10.25.156 installed (I was testing it versus 106), but I am booted with Native drivers on my 760:

 

 

$ kextstat | grep -i nvid

  120    2 0xffffff7f812c8000 0x2e6000   0x2e6000   com.apple.nvidia.driver.NVDAResman (10.2.8) 39AC3832-2295-3B7B-85F6-F69D66D46EC7 <115 106 98 12 7 5 4 3 1>

  121    0 0xffffff7f827c3000 0x1b3000   0x1b3000   com.apple.nvidia.driver.NVDAGK100Hal (10.2.8) 7064CB37-CB6D-34F7-974D-DD12944BD354 <120 12 4 3>

Do you think it could cause an issue even if not used?  That is something I did change yesterday - reinstalling 156.  But I didn't think it would affect anything if I wasn't using them.

[chris1111]

I currently have 387.10.10.10.25.156 installed (I was testing it versus 106), but I am booted with Native drivers on my 760:

 

 

$ kextstat | grep -i nvid

  120    2 0xffffff7f812c8000 0x2e6000   0x2e6000   com.apple.nvidia.driver.NVDAResman (10.2.8) 39AC3832-2295-3B7B-85F6-F69D66D46EC7 <115 106 98 12 7 5 4 3 1>

  121    0 0xffffff7f827c3000 0x1b3000   0x1b3000   com.apple.nvidia.driver.NVDAGK100Hal (10.2.8) 7064CB37-CB6D-34F7-974D-DD12944BD354 <120 12 4 3>

Do you think it could cause an issue even if not used?  That is something I did change yesterday - reinstalling 156.  But I didn't think it would affect anything if I wasn't using them.

I do not think you have anything to lose from removing them

[TheBloke]

OK I'll revert to 106, which I was using for a week or so without problems.

 

PS. I'm also getting multiple Firefox tab crashes.  Can't see any logs for them, but again I've not had that before tonight and now it's happened three times in an hour.

 

EDIT:

 

OK I've gone back to 106, though still booting with Native.  I also booted into Windows and ran IntelBurnTest on Maximum for about 25 minutes to see if I could spot any hardware errors.  No problems, ran fine.

 

I will see what happens on this boot..

[PPCnostalgic]

Uhm..... weird is all I can say. Try removing Ram sticks all but Dimm 1 & Dimm 2 and test.

BTW did you remove AppleTyMCEDriver.kext from your S/L/E folder?

[TheBloke]

Uhm..... weird is all I can say. Try removing Ram sticks all but Dimm 1 & Dimm 2 and test.

 

Yeah it's really weird.  So far today I've had no major applications crash, however I just had a couple of Firefox tabs crash in quick succession, and checking my system.log show that there have been several segmentation fault failures of Spotlight indexing processes:

 

Feb 21 13:39:36 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1291]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:40:48 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds[1404]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:41:55 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1405]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:45:03 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1489]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:45:31 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds[1595]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:46:37 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1596]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:47:12 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1681]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:48:13 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1714]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:51:12 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1789]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 13:52:33 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1859]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 14:09:32 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[1926]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]
Feb 21 14:13:13 Marvin com.apple.xpc.launchd[1] (com.apple.metadata.mds.index[2371]): Service exited due to signal: Segmentation fault: 11 sent by exc handler[0]

 

I had to turn Spotlight indexing back on because it turns out that Alfred, the spotlight 'replacement' I use, actually uses Spotlight indexing.   Re-enabling spotlight caused quite a bit of CPU from mdworker, mds, mds_store processes for an hour or two, but that's now gone away. 

 

So right now I am assuming that something happened yesterday to trigger the first hard crash/reboot.  That 'something' caused the Spotlight indexes to fry, which then caused all the high CPU from md* processes I complained about over the next several hours.  I had further hard reboots over the next hour, which were presumably from the same cause as the first reboot, perhaps made more likely by the spotlight indexing.  

 

Since then I've had no reboots and Spotlight indexing is under control.  But I have this ongoing problem of Segmentation Fault crashes at random times.  

 

Is it all related?  That seems very likely.  Maybe the reboots were triggered by Seg Faults of vital processes, which in turn corrupted the Spotlight indexes forcing lots of CPU to rebuild them, and making it more likely for further crashes if those processes crashed again while trying to do their job.   Once the indexing was mostly completed, the risk of reboot got much lower.  But the underlying Seg Fault problem remains.

 

Yes I suppose I should next look at hardware.  Maybe revert my OC, maybe remove RAM.   It does seem weird though that IntelBurnTest can run fine, given that normally shows up in minutes problems that take days, weeks or months to show in normal usage.   But I did only run it for 25 minutes, maybe I should run it for a couple of hours at least.

 

I've just checked all my /var/system.log archives, which go back a week, to Feb 13th.  There are mentions of "Segmentation fault" in nearly all of them.  So this is not such a new problem after all. 

 

The oldest reference, in Feb 13th's log, was Adobe's AAM Updater which is probably a process I had left over from El Capitan, so that might not be relevant.  No crashes on Feb 14th or Feb 15th. 

 

Feb 16th had one crash, from mds - similar to the many I've had today.  And Feb 17th had one from com.adobe.Install; I guess that was the day I installed Photoshop or one of my other Creative Cloud products.

 

No crashes on Feb 18th, then on Feb 19th the NVidia PrefPane seg faulted once.  Then we get to Feb 20th, yesterday, where the problem seemingly explodes, with 19 total crashes; mostly mds and other Spotlight processes, but also Activity Monitor, iTerm, uBar and others as mentioned before.

 

From all this I suppose I draw the conclusion that while things are definitely far worse now, it is perhaps not a brand new problem.  The SegFault of mds five days ago seems to suggest that.  But it's definitely getting more common and more severe.  That could point to hardware, or it could point to problems in the OS files;  this is an OS that was originally installed on 10.10 I think, and then was upgraded to 10.11 and then recently straight to 10.13. 

 

Right now the system is definitely usable, but there's always a question mark hanging over things - eg my Firefox tabs crashing today.    I will try some hardware reverting, and if that doesn't solve things I suppose I will need to think about a fresh install  

 

 

ADDED:

 

Yeah actually things are still pretty bad.  Still no foreground app crashes, besides Firefox tabs (but a lot of those - maybe 10 or 15 now in the last hour), but a couple more background ones - like com.apple.AddressBook.Sync.

 

And I thought I'd run a virus scan to be safe.  I downloaded the free trial of Kaspersky, which is a small installer which then downloads the main program.  After download it runs a CRC checksum on what it downloaded.  Which is good, because that download failed five times in a row - I can only assume as a result of the downloader crashing.  At the same time that was running, Firefox tabs were crashing repeatedly.

 

I'm now scared of installing anything else in case the installers crash in the middle and leave me in a half-installed state, and add to my problems.   I'm going to let this Kaspersky scan complete then I'm going to revert my BIOS to default settings and re-configure just what's needed to boot, no OC, and try that for a bit.

 

BTW did you remove AppleTyMCEDriver.kext from your S/L/E folder?

 

No I still have that file in /S/L/E. I'd not heard of it before but I've just researched it - seems it's related to ECC RAM and using the MacPro 5.1 SMBIOS?  

 

Apparently it's sometimes recommended for X58 users on MacPro 5.1 to add the flag -nehalem_error_disable to boot, which does the same thing as patching AppleTyMCEDriver.

 

So I've tried adding that to Clover andit boots OK.    I don't fully understand this, as I've used MacPro 5.1 from day 1 and I don't have any KextPatching for AppleTyMCEDriver.  I've never used ECC RAM.

 

I don't know if it's changing anything, but I do now have -nehalem_error_disable and I'll run the system for a bit to confirm if anything is differnt.

 

I'd be grateful if you could clarify what you were referring to regarding that kext - whether I should have a patch for it or whatever.  Thanks.

[PPCnostalgic]

You are right, that kext should be removed if you don't use ECC Ram but I still have it in my S/L/E too on the GA-X58 and it works fine despite I have no ECC Ram sticks. On the HP I have ECC Ram so it's good as is. Let's talk about your OC settings, how did you set your CPU Voltage? Mine is set to 1,225 + 0,1250 = 1,3500max and it never goes further than 1,3100v

I have Dynamic Vcore checked as well. If I were in your shoes I'd try lowering your CPU 1 step down (leaving every voltage setting untouched) so it should be 23x190 = 4,370Ghz that is quite a high clock for your CPU (mine is 33x133 + Turbo Boost = 4,389Ghz - 4,5220Ghz). If I go 1step beyond my system crashes the same way yours does unless I rise up voltage to 1,38v that's quite too much, the CPU will get very hot for a very little speed bump, not worth it IMHO.

[TheBloke]

OK yeah it looks like it's OC related.

I reset BIOS to defaults and then configured just the peripherals, leaving all OC settings at default.

To test this I am running the command: brew reinstall macvim

This causes the HomeBrew software library to recompile MacVim, which uses XCode's clang to build MacVim and takes about 70 seconds total.  It was running brew upgrade macvim a couple of days ago that first indicated I had problems - it took me three or four attempts to get it built right.  Then yesterday I tried it again, before I reset the BIOS, and it seg faulted almost immediately.  And interestingly, this then caused softwareupdated to seg fault in the background.

Since I reset the BIOS, I've run brew reinstall macvim in a loop more than 40 times with no problems.

So clearly I need to go back to square one with the OC.  I'm actually running new OC settings - 190x24 with Turbo enabled, CPU on Normal voltage with a +DVID of 0.243, RAM voltage of 1.68, and everything else on Auto.   Before last week I had 200x22, no turbo, +DVID 0.281 and custom voltages for nearly everything.  I'd run that OC for several months with no problems in Windows.  But then in macOS I saw I wasn't getting full SpeedStep - I only got x12 and x24 multipliers.  I was told that I needed Turbo enabled for that, and then I followed Emilo's OC guide for the X5670 on Gigabyte X58 motherboard.

Clearly it's not working properly for me, so I'm going to either need to tone it down a bit, or else go back to my earlier no-Turbo OC.  Assuming of course that does work fine in macOS like it did for months in Windows.

Anyway, assuming I can get the OC sorted, this won't have been the worst experience because it's also caused me to clear up my config.plist and also look through /S/L/E and /L/E and remove some unneeded kexts and upgraded some outdated ones that were still there from my original late-2015 install.    It didn't help with this problem but it's good to have a cleaner system going forward.

Thanks again for your help, PPCnostalgic.

[PPCnostalgic]

You said you don't see many CPU speedsteps... I had the same problem so I modified a kext to achieve more speedsteps. If you like to give it a shot it's attached to this post.

It works only with Mac Pro 5,1 SMBIOS that's our instance.

BTW I can say this kext works fine and it's stable, but backup your original kext just in case...

 

IOPlatformPluginFamily.kext.zip

[TheBloke]

You said you don't see many CPU speedsteps... I had the same problem so I modified a kext to achieve more speedsteps. If you like to give it a shot it's attached to this post.

It works only with Mac Pro 5,1 SMBIOS that's our instance.

BTW I can say this kext works fine and it's stable, but backup your original kext just in case...

 

Awesome, thanks.  With a Turbo OC, I currently get the following states:  x12, x21, x22, x23 and x24.   That's better than a non-Turbo OC which only gave me x12 and x22, but I still expected to get more states between x12 and x21.

 

If your edited kext achieves that then that's awesome, I will definitely try it.

 

Quick newbie question: where should I put your file?  Do I replace the original IOPlatformPluginFamily.kext in /S/L/E?  Or can I override it by putting your version in /L/E ?  Or should I have it loaded at boot by putting it in EFI/CLOVER/kexts/Other?

 

I'm still a bit hazy on what kexts should go in which locations.

 

Thanks again.

[PPCnostalgic]

Mine is in S/L/E and I replaced the original. You can manually put the kext in your S/L/E (replace original) and then run the attached utility

Kext Utility.zip

[TheBloke]

Thanks.  I have put the file in place and run the utility (which looks useful - thanks) and rebooted.

 

However I can't see any difference in states.  I am still seeing x12, x22, x23, x24.  No x21, so I think I was wrong in thinking I saw x21 before.

 

In Windows I sometimes see states between x12 and x22, like 3610mhz (x19.)   Maybe your fix to IOPlatformPluginFamily depends on certain BIOS config or particular CPU? 

 

Is there a thread somewhere discussing changing IOPlatformPluginFamily?  I ran a diff of file Contents/PlugIns/ACPI_SMC_PlatformPlugin.kext/Contents/Resources/MacPro5_1.plist which I could see was a changed file, and there were quite a number of changes between the original one and your one, including some of the data.  How did you know what changes to make?

 

I suspect this is the main change related to SpeedStep? 

           <key>StepDataDict</key>

           <dict>

             <key>MacPro5,1</key>

-            <string>SP12</string>

-            <key>SP12</key>

+            <string>SP1</string>

+            <key>SP1</key>

             <data>

Apart from that difference, most of the other changes seem to be related to GPU, which I'm not sure if even applies on my CPU which has no onboard GPU.

 

Anyway if I can't get more states it's not a big deal. It would be nice to have more but it's not important.

 

In other news I have re-instated the OC recommended by Emilo in his guide.   And so far, no segmentation faults.   Before I had started with his guide but then made some further tweaks, eg to RAM, and maybe that is what gave me the instability and segmentation faults.  Using his settings exactly, so far it seems OK - I have compiled macvim many times without problems.

 

I will keep an eye on it and if the issues come back I will have to tune down the OC a little.

 

Phew - I am glad I didn't have to re-install the OS at least!  

[TheBloke]

Update on my OP issue:  it wasn't OC.  It was a bad DIMM   One of the two new DIMMs I installed last week.

 

Took me rather a long time to work this out, because I had changed so many things in a short space of time.   And I got thrown off because /var/log/system.log archives showed Segmentation Fault crashes before the date of the new RAM install.   But I guess that must have been unrelated - there were only a couple before the RAM install, and they were always in mds, so maybe that just has issues.

 

Anyway, one of my new DIMMs records more than 4000 errors in MemTest86.  I have got a refund from the seller and right now am running with 32GB instead of 40GB.   I will probably at some point get another one or two DIMMs to take me back to 40 or 48GB.

 

I wish I had run MemTest86 on the RAM before I installed it   But I have literally never seen errors in MemTest86 before, despite running it on dozens of DIMMs over the years.  So I got complacent and assumed RAM errors were too rare to bother checking for;  if it detected in the BIOS, it would be fine.  Clearly not

[PPCnostalgic]

Glad to know you solved, X58 mobos are still reliable boards and 1366 Xeons are still capable processors too, expecially those 6-Cores.

BTW I guess you're putting so much load on your system, if 32Gb Ram aren't enough you are dealing with heavy workload... what's your primary use of your Hackintosh?

[TheBloke]

Glad to know you solved, X58 mobos are still reliable boards and 1366 Xeons are still capable processors too, expecially those 6-Cores.

 

Yeah this is by far the best and longest lasting PC I have ever bought.  Purchased August 2010, and still going strong.  I've changed GPU, disks/SSDs, RAM, CPU, CPU cooling, fans, but it's still same mobo and case so it feels continous.  And one of the reasons I haven't upgraded (besides never since having the same amount of money to throw around as I did in 2010) is that it's been powerful and upgradeable enough that to replace it I would have to spend serious money.  And even then, many upgrade options feel like partial downgrades - eg I am currently using 33 PCIe lanes and many modern systems don't have more than 24.  That's partly because they're PCIe 3.0 not 2.0, but still it means I can't easily upgrade without either losing some functionality (either my LSI controller or 10GB ethernet) or else having to spend yet more money on specialist mobos with more integrated peripherals.  

 

Some of the Intel Extreme CPUs have more PCIe lanes, but they're ridiculous amounts of money.  There is now AMD Threadripper which looks like the best option I've seen in a while, with tons of cores and 40+ PCIe lanes at a not-completely-stupid price.  But it's still a lot of money, more than I've had at one time in a while.  (I just checked and the motherboard alone would be £300-£400!)

 

So yeah, I think X58/LGA1366 was a particularly good system - powerful enough to remain modern for many years (with OC and component upgrades), but also early enough to not lose expandability through going to PCIe 3.0 and having M2 and dozens of USB 3.0 ports and all the rest that eats up PCIe lanes on modern systems.

 

BTW I guess you're putting so much load on your system, if 32Gb Ram aren't enough you are dealing with heavy workload... what's your primary use of your Hackintosh?

 

I may well not need more than 32GB RAM.  Up until now 24GB has been fine.  My normal day-to-day usage in macOS is no higher than 16GB;  for example right now I'm using 15Gb (mostly Firefox.)  But the reason I want extra RAM in macOS is for VMs.  I have my Windows 10 SSD available to boot as a VMWare Fusion VM, and I prefer to be able to give it 8-12GB of RAM so it's nice and speedy.  I won't use it daily, but I do want it sometimes for programs I can't easily get in macOS, or for accessing USB peripherals with no macOS drivers.

 

As for use of my Hackintosh.. well, it's my only decent PC.  I don't have a modern laptop, and I rarely use my iPad.  So I use the desktop for everything.  A lot of browser usage of course, now in Firefox (migrated from Chrome recently).  Dozens of tabs, several windows, usually a full screen video playing at all times.  On top of that, a lot of iTerm 2 terminals - that was the main thing that got me coming back to macOS versus Windows.  I use those to manage remote servers and for script development and the like.   Also, periodic use of Photoshop and Premiere Pro for photo manipulation and video editing respectively.  And occasional use of Adobe Audition, or maybe Logic Pro X now I'm in macOS, for playing about with sound recording/editing.

 

I also play games from time to time, but that's one thing I definitely won't do in macOS.  I will reboot into Windows 10 for that.

 

But yeah it's VMs that encourages me to go beyond 24GB.  I think 24GB is fine for all my normal usage except when I want a VM running.  And most VMs I can run on my home server (which has 172GB), so it's really only the Windows 10 VM for running desktop apps that needs extra RAM in my Hack.

 

TLDR: I probably don't need more than 32GB.  But I will probably upgrade anyway, just to be safe and to have the most possible   Oh, plus of course extra RAM is always helpful for file system caching.

 

EDIT: Actually, I'm using even more RAM than I thought.  I forgot quite how RAM-hungry 64bit macOS is.  Right now I am using 23.5GB, because I have Premiere Pro open doing some video editing.  That's without any VM loaded.  Add a 10GB Win10 VM and yes I would need more than 32GB. I probably could have Win10 use only 6GB and it'd be fine for most things I'm likely to use it for.  But I don't like to skimp