New build - suffer from random freezes

[mroek]

Hi,

First, just let me mention I am new here, and that I tend to keep my machines for rather long, so my previous build was 7 years ago (yep, you read that right). What this means is that I don't have much experience with the current generation of components, but I believe (at least I did when picking the components) that what I have built should be a good machine for my needs. And that is to have a solid, good performing workstation at a reasonable price. No gaming, but quite a bit of 3D CAD, and a bit of programming. And I tend to use virtual machines quite a bit for all kinds of testing.

So, here's a list of the components:

Asus Sabertooth Mark 1 Z170
i7 6700k
BeQuiet Pure Rock
32 GB (2 kits) of Corsair Vengeance CMK16GX4M.
GTX1050ti (Asus Expedition)
Intel 600P M.2 512 GB (mounted in the slot on the MB itself)
Seagate Firecuda 2TB 3.5" SSHD (ST2000DX002)
WD Black 1 TB (from the old computer)
EVGA Supernova 650 G2
Fractal Design Define R5 (using only the two stock fans it came with).

After putting it together (which was a week ago), the first I did was to update the bios to the latest version (3007). I have no intention of pushing this (obviously, with just a medium class CPU cooler), but I entered BIOS and changed the Ai overclock tuner to XMP to get the RAM up to speed. I didn't change much else in the BIOS, I mostly left stuff at Auto. I did enable the virtualization settings though (VT-x and VT-d).

Then I did a clean install of Win10 pro (on the SSD, obviously), and ran a number of tests and benchmarks (but not for any long time). I actually repeated this two more times (for various reasons, and I also wanted to try the Windows reset mechanism), but after the third clean install of Windows I decided it was ready to take over the duty, so at that point I removed the drives from my old machine, and temporarily copied the data over. One of the drives from the old machine went into the new machine permanently.

Then I started using the machine, and here's the problem: I have random freezes where it totally locks up. No BSOD, nothing in the event log, reset button doesn't work. Holding the power button for a few seconds turns it off, though. This has happened 4 or 5 times now, with no apparent cause. In fact, most of the freezes has been almost at idle, or at least at a light load. Those are the worst kind of problems, and makes me distrust the machine for doing anything valuable.

To investigate, I did play around a bit more with benchmarks, and I noticed that the CPU was given around 1.3V at load, so I tried setting it to adaptive 1.2V in the BIOS. This caused the max temp to drop during Prime 95 v 26.6 small FFTs, to somewhere just below 70 degrees on all cores (it would get closer to 80 degrees before). However, I still got a freeze some time later (not during any benchmarking), so I knew I still had an issue.

Then I tried removing Ai Suite 3 (which I had read could cause issues), and I had to really hunt down all the crap it leaves behind everywhere. I hate ****ty software withe a vengeance, how hard can it be to do a proper cleanup in the uninstaller?

That made me do some work with Q-fan in the bios, and that's also quite crappy, changing unrelated settings without warning, so to get things to stick I have to save and reboot back into bios quite often. I should also mention that trying to run the fan optimizer doesn't work. It gets to 99 or 100%, and then locks up. I even tried resetting the bios to factory defaults before running it, still doesn't work. I've read that's a known problem on other Asus boards as well, so I'm just doing it all manually.

I have also tried running Memtest86 over night, no errors.
I've run Intel Processor Diagnostic Tool, no errors.

The only time I've had a freeze during benchmarking was with OCCT (large data set), and after that happened, I went back into bios and reset everything to defaults, then changing back to XMP and the few other settings I had changed. At that point I noticed that when choosing the XMP profile, there's a box asking if you want the Intel stock settings or some Asus optimized settings. The first time I answered yes (to get Asus optimized settings), but this last time I chose no, to get the Intel stock settings. This was last night, and the computer has been running solid now for some 18 hours, but what I have noticed is that now it very seldom goes to turbo boost. Under load it stays at 4 GHz, under light load it sometimes goes to 4.2 GHz, but only very briefly. With the Asus optimized settings (choosing Yes after setting XMP) it went to 4.2 GHz under load, and stayed there. I also set the CPU voltage to 1.2 V manual (was 1.2 V adaptive).

I have just recently had it running OCCT for 3.5 hours (stopped it before posting this) without issues, so I kinda hope it has been fixed, but I don't like that it almost doesn't use turbo. That could be the key also, but if so, does it mean I really lost in the silicon lottery?

Even if it might be possible that the problem did go away, I'm posting it to have the opinions of more experienced guys in here. What more info would should I provide? I've attached the bios settings text file that did give me freezes (the current settings I will have to reboot and export to get). Anything obviously stupid in there?

Thanks for any insight you may provide. But please don't suggest reinstalling Windows, as I have already spent a lot of time transferring data and installing stuff I need from the old computer.

 

[ATMINSIDE]

Instead of running XMP for the RAM have you tried setting the timings manually?

 

[wingman99]

4.0GGHz is base clock 4.2GHz is intel Turbo with one core, then when you change the Bios settings Asus boosts all cores to 4.2GHz with any kind of load. If you still have a problem try running the memory at stock default 2133 to see if it is a memory problem first.

 

[mroek]

Instead of running XMP for the RAM have you tried setting the timings manually?

No, I haven't. You mean setting them to the same as the XMP profile says, just manually?

 

[mroek]

4.0GGHz is base clock 4.2GHz is intel Turbo with one core, then when you change the Bios settings Asus boosts all cores to 4.2GHz with any kind of load. If you still have a problem try running the memory at stock default 2133 to see if it is a memory problem first.

Ah, OK. So what I'm seeing now is actually normal, then? The computer has been running for 22 hours now, and no freezes (which is a new record), so it is perhaps possible that the Asus boosting of all cores was the culprit? Somehow that sounds weird, since these CPUs normally can overclock a lot higher than that.

Might be best to just leave it like this for a while, and see if the problem is indeed gone?

 

[wingman99]

Ah, OK. So what I'm seeing now is actually normal, then? The computer has been running for 22 hours now, and no freezes (which is a new record), so it is perhaps possible that the Asus boosting of all cores was the culprit? Somehow that sounds weird, since these CPUs normally can overclock a lot higher than that.

Might be best to just leave it like this for a while, and see if the problem is indeed gone?

The Asus motherboard boosting all the cores to 4.2GHz should not give a problem. What you are seeing with the Asus motherboard is normal.

I had a infrequent Freezing problem with all default settings with a skylake and I RMA the CPU to fix it.

I would just wait and see what happens.

 

[mroek]

The Asus motherboard boosting all the cores to 4.2GHz should not give a problem. What you are seeing with the Asus motherboard is normal.

I had a infrequent Freezing problem with all default settings with a skylake and I RMA the CPU to fix it.

I would just wait and see what happens.

And the CPU that you RMAd, did that also pass all relevant tests (like the IPDT)?

 

[ATMINSIDE]

No, I haven't. You mean setting them to the same as the XMP profile says, just manually?

Yes. Leave all the RAM settings default except the speed, timings listed on the sticks, and the voltage.

 

[wingman99]

And the CPU that you RMAd, did that also pass all relevant tests (like the IPDT)?

The Skylake CPU that I RMA'ed passed Prime95 blend and Realbench stress tests also MemTest86.

 

[mroek]

Thanks, both of you. It appears that Skylake freezing it more common than I thought, over at the Intel community forums there is more than one thread about it.
However, it seems impossible to know for certain if the CPU is bad or not. Some people report improvements if disabling C-states in the bios. In my case, I haven't had any more freezes since I started this thread, but I'm not convinced the issue is gone. Just for good measure, and after reading on the Intel forums, I'm going to take the CPU out and validate that all the pins are in perfect condition, and I will also remove the Asus installation tool that I used when I first installed it. Some people report that it might not be a good idea to use that, even though I don't really see what harm it could do.

 

[wingman99]

I tried all the C state off and the voltage adjustments in the end of 2 months of trying things I just RMA'ed the Processor. It was very rare that the skylake would Freeze that is why it took so long to try a CPU RMA every time thought it would work fine after a change it would Freeze in 2 weeks give or take.

In the future If I have a Freezing problem with a new rig I'm just going to RMA the processor after the basics are covered.

 

[mroek]

My issue doesn't seem to be resolved with disabling c-states either. I also removed the plastic Asus CPU tool and cleaned off all the old thermal paste (which was the preapplied paste on the cooler), and then reseated the CPU and remounted with a different paste (some older Arctic silver I had lying around).

Issue still present, so I decided to purchase the deluxe version of HCI memtest to really make sure my RAM isn't the culprit. Running it in Windows shows no errors, but somehow it seemed to accelerate the rate of freezing (which may well be a coincidence). Running HCI memtest natively (booting it) doesn't show errors either, but even then the computer has frozen, so it hasn't finished testing. One time it actually also threw a kernel panic (I assume it runs under a Linux kernel), so my issue most definitely isn't tied to Windows alone.

I have also tried setting the RAM timings manually as suggested by ATMINSIDE, but it doesn't seem to make a difference. I also played a bit with the VCCIO and VCCSA voltages, but still got freezes.

I guess I could try to RMA the CPU, but if this is a rather common problem, there's no guarantee that the new CPU is better. And I really can't say with certainty that the CPU is to blame, even though I believe the evidence points in that direction.

This is clearly the most annoying problem I have ever had on a new computer!

 

[trents]

My issue doesn't seem to be resolved with disabling c-states either. I also removed the plastic Asus CPU tool and cleaned off all the old thermal paste (which was the preapplied paste on the cooler), and then reseated the CPU and remounted with a different paste (some older Arctic silver I had lying around).

Issue still present, so I decided to purchase the deluxe version of HCI memtest to really make sure my RAM isn't the culprit. Running it in Windows shows no errors, but somehow it seemed to accelerate the rate of freezing (which may well be a coincidence). Running HCI memtest natively (booting it) doesn't show errors either, but even then the computer has frozen, so it hasn't finished testing. One time it actually also threw a kernel panic (I assume it runs under a Linux kernel), so my issue most definitely isn't tied to Windows alone.

I have also tried setting the RAM timings manually as suggested by ATMINSIDE, but it doesn't seem to make a difference. I also played a bit with the VCCIO and VCCSA voltages, but still got freezes.

I guess I could try to RMA the CPU, but if this is a rather common problem, there's no guarantee that the new CPU is better. And I really can't say with certainty that the CPU is to blame, even though I believe the evidence points in that direction.

This is clearly the most annoying problem I have ever had on a new computer!

Yes, but clearly not everyone with a Skylkake CPU is having this issue. Most are not. So the chances that you will get another bad one if you RMA are probably not that great. You've got to start somewhere. Likely it is either the CPU or the motherboard. I think I would be tempted to purchase a cheap celeron CPU just to test this.

 

[mroek]

Yes, but clearly not everyone with a Skylkake CPU is having this issue. Most are not. So the chances that you will get another bad one if you RMA are probably not that great. You've got to start somewhere. Likely it is either the CPU or the motherboard. I think I would be tempted to purchase a cheap celeron CPU just to test this.

That's something I hadn't thought of, but it is an interesting suggestion. A Celeron G3900 (or even a Kaby Lake G3930) isn't all that expensive, but also the rest of the system woul'd be less stressed with such a CPU, so it is perhaps still possible that the system wouldn't freeze with a Celeron and still freeze with an i7 if some other component is to blame. Or is that just very unlikely?

I guess this wouldnæt require any messing with my Windows installation, because I really want to avoid any reinstallation. Just swapping CPUs should be OK, right?

 

[t1nm4n]

Just to toss this out there, does it freeze whiel you are actively using the PC? I mean has it just locked up when you were moving the mouse or typing something? Could it be a sleep/hibernate issue?

 

[mroek]

Yes, it has frozen while being actively used. However, of all the freezes, the majority has been at low loads, but also some while stress testing. Currently I went back to XMP set in bios (but choosing the "Intel stock" settings), and left all voltages at auto (and c-states enabled again). Now it hasn't frozen for 5 hours, and it has been sitting at some load (10-20%) all the time.

My plan now is that for each new freeze, I'll change only one or two settings at a time, and then just wait for the next freeze, but making sure I always record which settings I had when it froze. In the end I don't really think this will be resolved until I try different components (and the prime suspect is still the CPU), but I'll have to try.

 

[wingman99]

A RMA with Intel is painless, they will even cross ship. Link: http://www.intel.com/content/www/us/en/support/contact-support.html#@14 If the memory won't help running stock 2133 speed no XMP, then I would RMA the CPU. Less than 1% out of all the PCs sold can make Intel specification alterations in bios. The PC should work just fine with optimized default settings, just like the rest of the Intel PCs in the world that run stock fine.

 

[mroek]

A RMA with Intel is painless, they will even cross ship. Link: http://www.intel.com/content/www/us/en/support/contact-support.html#@14 If the memory won't help running stock 2133 speed no XMP, then I would RMA the CPU. Less than 1% out of all the PCs sold can make Intel specification alterations in bios. The PC should work just fine with optimized default settings, just like the rest of the Intel PCs in the world that run stock fine.

Not sure if an RMA with Intel is as easy here in Norway, but I will look into it. What you write makes perfect sense, there is no reason it shouldn't work fine at default optimized settings, and I shouldn't have to spend countless hours chasing issues like this. If I had been trying to push limits then it would have been different, but I just want a good, reliable workhorse.

 

[trents]

That's something I hadn't thought of, but it is an interesting suggestion. A Celeron G3900 (or even a Kaby Lake G3930) isn't all that expensive, but also the rest of the system woul'd be less stressed with such a CPU, so it is perhaps still possible that the system wouldn't freeze with a Celeron and still freeze with an i7 if some other component is to blame. Or is that just very unlikely?

I guess this wouldnæt require any messing with my Windows installation, because I really want to avoid any reinstallation. Just swapping CPUs should be OK, right?

Should not mess with Windows, no. Since your motherboard and the motherboard chip set would remain unchanged it should not create a problem and should not cause you to have to re-activate the OS with Microsoft.

 

[mroek]

Should not mess with Windows, no. Since your motherboard and the motherboard chip set would remain unchanged it should not create a problem and should not cause you to have to re-activate the OS with Microsoft.

Thanks, as I had thought/hoped.

However, the machine has been running without issues for nearly 48 hours now (I haven't rebooted since my post #16 in this thread), and I have a suspicion that it has something to do with setting all the voltages back to auto. One other change I also made, was to disable sleep/spin down of the hard drives. Someone suggested that allowing this could cause freezes, so I disabled it. Whether it actually has any effect I don't know, and I'd need a lot more time to know if the issue has in fact disappeared, but for now I will not change anything until I get another freeze.

Setting the voltage(s) back to auto (except for the RAM, which is at 1.35V as per the XMP profile), I do see that the CPU voltage is jumping up and down as the CPU requests it, and it does get more than the 1.2 V that I had tried manually setting at some point earlier. While it might be a stretch, there is perhaps a slight possibility that my particular sample is starving at 1.2V under some loads, causing the freezes. Even at stock clock. At least that's what I'm hoping for...

Edit: With the CPU voltage at auto, it is requesting/getting 1.232 V when loaded with P95 v26.6 small FFTs.