I can confirm the issues described by Nicolas, since I ran into the same issues today:
CPU: Intel Core i7-6700
MB: ASRock H170M-ITX/DL (latest Bios)
RAM: Crucial CT2K16G4DFD8213 2x16GB DDR4-2133
SSD: Samsung 850 Evo - 500GB
PSU: Cooler Master V550
OS: Arch Linux
- no Overclocking -
I turned off my BOINC-Client (which worked at 100% cpu load) and the system froze immediately.
With all enabled cores, I can't boot into my OS anymore. 1-2sec after selecting the OS to start, the system freezes.
Then I tried to start the OS with the linux mainline kernel, which also freezes but provides Machine Check Exceptions (MCE). I analyzed them with mcelog:
What I tried
1) Memtest86+ passed without errors (3h)
2) I turned off particular UEFI-Settings in order to reproduce the error (namely: Turbo, Speedstep, HT, C-States, Number of Cores)
-> I discovered like Nicolas that with 1-3 enabled cores my system just boots fine and works after several hours at 100% cpu load still fine.
3) I updated my UEFI to the latest version and reset my UEFI settings
4) I tried the Intel Processor Diagnostic Tool which passed with no errors, strangely it detected a cpu frequency which is too high (4.5332 GHz): Report File
Our problems are strongly correlated (in impact and time) and not OS-dependent.
Interestingly we both use ASRock-Mainboards.
I think it's either the CPU (the MCE-logs point to it) or its the mainboard.
@Nicolas: Do you got details of your BSOD Messages? Maybe you can retrieve error details from the Windows Event Viewer.
Any help is appreciated :-)
Hello, thanks for posting.
So, the file below I put all the "error" and "critical level" that occurred yesterday, after the problem started.
I don't know exactly what I should look for.
Like you said, any help is welcome.
EDIT: I tried the Intel Processor Diagnostic Tool which passed with no errors too (only with 3 cores).
Hi NícolasFH and Skylimit,
Before doing the replacement of the processor I suggest you the following workaround; get into the motherboard BIOS and reset settings, make sure processor speed (Base frequency) is set to 4.0GHz, memory at 2133MHz. Save changes and work for a while. Let see if processor works better with default settings (no overclocking, 4 Cores and Hyper Threading enable)
If processor works better and you want to OC, Intel recommends to change only the processor multiplier in the BIOS.
Replace process, contact our support team
I am having a similar problem, my system spec.'s are:
Asus Sabertooth Z170S
CPU cooled by Corsair Hydro H115i
16G Corsair Dominator Platinum DDR4 3200
Samsung 950 Pro NVMe - M.2 512M
Corsair HX850i Power Supply
Win10 64 OS
EVGA GTX980 Hybrid Graphics
System ran without a hitch for a few weeks then the freezes started. Nothing regular time-wise but always while idling. No error messages, no BSoD's, just total lock. Sometimes not even the reset would work, had to perform a full shutdown. I have the latest drivers and bios from Asus for this board. First i tried removing unnecessary programs that came with mobo disk. Then I tried setting the uefi to optimized defaults because I was overclocking. Continued to get freezes. Next I ran Memtest86, version 4 first, after 1.5 hours the single CPU did fine. Then I ran version 6 with multiple cores: on Pass 1 / Test 5, Pass 41% and Test 74% ... freeze.
I did some research and found others in various forums having the exact same problem and ALL running Skylake CPU's. I hope this can be fixed between Intel and the mobo manufacturers with a bios update.
thanks for the replies.
I re-enabled all cores and the system worked ~24 hours. Immediately after calculations (parallelized, 100% work load) for my master thesis finished, the system froze again and I wasn't able to start it with all cores, receiving Machine Check Exceptions (MCE)..
I followed the advice of Mike C. Unfortunately, a BIOS reset has not solved the problem.
- Removed one memory module
- CMOS reset
- Reinstalled the CPU cooler and applied new thermal paste
The system worked for ~10 hours with all cores. Then, immediately after I've stopped my BOINC client. The system froze again, OS start with all cores not possible, receiving MCE's..
So, in my case, the system freezes reproducible, when the cpu load drops from 100% to zero.
If I'm using CPU Voltage Offset Mode with +50mV, I can boot my OS with all cores, but it's not stable.
In regards to your issue, I notice the application BOINC client is still under development and it could show an unstable behavior. The application is for testing purposes.
The processor is unlock; however, if processor is working with customized settings, the system could show BSOD or random flickering issues.
Hi Mike C,
it seems your BOINC client research was over-hasty. BOINC is developed by scientists of the University of California (Berkeley) since 2002 and there's a stable version, which I'm using. I've reproduced the described problem with my own application written in scala. Further, considering the Machine Check Exceptions, there is strong indication that it's an application-independent behavior. Even if an application/OS is responsible for system freezes, it should never has impacts for starting other independent Boot-Devices e.g. Memtest86+ (freezes with all cores, described by NícolasFH and Buzzbomber36) or Linux-Live-Images (currently tested by myself: freezes with all cores too).
I'm using the Non-K Version (see my specs) of the Intel i7-6700, which is not unlocked. I'm running the CPU with default BIOS settings. That's all I want: running the processor under its specification.
The evidence available indicates that either the CPU is faulty, the Mainboard is faulty or the BIOS is faulty.
Let me reformulate the questions concerning the topic:
- Can we change settings in order to use the promised four cores?
- Can we do tests / change settings to identify the faulty component?
- Can we run tools to give you further information?
- Is the problem known and, if so, are currently Intel engineers working on it?
Over 2-days without a freeze. Supposition: Originally the XMP configuration wasn't as stable as I presumed, thus the freezes. When I disabled OC, the DRAM voltage went to 1.2V. I noted this a couple days ago as I was roving through the bios settings. My particular memory sticks are mean't to be supplied with atleast 1.35 volts. After restoring the proper voltage to the DRAM, I have yet to see another freeze. Fingers crossed.
Using default processor settings the system should work with the 4 Cores. Now, new motherboards has the option to overclock the processor even if it is a unlock model. It is very important to verify processor base frequency in the BIOS. It should be 4.0GHz.
In regards to the memory supported by the memory controller embedded on the processor is the following:
DDR4 2133MHz 1.25v or DDR3L 13333/1600 1.35v
Intel has a tool to stress the processor; it will tell us if there is anything wrong with processor. Run the application alone, close the rest of applications.
Intel® Processor Diagnostic Tool v. 126.96.36.199
I ensured the proper base frequency in BIOS: It's 3.4GHz (like expected for my Non-K i7 6700).
In my OS under full load it's 3.7GHz, like expected for Base+Turbo.
Thanks for the hint with the memory voltage. My standard voltage for my DDR4 2133MHz memory was set to auto with 1.2V only. I changed it to 1.25V. Unfortunately, it doesn't solve my problem.
Here is again a recent IPDT-log (UEFI-Version, All Cores): http://pastebin.com/NwLTj6L3 -> The 30s "stress" test passed.
I've changed the PSU, but the failure remained.
Since I can only specify the number of actives cores and not the specific active ones in BIOS, I did some further tests:
I've deactivated HT and deactivated respectively one core via linux kernel configuration. The config with cores 0+1+2 is the only configuration, which is stable for handling cpu load drops. Additionally, I took a closer look at the machine check exceptions. It's always core 3, which is responsible for throwing the MCE's.
Core 3 is defective and I'm going to replace the CPU.