My newly built PC crashes randomly a couples of times a day. Before I condemn and return my i5-13600K CPU, is there anything else I should check first? – see list below.
I’m reluctant to blame the CPU as it’s never been overclocked, the Intel microcode update BIOS was installed straight after the build and before Windows 11 was installed, and the CPU does not overheat or crash when running Prime95 for half an hour. Crashes don’t seem to occur when the system is working hard, but often occur when the PC is left idle with no apps running.
Specs, (NB All items are new):
Intel i5 13600K, Noctua NH-U125 cooler, Gigabyte Z790 UD motherboard, 2 x Kingston Fury Beast 8GB 5600MT/s RAM, Samsung 870 QVO 1TB SSD, CiT FX Pro 80 Bronze 700W PSU, Corsair 4000D case.
This is what I have tried to pinpoint the fault without success. Have I missed anything?
• Latest motherboard BIOS installed before Windows installed (F12e)
• Remove Graphics card and use CPU graphics. No other PCI cards installed.
• No clues in the Windows System Event log, which often contains zero entries for the half hour before a crash if the system is idle. After the reboot you get an entry ‘The previous system shutdown … was unexpected’.
• Run Prime95 several times. HWiNFO says; Peak core temp 89C (11C to TjMax), max core VIDs are 1.460V, no thermal throttling or critical temperatures, CPU is power throttling for IA and Ring Limit reasons.
• CPU cooler removed, socket checked for damage, CPU re-seated with new paste. Cooler refitted.
• Case shaken (while cooler/CPU removed) in case of debris. None found.
• CPU passes the Intel Processor Diagnostic tool tests.
• MemTest86 run twice with zero errors.
• Main SSD passes an extended SMART test.
• Windows 11 shows no corrupt system files after using the DISM.exe and ‘sfc /scannow’ commands
• Mains cable changed and an UPS used to check to for power line issues.
• SATA cables swapped. PSU cables unplugged and re-plugged.
DDR seems to be overclocked from the supported 4400 to 5600, DDR voltages and timings are unclear, Prime95 coverage is unclear, and (last I checked) MemTest86 only covered the first eight cores. So Prime95 changes, y-cruncher, OCCT, or MemTest86+ might uncover something.
CiT doesn’t list with Cybenetics, I’m not finding anything else about the PSU, and its age is unclear.
Many thanks for the suggestions. I haven’t touched the RAM settings, but will double check the BIOS settings just in case. The PSU is new, but a cheap brand. However, It’s rated at 700W and I only hit 200-230W when running the Prime95 test as I’ve removed the graphics card for the time being.
Thanks again for your help,
This is what I was experiencing as well. My 14900K would be basically 100% stable when the system was under gaming or more extreme heavy load. But when only 1 application (in my experience Thunderbird e-mail client most often) was boosting only 1 core to its highest clock, the system would BSOD regularly. It was essentially repeatable.
Install Intel XTU or head into the BIOS and limit the P-core multipliers (I was dropping mine down to 52x or 53x by the time I RMA’d it late August) so the first few cores never exceed the All Core multiplier and see how the system behaves for a few days to a week. If it’s 100% stable with the new settings, you likely have a small bit of damage already to one of the P-cores. You can try disabling or slowing down the E-cores too. Mine continued to get worse for about 6 months while I waited for new BIOS versions to become available before RMAing.
But by that point my system had crashed enough times and the CPU had corrupted enough data on the SSD that I couldn’t boot Windows anymore, and I had swapped out to a 12900K and reinstalled. If you can run it, their utility may give you some useful information.
That’s a good thought. I’ve been running y-cruncher for 30mins with no errors and sensible temps/core VIDs (84C/1.381V), so it looks like it’s stable under load. Thanks for the suggestion.
That’s helpful to know that others have experienced their system being ‘liable to crash when idle, but stable under load’. Intel XTU shows that my P-core multiplier is currently 52x and E-core is 39x, so I’ve left these alone. I’m not aiming for max performance - just a regular powerful PC.
NB When I open Thunderbird it does indeed max out all the P-core clocks to 5.2GHz.
I think we’re heading towards RMA’ing the CPU.
Thanks for your suggestions and shared experiences.
If this is KF556C40BBK2-16 40-40-40 1.25 V is the 5600 profile and I wouldn’t expect 40-39-39 1.1 to be stable. If you meant they’re running at 4800 an overclock validation’s still required as it’s not supported by the 13600K on 2SPC boards.
Mine are KF556C40BBAK2-16 but have the same specs are the KF556C40BBK2-16. I will dig deeper using your advice and double check. Thanks for helping out.
Yeah, DIMMs’ RGB and non-RGB versions seem usually to differ only in the RGB module and its attachment. Kingston uses the same board drawing for both, too. Absent evidence otherwise, I tend to assume it’s the same DRAMs, PCB, and thermal pads going through the same binning, just with the RGB version running slightly warmer due to LED heat (if lit) and the closed top.
My KF556C40BBK2-16 pair are running at DDR5-4800 CL40-39-39 @1.1V. I’ve now tried the XMP1 profile (DDR5-5600 [email protected]) and XMP3 profile (DDR5-4800 [email protected]) but the PC crashes after a few hours in both cases.
Also could you please explain the term ‘2SPC boards’. Many thanks.
Two sockets per channel, in this case you have 4 memory sockets in a dual channel setup.
Officially, in this case the Z790 only supports 4400MT RAM speed.
I think it’s unlikely this is the reason for your crashes, though. Idle crashes I believe are more likely to occur from CPU instability, rapidly clocking up from idle to some background request. But it’s good practice to run everything at official spec when debugging issues so it doesn’t harm to try at 4400MT.
Another thing to try might be limiting power limits to something really conservative (65W?), or fully disable turbo boost to see what happens then. If it’s stable at base clock/lower power limit, but not at normal power limits and default boosting, something is likely wrong with the CPU.
Thanks for all your help!
I’ve been a bit slow but finally understand that I’ve been optimistic on my chosen memory speed and 4400MT would be the safer figure for an i5 13600K.
Also, the CPU vendor’s tech support are strongly convinced that it’s not the CPU and suggest a fresh Windows install. So I’ve put Windows 10 (not 11) on a spare drive, and it’s been stable for 4hrs so far. Fingers crossed. PS The previous Win 11 was a fresh install. Will post an update tomorrow.
I was able to run ycruncher for weeks and yet minecraft would die after < 1 hour when a cpu started to degrade. it had to do with the cpu going in and out of high power states it seemed.
the “test” of repeatedly extracting nvidia drivers is a good windows-centric test, hilariously, moreso than ycruncher
Eh, Intel’s the one who decided to market their processors as supporting 5600 by redefining 2DPC to mean 2SPC in footnote 1 of Table 25. They got away with it for about a year and a half, during which pretty much everybody assumed Raptor had a more capable IMC than Zen 4, before the degradation issues increased awareness of the fine print.
As an aside, Intel does the same with Arrow (Table 49), where 6400’s supported only for 1SPC CUDIMMs with certain clock layouts and sizes, and the datasheet ducks out of indicating anything about 2R.
Good news is that the fresh Win 10 install ran for 19hrs with no crashes (used to be every 30mins). The slightly less good news is that the old Win 11 install then ran for 7hrs without a crash. But it then proceeded to crash twice while I was writing this update (it must be afraid of Wendell’s hard stare).
I suggest we close this topic while I do some extended testing and murmur incantations to the God of Windows. Thanks to everyone for their help. It’s been very useful!
Hi @Kiltrash
I was having the same issues. Mine worked fine when running with a heavier load but when it was in idle just randomly crahses.
This behavior didn’t occur before Intel’s desperate microcode updates.
After almost losing my mind and not having anyone to lend me a processor, as everyone around me is using AMD, I decided to check some settings in the BIOS. With loads of try and errors I disabled the C-States, and everything went back to normal.
I’m not sure if this is an issue related to low power consumption and my PSU, or if it’s a bug with this feature on Gigabyte motherboards.
I would give Windows 11 another chance after changing this setting. In my case I still need Windows for Adobe Lightroom and play games when I have some free time. Other than that, I’m almost always on Fedora, and those crashes also happened there.
Maybe Windows 10 handles power management differently, but that’s beyond my knowledge.