PC freezing under load..and its not the GPU

[Spud1]

Hi,

Looking for any advice/suggestions.

Basic symptoms are that seemingly randomly, when my PC is under heavy load e.g. gaming @ high settings, after a period of time (5-30 minutes) the system will freeze and be completely unresponsive, leading to a hard shutdown to get anything back up and running. This has been happening on and off for a while, and I initially assumed it was due to the hot summers. This was backed up in my head as I could run the system without any issues if I took the window off my case and pointed a fan at the case.

Sounds like an obvious overheating issue, right?

This week my RTX2080 arrived, and I had also ordered a new case to go with it (bequiet dark base 900 rev.2). So I spent all of last night completely rebuilding from scratch into the new case, making sure airflow was near perfect etc....and guess what - same symptoms!

Since then i've been checking everything I can - I added another fan (but temps *seem* ok to me), lowered memory timings, tried a mem test (not per stick yet, but thats my next task) made sure nothing is overclocked etc..but no joy.

Current specifications are:

BeQuiet dark base pro 900 rev.2 with 3x140mm and 1x160mm fan (2 intake at the front, 1 exhaust at the back 1 exhaust at the top) - all pointing in the right directions
Gigabyte GA-Z170-Gaming-K3 motherboard
Intel i7-6700K @ 4ghz (stock speeds) with Arctic Freezer 13 CO as a cooler
32GB Corsair vengence 2133 DDR4 memory
2x 7.2k SATA Hard drives
3X SSDs
630W ThermalTake Smart SE PSU


I was running a temp monitor when it crashed last time - here is a screengrab of the temps etc at the point of crashing - https://imgur.com/a/pM066u5

I can get the crash to happen by running the Shadow of the Tomb Raider benchmark 4-6 times, so its repeatable.

Any thoughts/suggestions? My top two candidates for being a problem are the PSU (According to a calculator I need 520w..but thats close to 630w so..) and Memory, which I am about to go and test.

[Kumagoro]

After checking the usual and not getting anywhere try changing the bios battery for a new one. I have seen many a weird issue caused by a duff battery.

[Phage]

I'd be checking the PSU and ram sticks too.
I'm surprised your 6700k is at 100%... Is it OC'd ?

[Spud1]

I'd be checking the PSU and ram sticks too.
I'm surprised your 6700k is at 100%... Is it OC'd ?

100% is the "max" value - it does spike that far in DX12 games but the load at point of crash was ~50%. Nothing is overclocked (aside from whatever "stock"/factory overclock Nvidia may have applied to the GPU.

@Kumagoro good idea thanks, i have a spare and it certainly can't hurt!

I've not been able to find any problems with the memory stick individually I can run an extended test overnight but in the meantime I am going to wipe and re-install the PC onto a clean disk so i can rule out a dodgy driver or disk...
Replacing the PSU is the 2nd to worst case scenario (worst being the motherboard!) as re-cabling will take hours..and its another £100 or so chucked away..but giving myself until Sunday night before I bite the bullet and order one

edit: thinking about it some more, I wonder if the CPU/GPU both spike to 100% and that's when the crash actually happens..they just drop back as soon as that hits which is why the monitor shows much lower % utilisation as "current". Hmm more testing needed, but will do that rebuild first! I will update this thread as i go..

[kompukare]

edit: thinking about it some more, I wonder if the CPU/GPU both spike to 100% and that's when the crash actually happens..they just drop back as soon as that hits which is why the monitor shows much lower % utilisation as "current". Hmm more testing needed, but will do that rebuild first! I will update this thread as i go..

The obvious thing to do with a monitoring tool is to dump it on its own monitor unless you can get your repeatable crash to work withouta full screen game.
Alternatively, your monitoring tool might be able to output to CSV.
Power calculators tend to overdo it though. Would imagine that your actual max usage would be closer to 400W (250W for the GPU, 100W for the CPU, 50W for the rest) which is well below your PSU's rating.

[Spud1]

The obvious thing to do with a monitoring tool is to dump it on its own monitor unless you can get your repeatable crash to work withouta full screen game.
Alternatively, your monitoring tool might be able to output to CSV.
Power calculators tend to overdo it though. Would imagine that your actual max usage would be closer to 400W (250W for the GPU, 100W for the CPU, 50W for the rest) which is well below your PSU's rating.

thanks - yeah thats how I got that photo (left the monitoring on monitor #2 so it woudl show when it froze). Agreed on the PSU - unless its significantly under performing I should have room on it.

[Phage]

If it doesn't do it with the fan on it.... VRMs maybe ? I'd monitor those.

[kompukare]

thanks - yeah thats how I got that photo (left the monitoring on monitor #2 so it woudl show when it froze). Agreed on the PSU - unless its significantly under performing I should have room on it.

Okay, but Open Hardware Monitor also allows a plot: you have to turn it on on View>Show Plot and then check the items you want (it doesn't default to plotting everything as that would get too busy). Guess if you check CPU and GPU clocks, loads, and power.

Sometimes I've seen freezes from HDDs going bad too. Although those are usually momentary. Still no harm in freeing up one of your SSDs, disconnecting all other drives and doing your fresh install onto there.

Edit: although Phage's suggestion is better as I forgot you said a big fan stops it and HDD issues are unlikely to be heat related. And yes, OS/install issues are also unlikely to heat related.

[Spud1]

The problem *might* now be solved following my re-install. I wiped an SSD and did a clean install on to that, just installed the basics + steam, all windows & bios updates etc.
Just managed 2 hours straight on Tomb Raider - never got beyond 30-40 minutes earlier!

I've changed so much today that it's tough to say what's solved it, but I have noticed one of the motherboard temps ("temperature #3" in the image in my OP) has dropped significantly under load from 70 to 50 degrees, which may be related?

I will need to keep an eye on it for now, put it through some more stress tomorrow (going to run memtest overnight too), see if it holds up. Thanks for the advice/suggestions so far!

[Zak33]

even though you may have solve it.. I suggest you start with running Memtest so you know that the ram, motherboard and cpu work perfectly together on all addressable areas of that ram.

[philehidiot]

For what you've said, I suspect this was a driver issue possibly caused by installing immature drivers buuuuuuuut if it recurs I would suspect that PSU as being potentially the culprit. The max output is just that and not necessarily indicative of sustainable output, it might be just on the limit at cold and the riding temps cause it to tip over. That does seem a lot of spec for a 630W PSU. I run something similar and calculated my peak load at just under 620W.

My suggestion would be to bring down the power consumption and retry - easily done by unplugging some HDDs, SSDs, peripherals, etc.

I switched from Nvidia to AMD and, despite following the usual driver removal procedure that has served me just fine since pre-Voodoo3, Windows 10 just HAD to get involved mid way through install and make a hash of it.

[hal1000]

It's a long shot - but try seeing if your getting any Latency issues (they can build up over time especially in Windows 10 machines) - google LatencyMon (it's a free app) and super easy to use and will tell you a lot interesting stuff about what is going on with your PC etc.

[Jonj1611]

Why are you answering posts from years ago?

[Spank]

Free shipping would be my guess

[WithTemper]

Some kind of app advertising I guess

[wazzickle]

Free shipping plus competition entry is valuable