• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Unstable following CPU and motherboard upgrade

Linkex

Linkex

Linkex

Associate
Joined
4 May 2011
Posts
1,065
I picked up an i7-4790K 4.00GHz (Devil's Canyon) and a Gigabyte Z97X-UD5H motherboard. Running stock settings on everything. When I first installed it, it was really quite unstable with random crashes (In the form of immediately powering off and restarting). The crash could be reliably produced by running prime95. This would produce the crash immediately – click start, heartbeat pause, power off.

I was advised to update the BIOs, which I did, and this has improved matters. There are no more random crashes so far, and prime 95 will run for up to a couple of minutes, but then does still crash. Temps run around 25-35 idle, climb up to around 90-95 under full load, but seem stable at that point. It’s a retail cooler with some older thermal paste, so I think that’s about right? i.e I don’t think it’s a thermal cut off triggering. A friend suggested increasing the vcore voltage, but that feels odd to me – if I’m not running an overclock, shouldn’t it be stable at stock settings without uping the vcore? That said, the boards vcore is currently set to ‘auto’ – could that mean it’s changing the vcore dynamically and undervolting it, causing the instability? As you might have guessed, I’m new to fiddling with CPU settings – I tend to leave them stock.

Anyway, on to the questions:
Are the temps within acceptable range for a retail cooler on some thermal paste a couple of years old? 25-35 idle, up to 90-95 under full load? Could they cause a power-off if they are too high?
Is it likely that I need to increase the vcore? If so, would that be indicative of a fault, or just expected behavior?
If increasing the vcore is the right thing to do, how do I know what the default should be? The vcore comes with a list of options, I’m assuming I set it to whatever the default should be, test it with prime and step up if unstable until I find a stable option?
 
You shouldn't need to automatically increase Vcore if running at stock, no, it should just be plug and play.

Temps are within limits, though do seem on the high side under load, but Ive never run one with a stock cooler so Im not sure how they compare Im afraid! I would assume that the stock cooler is designed to run it to any temp within its 100C limit so 90C wouldn't be out of the realms of possibility.

Have you checked everything is connected/seated correctly? All fans, cables etc. And have you got another PSU to try?
 
cwgk91 said:
You shouldn't need to automatically increase Vcore if running at stock, no, it should just be plug and play.

Temps are within limits, though do seem on the high side under load, but Ive never run one with a stock cooler so Im not sure how they compare Im afraid! I would assume that the stock cooler is designed to run it to any temp within its 100C limit so 90C wouldn't be out of the realms of possibility.

Have you checked everything is connected/seated correctly? All fans, cables etc. And have you got another PSU to try?
Click to expand...

To clarify, the boards vcore is set to auto because that's what it came with, I didn't change it from that. All instability has occurred with it set to auto.

Ok, lets see. Have checked all seating multiple times. Reapplied the thermal paste twice. All fans are running, I've run memtest with no errors, and tried running individual RAM modules, swapping slots etc. I've replaced the PSU. Tried progressively removing parts until down to board, cpu, ssd, RAM and PSU. I think that's covered it, but is there anything else I should have tried?

Something else to mention, the RAM is Kingston Technology 8 GB 2400 MHz CL11 DDR3 HyperX Beast which is 1.6v. I thought that meant that it was rated for 1.6v but should be happy at 1.5v, and it indeed ran on my old i5 rig for many months with everything stock. Now I wonder, could that cause this?
 
I can't see it mentioned in your post - did you re-install Windows when installing the new mobo/cpu?
 
Doh, should have mentioned that. Yes, I re-installed 7 (Several times) and then moved to 8.1 to see if that helped (It didn't) :)
 
When you took off the heatsink what was the contact like? (ie how thick/thin was the spread paste). Also was the fan spinning up when testing?

With the ram concern, you could up the ddr voltage in the bios to 1.6.
Alternatively use cpuz to see ram timings and compare to bios set.
 
eddyr said:
When you took off the heatsink what was the contact like? (ie how thick/thin was the spread paste). Also was the fan spinning up when testing?
Click to expand...

Originally was the pad that comes with it. Second time it was too thin and the temps went up, second time, I've gone a bit thicker, probably too thick if anything, but this old paste is like bluetac in consistency. As mentioned, its sitting at 35 idle and not more than 95 under full load. My understanding was it starts to self throttle at 100, which tells me its not a thermal cut of, unless someone knows otherwise?

eddyr said:
With the ram concern, you could up the ddr voltage in the bios to 1.6.
Alternatively use cpuz to see ram timings and compare to bios set.
Click to expand...

I have updated the bios to set the ram to 1.6, but prime still crashes it. Is there any benefit of running the higher voltage, in terms of resolving this issue, or can I go back to 1.5?



As far as I can tell the settings are correct for the RAM module, at least according to this: http://www.kingston.com/datasheets/HX324C11T3K2_8.pdf - but I'm not 100% sure how everything map so I may be missing something.

I've checked all the power connections and everything looks solid.
 
Higher voltage will probably increase temps and you are already nearing the max temp for haswell, thats the problem.

Any way you can try a different cooler? Sounds like you need new paste at any rate.

have you tried any other benchmarks? Aida, realbench, etc?
 
cwgk91 said:
Higher voltage will probably increase temps and you are already nearing the max temp for haswell, thats the problem.

Any way you can try a different cooler? Sounds like you need new paste at any rate.
Click to expand...
Only by buying one (And yes, need new paste) which I'd rather not do at this stage - I'm not overclocking, at least not yet, so the retail cooler should be fine. Plus, because the power-offs aren't temp related, it wouldn't seem likely to fix the issue.

cwgk91 said:
have you tried any other benchmarks? Aida, realbench, etc?
Click to expand...

No, will download some now.
 
Tried Realbench. Went through the encoding, temps touched 100 but didn't go over so looks like throttling is working at least, but seemed stable. Then on the third test "OpenCL" the system locked up. That ones new, it just hung, rather than restarting itself. Didn't even realise it had hung for a few mins, but it was totally locked up
 
OpenCL works mainly on GPUs I think. Have you tried running it just using onboard graphics and disabling your graphics card?

Also can you tell us the Vcore while realbench is running the encoding section to get an idea of what the load voltage is?
 
Have been running with onboard gpu since I started removing components, perhaps that's why it hung? May try re installing the card and seeing what happens

I'm assuming that its the "VID" reading in hwmonitor that I'm looking for? If so, it seemed to hover around 1.218 - 1.220 for the most part. Min of 0.675 max of 1.242. If that's the wrong one, can you let me know which reading it is please?
 
Looking at the datasheet ram should be more than fine with 1.5v at 1600mhz unless a faulty. You could use memtest loaded on a usb drive to check for that (run overnight).

Not so familiar with hwmonitor but VID is the voltage the cpu informs the bios for a given freq. What the bios sends can differ if manually alterd or if for eg the bios guys decided to add a safety margin. Cpuz should report voltage being provided or alternatively try gigabyte tweaking software for the board (not sure but assume there must be?).

Temp does seem a tad excessive for stock clocks and volts.
Just throwing out ideas, you could try disabling c1/3/6 states and turbo and/or force constant voltage, see where that gets you.
 
Last edited:
Yeah, check CPUz while running and see what that says if possible! 1.22v with stock cooler could well give 90C temps while running benches, though anyone running one with stock cooler correct me!

Something definitely isn't working though if it still hangs. Try putting your gpu back in and see if that does anything different. Also, try pressing the 'load optimised defaults' button in the bios just in case you've changed something by accident? Long shot I know :D
 
tl;dr - Its fixed. For those wanting to know, here what happened:

Check CPU-z voltage: Done. Got similar result, 1.218 all the way through. Also to note that I ran cpu-z and hwmonitor side by side and they reported the exact same voltages all the way through, so beleive they are both querying the same thing

Try pressing the 'load optimised defaults' button in the bios - Done, see below

Temps - This is where things start to get interesting. So I replaced the thermal paste. The temp still hit 100, but only just, right at the end, where before it hit 100 by 30%. At this point, I tested the GPU bench, which still hung. As I was restarting anyway, it seemed a good time to try the bios defaults. Hit default and had a look through. Wait a second, whats this intel turbo boost thing (Which is enabled by default apparently)? Lets try disabling that. Reboot and run the benchmark again, and max temps are 76. So basically, the boost was automatically OCing the damn thing until it hit 100.

Anyway, I've now had prime 95 running a stress test for 5 minutes and still going. I know thats not much in normal terms, but is enough to give me confidence that the rebooting issue is fixed. The temps are capping at around 83-85, and I'm sure will drop further as the new paste beds in. The GPU test in realbench no longer hangs.

So, there are two points I would like to make:
1 - Thanks to everyone who contributed to this thread, your efforts are much appreciated. I'm sure the OC staff would have figured this out if I had returned it, so you've saved me both postage and embarrassment :)
2 - If you are having stability issues on a system and it has "Intel turbo boost" enabled, try turning it off.
 
You must log in or register to reply here.
Back
Top Bottom