Think my CPU has a strange fault

McGraw

Member
Oct 16, 2014
36
0
0
Hi.

I have a 4790k rig. One day it crashed and wouldn't boot back up so I sent the board back to be fixed.

It came back and was working but the bios kept resetting from power on (not restart).

Every component has been changed, and I mean everything, board, ram, psu etc. except the CPU and the problem remains. It's had multiple CMOS resets and BOS flashes including previous versions.

When powered on with XMP or previously stable overclock, the bios says unstable and must enter bios. If I go in there and put the settings back and restart, it boots fine and is rock stable.

Sometimes it happens at stock.

What could be happening? Should I RMA my CPU?

Thanks.
 

PliotronX

Diamond Member
Oct 17, 1999
8,883
107
106
I had a strange message with my 4670k about memory channel A failing at a previously stable clock and I had to reset CMOS to get it to boot until I figured out that the case cover was applying uneven pressure to the heatpipes of the huge Noctua D15 because I run it delidded (direct-die). Might want to double check the pins of the board(s) and clean the contacts on the CPU if contact is an issue (quite doubtful, but I've seen some weird sh** in my day).
 

McGraw

Member
Oct 16, 2014
36
0
0
It had a massive overclock of 4.5ghz at VID equivalent whilst running an XMP profile.

Fair enough, I'll suck it up.
 

know of fence

Senior member
May 28, 2009
555
2
71
Even when you don't change the (100 MHz) BCLK, RAM timings and CPU overclocks are linked, people loosen their RAM timings to reach stability with their overclocked CPU and vice versa: XMP profiles would work with stock CPU speeds (by no means is it 100% guaranteed) but not when you start messing with the CPU multipliers.

Maybe someone can speak to what is required to reach stability while overclocking both RAM and CPU. It is not a strange fault, just something that requires next level OC sophistication.
 

Sheep221

Golden Member
Oct 28, 2012
1,843
27
81
Even when you don't change the (100 MHz) BCLK, RAM timings and CPU overclocks are linked, people loosen their RAM timings to reach stability with their overclocked CPU and vice versa: XMP profiles would work with stock CPU speeds (by no means is it 100% guaranteed) but not when you start messing with the CPU multipliers.

Maybe someone can speak to what is required to reach stability while overclocking both RAM and CPU. It is not a strange fault, just something that requires next level OC sophistication.

I think his mem controller is screwed up as a result of high memory overclock and multiplier overclock, if you do OC the memory you also increase the voltage, that could damage the mem controller on-die.
 

PliotronX

Diamond Member
Oct 17, 1999
8,883
107
106
He would have been having to pump around 2V into it for quite a while to affect the MC; Haswell and DCs MC are much more resilient than past generations. There isn't a 1.65v hardline for affecting its life.
 

McGraw

Member
Oct 16, 2014
36
0
0
Thanks for the replies.

It's been at 1.224 vcore and 1.66 on the ram (as per the 2400mhz XMP)

I just don't understand how it can boot fine and be completely stable from a restart but appear to be unstable to the motherboard from a cold boot.
 
May 11, 2008
19,303
1,129
126
Thanks for the replies.

It's been at 1.224 vcore and 1.66 on the ram (as per the 2400mhz XMP)

I just don't understand how it can boot fine and be completely stable from a restart but appear to be unstable to the motherboard from a cold boot.

Maybe temperature ?
When you have it running for a while and do a restart, temps are perhaps higher than from a cold boot.
 

wingman04

Senior member
May 12, 2016
393
12
51
Try not using the XMP setting then you might know if it is memory controller. If it works ok then do a mem86 test with XMP. After all that you might have to replace your CPU.
 
Last edited:

McGraw

Member
Oct 16, 2014
36
0
0
It passes every stress test I throw at it when oc'd and the intel tool above comes back fine.

The BIOS complains at startup with xmp/no oc, oc/no xmp and both (and neither the first time it starts up after resetting to optimised).

Fortunately the bios settings save so all I have to do now to turn my pc on is wait for it to turn on then turn off on its own about 5 secs later, choose enter bios then click exit.

Think I can live with that for now.
 
Last edited:

wingman04

Senior member
May 12, 2016
393
12
51
Sounds like a board problem, What board is it? Have you tried to flash to a new Bios?
 

McGraw

Member
Oct 16, 2014
36
0
0
I've had the board replaced, tried different batteries and multiple BIOS flashes/resets.

I'm starting to think that it could be an inherent board fault (I'm reluctant to name the manufacturer as they have been so good with their support and it could still be my CPU) based on some other forum posts I've found so will be taking it to a local shop so they can try my cpu in their board and vice versa.

Going to try a beta bios I've managed to find too.
 

know of fence

Senior member
May 28, 2009
555
2
71
Did you select RAM from the motherboard compatibility list, just curious.

Maybe enabling hibernation/suspend-to-drive from the power options will work as a workaround, instead of shut-down.

Is there a way for you to reach stable/ as intended operation and work gradually from there.

Setting things in BIOS doesn't mean jack, software and OS can override pretty much anything. So that's in my experience where most of the strangeness comes from. When you say the system is rock stable, you also need to mention that it's also running at whatever clock or frequency it was set to.
 

Flapdrol1337

Golden Member
May 21, 2014
1,677
93
91
RMA will be pointless, intel will test it at stock with memory at 1600 and conclude it's fine.

Did you update the bios to the latest version?
 

McGraw

Member
Oct 16, 2014
36
0
0
did you select ram from the motherboard compatibility list, just curious. - no, it has worked fine for 18 months and two other sets of ram do the same.

maybe enabling hibernation/suspend-to-drive from the power options will work as a workaround, instead of shut-down. - i'm hoping i can just use sleep from now on but it won't resume from sleep at the moment and i'm not sure if this is becuase my windows install appears to be corrupted (it takes about a minute to boot since the new board). Sleep was a problem with the first board, it would sometimes resume and lose the sys fan settings which control my h60 pump.

is there a way for you to reach stable/ as intended operation and work gradually from there. - i do need to try multiplier and voltage changes independently. Even the slower xmp causes boot complaint.

setting things in bios doesn't mean jack, software and os can override pretty much anything. So that's in my experience where most of the strangeness comes from. When you say the system is rock stable, you also need to mention that it's also running at whatever clock or frequency it was set to.

-cpu-z, hwid and aida all confirm vcore, ram clocks and cpu clocks to be what i expect. In otherwords, it is definitely overclocked and stable.

.

rma will be pointless, intel will test it at stock with memory at 1600 and conclude it's fine.

-maybe not if it won't resume from sleep, it won't boot at 1600 either, only 1333.

did you update the bios to the latest version?

-yes running latest and have tried earlier.
 

john3850

Golden Member
Oct 19, 2002
1,436
21
81
Many of the smarter people here will not trust a RMA MB and buy a new mb to be safe then they sell the RMA MB.
I believe your problem is memory related or even a bent socket pin.
 

McGraw

Member
Oct 16, 2014
36
0
0
It's working I am so happy!

Beta BIOS sorted it.

It now cold boots fine and much faster.

Windows starts in two seconds again and it resumes from sleep!

Thanks everyone for your replies.