System freeze with prime 95

[blacknemesist]

Hi.

My newly acquired system worked well for 3 days but then something happend!
Suddenly it started freezing(no shutdown,no BSOD) on facebook.It happend randomly.Then it started freezing during you tube videos.Yesterday it freezed with prime 95 not 5 seconds after I started it.
The CPU is running stock until I have my Case fans.

Here is my system:

- i5-3570k @3.4ghz turbo 3.8ghz
- ASrock Z77 extreme 4
- Noctua D14 SE2011 with the PWM fans and 1155 socket backplate
- NZXT Phantom 410
- XFX XXX 750w
- MSI gtx 670 power edition OC
- 2 x 4gb G.skill 1600mhz sniper low voltage(run at 1.25v)
- HDD sata 3 samsung 2 TB
- Windows 7 SP1

I ran Prime 95 and MSI Kombustor when I got the computer and the CPU was getting in the mid 50's and the GPU was about the same.
A few days went by and with them the freezes.I didnt flash the BIOS and was using the 310.70 GPU drivers.Didnt install no software either.
After the first freeze I ran memtest86+ and passed.

All I can think about is that either the PSU or Mobo have gone bad.Since the system barely pulls any power I am guessing the mobo was defective to begin with and is not feeding the CPU the power it needs.
On the other hand,the PSU could also be the problem.

I sent the machine to the store so they can figure it out since they were the ones who assemble it and assume full responsability for malfunctioning hardware for 2 years(thats why I had them built it in the first place).

The Cpu was also acting strangely.The load on core 0 was was 10-15% more than the rest.The temps on idle were also ~15º higher than the rest,usually 30-32º against 17-20 on the other cores.

Any ideas on the possible cause for this?
Also the system was close to the wall socket.I have 2x 5 plugs connector on it,one connected to the other so 9 available power sockets.My old machine never complained and I had a 700w PSU on it.

Thanks.

 

[C12Friedman]

It is not unusual for a CPU to show a minor temperature spread between cores but a 15C difference (we are talking C here right?) is pretty extreme. To me it sounds like either a very low binned CPU or possibly even defective. It could also be driver issues, hopefully the shop will get it figured out and let you know what had to be done.
My $0.02

 

[blacknemesist]

Thanks for the quick reply.

It is in celsius.The ~15ºC difference is between the core 0 and core 3 and is was not constant but the core 0 was always 10ºC hotter than the other cores at downclocked 16x100 =1600 mhz idle.When it went to 3.4-3.8 the temps became closer.

Could it be a mobo/psu problem since it jumps from 16x to 38x instantly during prime 95 and maybe the mobo is not delivering enough voltagge to the cpu?I wish I had ran MSI afterburner yesterday when the instant freezes began to rule out the PSU.

What about the wall socket being shared with a 21'' monitor,router-modem,2 external hard drives and a phone?Maybe there wasnt enough power going into the PSU?Although it worked at 45x prime 95 and kombuster simultaneously during the first day...

Could the case being close to the wall socket cause any problems?The wall socket has the power connectors in and it is made of rubber.

I'm still betting on the mobo or the psu.
The you tube issue can be drivers or the OS as i've seen that the issue happens in alot of systems and is caused by flash software or the gpu drivers.

Still,freezing the system with a sound loop strikes me more as a mobo problem since im using the mobo integrated audio.
My laptop with the same windows 7 SP1 and same GPU drivers doesnt crash.Browser is the same,flash version is the same,it has to be the mobo.During youtube videos the CPU barely works,it goes to 16x because it doesnt need anything else and the GPU was stressed with kombuster after the first few crashes.

I really hope it is the mobo.

 

[gamerk316]

Boy, these are the issues that are HARD to diagnose. Typically, hard lockups points to a mobo/RAM related problem. Make sure the RAM is configured right. If so, try a small voltage bump to the RAM/NB and see if that helps. Failing that, pull two sticks.

If a voltage bump + two fewer sticks (I'm assuming you have 4) doesn't help, then I'd suspect the mobo.

 

[blacknemesist]

The machine is still on the store and they are testing both the GPU and PSU.
Again,very odd they are looking on the GPU since to my knowledge prime 95 doesnt stress the GPU one bit.The PSU however can be the cause since hard hangups point towards power failure aswell.
What I can't seem to put my finger on is why was I able to stress the CPU and GPU at the same time and everything went fine.

As far as I can see the mobo is the central problem and may be giving off false negatives on other components.They are testing with another GPU right now to see if the problem continues but if it is really the mobo it may not present the same symptoms unless they are using the same GPU model that is in the machine.My fear is that unless they go with something equal to my GPU this problem may resurface at any time.

I know they will get it right and the store covers 2 years of warranty on store plus i can RMA the parts that have manufacture warranty after that but I don't want to have the machine going back and forth to the store because what seems to be a gpu/psu issue can be in fact a mobo going rogue.

If it was me I would test the PSU mobo cpu and gpu on different stable machines to be sure but I understand how expensive and downright impossible it can be to do so for every setup that cames through the door.

Hoping it is the PSU and that the damage is isolated on it and have not messed up with the mobo because that would mean that at any given time any other components can fail.

Still as you 2 said,it is hard to trace something back like that.
My older machine which I sold at one point fried both the GPU and one ram module and from my almost 7 years experience with it it has fried at least 3 ram modules but the mobo was never to blame.Thinking back on that,the mobo could very well be putting extra pressure on the ram.

Since my CPU shifts multipliers from 16x to 38x would the ratio between ram and cpu fsb be changing aswell?Normally the ram works at 800mhz on bios but when it jumps to 38x the mobo is probably setting the ram to 1333 or 1600mhz at the same time.I think it does since achiving 1:1 is perfect I wouldnt be surprised that the mobo tries to do that when the cpu multiplier goes up.Does it make sense for a mobo to do it?

 

[gamerk316]

My opinion is that its VERY unlikely to be the GPU. If the display driver were to hang, for instance, you'd get a BSOD (either 0x116 or 0x117). Motherboard or PSU seems more likely. My bets still on the mobo though.

 

[blacknemesist]

Thats what I think aswell,with the GPU experience I had I got either BSOD or plain computer shutdown.
Following the mobo logic,if they found something that could be affecting system stability coming from the GPU it can also mean that it can be the GPU,PSU,RAM and mobo,any of those or a combination of them.
As I said the system freezes shut without any call for the GPU,prime 95 blend stresses the cpu and ram but leaves the gpu alone.Also it crash exactly when I started the test which for me means that the problem is tied to the CPU,RAM,mobo or PSU.But why then would it freeze at idle?Better yet with not freeze when I was installing windows and using a % of the cpu?To my knowledge a cpu problem or system instability or errors would be pretty clear from the randomness of crashes but since it only crashes on instant 100% load or going from no load to some load(youtube streaming the video,maybe the igpu kicking in) it seems more like the 16x to 38x multiplier jump and the extra power demand is causing the crashes.
So PSU or mobo.
PSU is next on the list to test.

 

[blacknemesist]

Best answer selected by blacknemesist.

 

[blacknemesist]

Got word from the store and my GPU is fine.
They are now testing my HDD for problems so I figure they already tested the CPU and PSU.
Just hoping is not an HDD problem,700gb of data trashed doesnt look good.

 

[blacknemesist]

Dammit,it seems it was ram all along.Guessing the 1.25v don't work very well with the mobo.The store is testing the new ram modules(they tested everything and all passed) and should be an answer soon.If the problem happens again im going for regular sniper modules.