Sign in with
Sign up | Sign in
Your question

System unstable? Faulty CPU or RAM?

Last response: in Systems
Share
January 7, 2013 6:20:37 AM

I've just built a new system and I'm at a point now where I'm not exactly sure if the system is stable, and if not, which component is responsible (coming from a state of definite instability).

Before I go into my processes and details, here's a quick rundown of the (relevant) components:

Mobo: ASRock b75m-dgs
CPU: intel celeron G530
RAM: http://www.mushkin.com/Memory/Silverline/996770.aspx

Now, the first test I ran was the latest memtest86+. I ran it for about 9 hours with 7 or so passes with no errors. At this point I thought I had perfectly good RAM (which I would come to read later that passing memtest might not be sufficient enough to say this).

Onto the second test, which was prime95 small FFTs (to test the CPU), which also passed with no problems - only ran for about 4 hours though (partly because I got impatient reading that blend test should prove to be a tougher test)

So lo and behold, the first time I run the p95 blendtest, it consistently fails within the first minute or so. At this point I was sort of lost because I thought I had successfully tested both the memory (memtest) and CPU (small FFTs) separately (albeit not as thorough on the CPU).

I was about to give up and make a thread asking for help until I did some digging/reading and double-checked some commonly wrong settings the BIOS auto-sets. And lo and behold, the mobo had set the timings on the RAM to 7-7-7-20 rather than the specified 9-9-9-24 (although I would later find out this didn't really matter much and was probably the motherboard compensating for the CPU not being able to handle the RAM at 1333 and running it at 1066 instead). At this point I thought I found the error and should pass the tests fine! And of course, nope! P95 blend test still failing consistently within the first minute or so.

At this point I was again about toe make a help thread, but figured I'd start fiddling around with voltages. Namely, I spent a lot of time fiddling around withe the DRAM and Vtt voltages. They were automatically set at 1.585V and ~1.08V, respectively, by the motherboard. Eventually I settled on 1.65V and ~1.2V, respectively. Take note that the RAM is specified for only 1.5V! Also keep in mind that I kept the loosened/specified timings of 9-9-9-24, even after realizing it probably wasn't error-producing factor.

So NOW I'm at the point where I can run p95 blend test for ~2 hours (got impatient/on a time constraint). I can also pass Intel Burn Test on standard + high settings for 10-20 passes. HOWEVER, these results are not consistent! At first it seemed like the pattern was I would fail these tests if I performed them RIGHT after rebooting the system. However, it eventually became evident that's probably because I had just finished/failed a previous stress test, so the system was still hot. Testing this heat-based conclusion, I would wait a while, then reboot, then run the tests immediately - and I would pass!

So it LOOKS like this would be a cooling problem right? But according to HD Tune, which I've been using to monitor the system during ALL the above-mentioned tests/passes, the CPU only ever reaches ~42-48 degrees celsius. I've also touched the RAM sticks (being paranoid about them being at 1.65V) and they were very cool as well. My hunch now is maybe it's the b75 chipset - because I could not keep my fingers on the heatsink for even a second because it was THAT hot. Oddly enough, HD Tune reports consistent temperatures of ~30 degrees celsius for the mobo sensors.

Sorry for the long post, but I figured giving a detailed description of my steps/processes would help in determining the problem!

So what do you guys think, is this system stable? And if not, what's at fault? CPU? RAM? B75 chipset? PSU?
a c 136 B Homebuilt system
a c 86 à CPUs
January 7, 2013 7:45:53 AM

Set the RAM to 1.5 volts , or you may be damaging the processor .

Set everything else to default speeds too


Start worrying if an actual program gives you a problem and not a synthetic stress test .
m
0
l
January 11, 2013 7:12:40 PM

Sorry for the late reply! Unfortunately, I had already run the system at the higher voltages for a couple of days. I was getting intermittent BSODs so I'm now running everything at the AUTO settings. However, I'm still getting intermittent BSODs under normal usage...what could be the problem? Have I damaged the CPU/internal memory controller already? Or could it still be faulty RAM?
m
0
l
January 23, 2013 7:12:53 AM

A quick update since I don't want to create a new thread: It was indeed the RAM in the end (one of the modules were defective)

Basically, all the module/slot combinations passed ~7hrs memtest, but only one of the modules (no matter which slot) would consistently fail prime95.

Changed out the RAM, voila - problem solved!
m
0
l
!