Help me diagnose a computer crash?

[tuffluck]

i built a new computer about 3 months ago. it was working just fine until recently. from the get go i OC'd the CPU and Vid card without problems. crashes are now showing up which makes me think the problem is not the OCs, but that could be totally incorrect. the types of crashes i have experienced in the last week are 4 in total:

1. crash watching youtube...sound card makes a loud buzz, computer crashes/restarts
2. crash while playing BF3, same buzz then PC restarts (this happened 1 time before the last week too)
3. the two weirdest ones happened 2 out of the last 3 times i shut the PC down. it sits on for a few days and i play games/browse without problems, then when i go to the windows logo and click "shut down" it shuts down...but in a crash form. then it restarts and then i get the "open windows in safe mode?" screen before i get back in and can actually shut it down properly

are these crashes indicative of an unstable OC? one time while playing with OCs i got a crash that resulted in a reboot, but i got the bios message "we have changed your config because your OC was unstable," and i have gotten no such messages in the last week of failures. i'm starting to wonder if the problem is my motherboard or my PSU. can anyone help me? again, all components brand new and purchased just 3 months ago. here is the equipment:

1. CPU - i5-2500k 4.0ghz @ 1.22v
2. MOBO - Gigabyte UD3H
3. RAM - G.Skill Ripjaws 8GB DDR3 1600
4. PSU - OCZ ModXStream Pro 700w
5. Vid - Radeon Sapphire 7850 OC 1200/1450 @ 1.176v
6. SSD - OCZ 60GB
7. HDD - Seagate 7200 250GB

also, is there any type of software i can run to test motherboard, PSU, and RAM for errors? crashes are tough, but i'm really hoping i can isolate where the problem is. just FYI, OCs passed stability tests and all temps look perfectly normal during idle and load. none of the crashes produce BSODs either. thanks for the help.

 

[Pinhedd]

These are all indicators of unstable overclocks. Overclocking is dangerous and should never be considered a "free speed boost".

Stability tests don't really mean a whole lot in the long term because overclocked components have less timing headroom and degrade much quicker. A stable overclock will not necessarily be stable 3 months later.

Please run Memtest86 WITHOUT any setup overclocked components to test your memory for errors. CPU instability will usually result in a BSOD, memory or GPU instability will usually result in a hard freeze. This is just correlation, it is entirely possible for any component to freeze or cause a BSOD.

Then, please run Prime95 WITHOUT any overclocked components (in setup or application) in blend mode to test your system for general stability. If it cannot pass 24 hours it is not stable and should not be overclocked. Intel does not warranty desktop processors to be completely free of errors over prolonged periods at high loads.

 

[pauls3743]

What power is the PSU?
How many graphics cards?
I would still update the firmware on the SSD. I had a pair of OCZ Vertex 3's running in a server, for three months, that I built, without issue. Swapped them into another machine,as I'd never intended to install them into the server, and they crashed out after a couple of hours running. I'm still testing them but I am happy so far

 

[Flareside]

Pinhedds got it, run a stability test without anything overclocked for at least 24 hours to ensure the parts are not failing at that load before OC. Is there a reason you are wanting to OC? It looks like your system should be able to handle most tasks without any problems?

Flare

 

[tuffluck]

so i ran memtest and within 30 minutes it got 35,000 errors...

i tried just one stick in both dimms, no problems. tried the other stick and immediately got the same 35,000 errors in under 3 minutes. so i think that must be the problem, right? RAM was never OC'd...

i will RMA it, but i need the PC next week so will wait at least a week to RMA it. is it okay to run on just one stick in the meantime?

 

[Flareside]

I would RMA it now as you can run on a single stick without any issues. Just another thought, just because your RAM wasn't OC'd it still might have been effected by everything around it running faster and hotter.

Flare

 

[Pinhedd]

Unfortunately you will have a hard time RMAing something which has been overclocked and/or run out of spec. This is the risk that users take when overclocking, you void your warranty. Since the Sandybridge processors were not designed for or tested with 1600Mhz RAM (although they can usually do it effortlessly there are always a few edge cases) you will need to test it with 1333Mhz DRAM at JEDEC timings. Your GSkill Ripjaw RAM with non-JEDEC timings will most likely not convince anyone at Intel or Gigabyte that their products are faulty especially when they were overclocked.


Good luck

 

[Flareside]

From what he said the RAM was never OC'd and appears to be the culprit so He should be able to return it I would think.

Flare

 

[Pinhedd]

From what he said the RAM was never OC'd and appears to be the culprit so He should be able to return it I would think.

Flare

Not necessarily. Memtest86 will detect errors, it cannot tell you where in the subsystem those errors originate. If there are a few errors then this usually points to a few bad memory blocks that will always manifest in the same address range. If there are a ton of random or sequential errors then this usually points to timing problems or controller failure. I've had similar results before myself, lots of errors but it turned out that I killed the memory controller on 3 motherboards due to NVidia's shitty northbridge. The DRAM from that fiasco is still in use today.

OP's methodology is good and certainly lends weight to it being one dead stick and nothing else but it's far from conclusive enough to warrant an RMA right away. At the very least it should be tested in another machine that hasn't been stressed