Mat3

Distinguished
Sep 21, 2011
3
0
18,510
I bought a new system a couple months ago and had the store put it together.
Phenom II 965
Radeon 6870
Antec 620 high current gamer power supply
Asus M5A88-V Evo board
8GB RAM 1600Mhz Ripjaws
Everything new, nothing over-clocked.

The are two problems, right from the day I brought it home, always when I play any games.

First, the computer will often lock up. It's an intermittent problem, completely random. Sometimes I'll play for 10 minutes and it freezes, other times I'll play for hours before a lockup, or sometimes it just works and doesn't freeeze at all. When it locks up, there is no sound and no response to the keyboard. I need to hold down the power button to turn it off and on again. Once I get it
rebooted, I can continue gaming as if nothing happened. There is nothing in the Windows Event Viewer to indicate what the problem is.

The other problem playing games, far less frequent but just as random, is that the computer will turn itself off. I can turn it back on without issues.

It's too random to be overheating, plus I've kept an eye on CPU, mobo and video card temps, nothing out of the ordinary there.

Some of the things I've tried were downclockig the memory to 1333 so that the memory controller would not be overclocked. I've also adjusted the RAM timings to match what is recomended for AMD CPUs. I've updated the BIOS a couple times, and everything else should have up-to-date drivers. I've also tried giving the CPU a slight voltage bump (from 1.4 to 1.42), no help there.

I've run various tests (Memtest, prime95, furmark, all three at the same time) but nothing seems to trigger a problem like a game does, but since it's random, who knows for sure. I didn't let these run for more than a couple hours, I suppose I could let them run longer...

At first I suspected faulty power supply. I decided to see what would happen if I downclocked the CPU and video card. CPU normally runs at 3.4 Ghz, I capped it at 2.7. The video card normally runs at 900, I capped it at 825. When running with these settings, I never encountered an issue. I had it set like this for over a week, so even though it's random, I'm confident it's stable with these settings.

Next I tried downclocking just the video card, but putting the CPU back to default speed. These settings were not stable, I still got lock ups.

When I switched it (CPU downclocked to 2.7 again and video card back to normal), the system was stable.

My conclusion from this was that I had a faulty CPU. I thought I could narrow it down to a particular core, so I started assigning game threads to particular cores with the CPU back to full speed.
The game I'm playing now is Star Wars: The Old Republic. It has two main threads, I can lock each one to a particular core. For example, one on core 0, the other on core 1, leaving the other cores idle. Unfortunately, I tried every core extensively, they all seem OK. The PC never seems to lock up like this. (If I don't set the affinity to particular cores, the threads just seem to bounce from core to core at random and the PC will eventually lock up.)

So.. any thoughts? Bad CPU? Maybe the L3 cache or something else outside the cores is not quite right? Any help is appreciated.

Thanks


April 4th update: It still hasn't been resolved but I'll keep this post updated in case it helps someone else. I can rule out the power supply, as I tried a different one in the system for a few days and I still had three freezes and one shutdown in that time. Before that I also tried adjusting the RAM voltage to 1.55 (auto is 1.5) with the RAM downclocked at the same time, that didn't seem to help. I am virtually certain it's the CPU or mobo, hopefully I'll find out which one soon. (AMD CPU and ASUS mobo combination seems rather popular for showing up in threads like this one.)

May 8th final update: The shop replaced the board and the cpu. They said the cpu had excess thermal paste and something about bent pins (?). So I guess I'll never know if the mobo or cpu was faulty, but it sounds like they f*** up when they built the system. It seems to be OK now though.
 

Isaiah4110

Distinguished
Jan 12, 2012
603
0
19,010
I got home with a bad CPU the first time I built a PC. I didn't experience any of that though, the system simply would not post. That build had an OEM processor (AMD). As a result I have only purchased Retail Boxed processors since then and have never had a problem. I have never experienced a bad processor causing those symptoms, but would your processor happen to be OEM?

When you ran memtest, did you run it overnight (to complete multiple passes) on the extended test settings?

Your PSU should be strong enough to handle those components, but do you happen to have a spare lying around (with at least the same output levels) that you could swap in to test?
 

Mat3

Distinguished
Sep 21, 2011
3
0
18,510


I think it was a retail box (all the components were purchased separately and then they put it together). I did run Memtest by itself overnight, but I had the CPU capped at 2.7Ghz when I did. At the time, I was just testing the memory and didn't want the cpu running at max the whole night.

I don't have any spare parts to swap in and out unfortunately.
 

Isaiah4110

Distinguished
Jan 12, 2012
603
0
19,010
I've never ran memtest myself, but I have run Windows Memory Diagnostic before (can create a bootable CD for it very easily) and had success identifying bad RAM chips.

At what store did you buy the parts and have them build it?

I would recommend disconnecting everything except the CPU, RAM and GPU, set each component to its native clock speed and run the extended memtest or Windows Memory Diagnostic overnight. There is a chance the memory only has the error with the processor in its native state. Worst case scenario, the computer freezes, and that narrows down which piece is causing the problem. Then you can take the PC back to the store, explain what is going on, and they should swap out whatever part it seems to be for you. If the problem recurs, then they should swap out another part and so forth and so on until you have all good parts.

If nothing happens while running memtest overnight then you should still consider taking the PC back and having them replace one or all of the parts that could be the problem. I would also think they might be able to test and determine which part it is exactly at the store. It does them no good to have to swap parts one at a time for you and keep returning the old parts to the manufacturer.

In any case, you definitely have something going wrong...

Lastly: Have you tried reformatting and reinstalling everything? It doesn't sound like a software problem (I don't know why the problem would only show up with the hardware in its stock settings and not at reduced settings) but if replacing hardware fails then it might be worth a shot in the end...
 
G

Guest

Guest
It's the RAM voltage. I had the same problem with Patriot memory. Crank the memory voltage to 1.65, or 1.66 and see what happens. 99% of the time the memory manufacturer will list the required voltage on the stick of memory on a label. Sometimes it's 1.5V (or states it is) but that just won't do. Up to 1.65V is fine. I sometimes go as high as 1.7V (some Patriot memory requires this.) So experiment a little. But do NOT go passed 1.70V! JEDEC states that memory modules must withstand up to 1.975 volts before incurring permanent damage, although they are not required to function correctly at that level.

Incorrect voltage is very hard to diagnose since even if the voltage is incorrect, often times the memory itself will pass testing with Memtest86, Memtest86+ and Windows Memory testing. Just trust me on this one, you are going to want to either up your voltage or swap your memory sticks and make sure you set the timings by hand. Or just run standard speed 1333 Kingston's (I've never heard of anyone having an issue with those.)
 

TRENDING THREADS