Computer randomly shutting off

[firebirdude]

My computer specs are in the sig.

It randomly just shuts off every now and again. Maybe once or twice a week. I don't remember this happening until I installed the second set of RAM (2x2GB Kingston) about a month or so ago. But my troubleshooting spidey sense tells me it's more of a power supply fault. The computer is positioned behind a bunch of junk. So it's kind of a pain in the neck to remove it, open it up, remove the new ram, and get it back together to test. Even then, the fault only occasionally shows face. So I thought I'd get some of you guys thoughts first.

 

[Patrick Wolf]

Probably the PSU. I've never mixed memory before so I don't know what issues that could cause. Are all the sticks rated at the same voltage?

 

[fffblackmage]

You could test with the ram with memtest 86+.

Does Windows give you an error message or memory dump? Those might help in identifying the issue.

 

[firebirdude]

Does Windows give you an error message or memory dump? Those might help in identifying the issue.

Nope. Just typing along and boom. Everything goes dead. Computer doesn't turn itself back on at all.

Going to run Memtest now and will post up results.

 

[firebirdude]

Ok. Hour long full Memtest complete. It recognized all 4 sticks and I had no failures.

Some info on the RAM:
OCZ OCZ2N10662GK DDR2-1066 2.1V-2.3V 5-5-5-15
Kingston KHX8500D2T1K2/4G DDR2-1066 2.2V 5-5-5-15

According to memtest, here are the settings set in bios:
CPU FSB 277
RAM 346Mhz(DDR 693)
Cas 5-5-5-15

I know a little bit about these numbers, but I get a little confused on how the RAM speeds move in relevance to FSB setting... then you throw in dual channel settings and I'm lost. Does the DDR 693 mean my ram is severely underclocked? Or because it's in dual channel mode, it's actually double that and overclocked?

And most importantly, you guys think any of this is my problem?

 

[fffblackmage]

Bah, memtest is kinda weird when reporting numbers. I've seen it report different timings from what I manually set in BIOS.

Are those numbers of your's consistent with what's in BIOS?

Dual channel has nothing to do with the memory's clock frequency. The ram does appear underclocked though. Either memtest is reporting some weird numbers or you just didn't manually specify the memory speed in BIOS. I don't believe DDR2-1066 is an official standard from JEDEC, so the ram is probably trying to run at a standard DDR2-666.

 

[firebirdude]

CPU-Z shows:

Is it just me? Or does it look like everything is underclocked?

 

[fffblackmage]

The CPU is probably fine; That's just SpeedStep at work (underclock the CPU to save power when idling). However, 1.536V is rather high just for idle. Again, that could be some weird software issue though. Should be fine if you've set CPU voltage to 'auto' in BIOS.

Memory is underclocked, but that seems unlikely to cause issues... anyway, that's a simple fix. Just go into BIOS and manually set the numbers. You'll need to increase voltage to 2.2V-ish for DDR2-1066 speeds though.

If you're still having issues, then remove one set of ram. Perhaps there's some sort of conflict with mixing ram.

Since you suspected the PSU, you could also swap the PSU, but you'd need a spare PSU to swap with.

 

[mfenn]

CPU-Z's DRAM Frequency is the clock rate before DDR doubling, so you are running at DDR2 667. This is because 333Mhz is the greatest common SPD setting that your Kingston and OCZ memory share. The OCZ can definitely handle DDR2 800, and the Kingston can probably handle DDR2 1066 with some more volts, so just do as blackmage said and set it manually in the BIOS.

As for your power off issue, I'd suspect the PSU given the symptoms and the fact that Xclio isn't exactly known for quality.

 

[mnewsham]

Is it just me or is his CPU voltage too high? I could get 3.4Ghz with a 1.4v on my q6600. His is 1.5v What is your OC? 😕

 

[firebirdude]

The CPU is probably fine; That's just SpeedStep at work (underclock the CPU to save power when idling). However, 1.536V is rather high just for idle. Again, that could be some weird software issue though. Should be fine if you've set CPU voltage to 'auto' in BIOS.

Memory is underclocked, but that seems unlikely to cause issues... anyway, that's a simple fix. Just go into BIOS and manually set the numbers. You'll need to increase voltage to 2.2V-ish for DDR2-1066 speeds though.

My system voltage is set to Auto in BIOS and in the grayed out part (because it's set to auto) it says Normal CPU Voltage-1.325V. BIOS also confirms Vcore at 1.556V though. CPU temp is at 55 Celsius. Seems high for no real work. HWMonitor show the actual cores all around 60 Celsius. And I am using an aftermarket heatsink. Can't remember which one, but it was very high quality a couple years ago when I bought it. Large. Copper with tons of tiny fins. Fan is still spinning. Should be working well.

As for CPU overclock. It's 266x9 which equals 2.4GHz. Looks like stock. I've messed with it in the past when I first assembled the computer, but honestly I really don't remember what has happened. It has been a solid 2 1/2 years.

I think I'm going to switch the system voltage to manual and throttle it back. Don't know why it's running high. I'm also going to bump up the memory frequency a notch. I'm also going to keep my eye on HWMonitor to watch the PSU voltage during idle and load. Right now, they're all within +-5%, but the +12V rail is close (12.54).

 

[firebirdude]

Ok. I switched system voltage to manual and put it to 1.425V. I also bumped up the FSB to 295x9=2.65GHz. With these new, faster, settings all my monitoring software shows CPU temps in the 49-50 Celsius range. I have no clue why the Auto setting put the Vcore voltage to 1.53. BUT The most puzzling thing is that all my monitoring software, including inside BIOS, shows the Vcore voltage at 1.38V. It remains at this with or without load. What happened to 1.425V???? Why are all my set voltages showing up lower? And this means the "Auto" setting before must have had that Vcore set at like 1.65V before!! On top of this, you can see in the CPU-Z screenshots that the RAM was at 1.8V for some reason. I bumped the DRAM voltage up by "+0.2V" and the monitoring software shows 2.0V now. Why is the default so low? I have to get into the "red numbers" in BIOS to even get to the 2.2V recommended by the manufacturer.

I've checked to make sure I have any integrated "motherboard tweakers" disabled. Shouldn't be any automatic tweaking going on.

 

[fffblackmage]

Ok. I switched system voltage to manual and put it to 1.425V. I also bumped up the FSB to 295x9=2.65GHz. With these new, faster, settings all my monitoring software shows CPU temps in the 49-50 Celsius range. I have no clue why the Auto setting put the Vcore voltage to 1.53. BUT The most puzzling thing is that all my monitoring software, including inside BIOS, shows the Vcore voltage at 1.38V. It remains at this with or without load. What happened to 1.425V???? Why are all my set voltages showing up lower? And this means the "Auto" setting before must have had that Vcore set at like 1.65V before!!

Hm, I guess it's not a glitch then. D=

In my experience, getting a slightly lower voltage than what I explicitly set in BIOS is normal. Weird, but normal.

On top of this, you can see in the CPU-Z screenshots that the RAM was at 1.8V for some reason. I bumped the DRAM voltage up by "+0.2V" and the monitoring software shows 2.0V now. Why is the default so low? I have to get into the "red numbers" in BIOS to even get to the 2.2V recommended by the manufacturer.

CPU-Z doesn't display ram voltage. Those numbers in the table just shows the different profiles the ram could be set at. Like for DDR2-800 speeds, you can use 5-5-5-18 timings with voltage set at 1.8V (that combination of clockspeed, timings, and voltage is a JEDEC standard).

You'll notice that the timings for DDR2-1066 is listed under "EPP" or Enhanced Performance Profile. It's not a JEDEC standard, which means you'll get better performance (usually not much), but you compromise reliability and stability.

The default is "low," because 1.8V is the JEDEC standard for DDR2 ram. Anything over that is not officially supported. The only thing JEDEC requires is that the ram is still functional up until 2.2V (IIRC). "Functional," but that doesn't mean the ram will actually work properly (error-free), which is why I recommend ram that does not require a higher voltage than the standard - 1.8V for DDR2, and 1.5V for DDR3.

 

[firebirdude]

So just to update.

I removed the heatsink. The thermal paste was a little crusty. I've seen worse, but it certainly wasn't like toothpaste anymore. After cleaning both the heatsink and the processor with isopropyl alcohol, I applied AS5 exactly how AS recommends. Blew the heatsink fins down to remove dust and reinstalled it.

Be sitting at idle for about an hour with a temp of............................. 37 Celsius!

Started this thread at about 55 Celsius, dropped to about 50 Celsius by switching off Auto Vcore voltage. And now I just lost a solid 12-13 degrees by renewing my thermal paste. Will this solve my random shut downs? So far so good, but time will tell. Rock on!

 

[mnewsham]

So just to update.

I removed the heatsink. The thermal paste was a little crusty. I've seen worse, but it certainly wasn't like toothpaste anymore. After cleaning both the heatsink and the processor with isopropyl alcohol, I applied AS5 exactly how AS recommends. Blew the heatsink fins down to remove dust and reinstalled it.

Be sitting at idle for about an hour with a temp of............................. 37 Celsius!

Started this thread at about 55 Celsius, dropped to about 50 Celsius by switching off Auto Vcore voltage. And now I just lost a solid 12-13 degrees by renewing my thermal paste. Will this solve my random shut downs? So far so good, but time will tell. Rock on!

Wow congratulations on the big drop in temp. Hope it solves your problems!