Overclocked C2D freezes during games

[DoodieBody]

Hey all, I'm back

I've been working on overclocking my E6300, biostar 965pt, 2gigs A-Data ddr2-800, 8800GTS setup. The RAM was suggest by 4 people all having the same luck of getting it to about ddr2-900 on the same mobo. I'm running the latest BIOS.

I'm having a fairly odd issue. Running at 2.8Ghz, it's perfectly stable, through both stress tests and gaming. Running at 2.9GHz, however, is a totally different story.

I began my testing at 2.9GHz. I first loaded up coretemp to keep an eye on my temps. Dual Super Pi 32M, passed. NEXT! Orthos blend while running atitool's artifact scanner (which strains my videocard and really cranks up the temp on it) at the same time. 2 hours, no errors. My temps were 58C max on CPU, 73C on GPU. That is my worst case scenario. Those two tests run my components waaay hotter than any game does (during gaming I'm generally around 51C on the CPU and 66 on the GPU). Computer seems relatively stable right?

I load up some Counter-Strike: Source. Within 5 minutes it freezes...huh? I load up EQ2, within a couple minutes, freeze. Any game I play freezes between 5 seconds and 20 minutes into playing. If I lower my CPU to 2.8GHz, it's 100% stable.

I have all CPU features off, I've tried vcore between 1.3 and 1.35, no change in stability. I've tried vdimms of 2.0,2.1, and 2.2. Timings of 5-5-5-15, set to ddr2-533 multiplier. PCI-e locked at 100 (PCI isn't an option, but I did some research and it is auto locked on this board.)

When the computer freezes everything locks up and an annoying buzzing sound comes out of my speakers. The sound is the same every time.

How is it possible that it's completely stable during extreme stress tests for hours, yet chokes after seconds while gaming?

UPDATE: I've tried voltages up to 1.4V, no change. I also loaded an older version of the BIOS, again, no change. I tried running 3dmark 06 for a change. I set it to run each test twice, just for some extra stress. 2.8GHz, no problems, ran right through it. 2.9GHz, froze before it even finished loading the first test.

 

[IntelUser2000]

My thinking is that programs that stress your system is like a synthetic benchmark. All Stress programs really need to do is just that, stress the system much as possible. In real world, you worry about everything else, development costs/time/game play/keeping it up-to-date.

You'll notice the worst of the worst graphics processors can run synthetic 3D benchmarks without errors, you just can't see that in real world. I'd assume its the same thing.

Stability is not just about temperatures and running within acceptable range, it'll also have to do with how it runs the code etc, stress programs must be more synthetic I'd reckon.

 

[DoodieBody]

True, but in practice it's generally accepted that stress programs are much more likely to crash out an unstable computer than regular use would. An error here and there might not get noticed in a real program, but in a complex math equation that a stress test uses, even a slight error would fail the program.

While you may have a point, generally speaking a computer is more likely to fail stress programs than they are real world programs.

 

[IntelUser2000]

DoodieBody:

I think you've got a point too, but we may be looking at an exception here.

Just run it at 2.8GHz then, so many things go on a computer there are nearly unlimited numbers of errors that can pop up. The 2.8GHz COULD just be a hardware limit, maybe a mobo or CPU??

 

[Mallet]

what power supply are you running?

 

[DoodieBody]

http://www.newegg.com/Product/Product.asp?Item=N82E16817189005

It's received pretty stellar reviews...and it should be more than good enough to power my system. I thought about the PSU as well, but that almost exclusively leads to reboots, not freezes. And would the CPU really draw that much extra power going from 2.8 to 2.9 with no increase in volts?

 

[Markfw]

My guess is that stress testing doesn;t stress the GPU, and gaming does, plus uses more power. My best guess is that power supply may not be enough.

 

[DoodieBody]

But it runs 100% stable, even during gaming at 2.8GHz. I wouldn't see how upping the fsb to 415, but not changing the voltage, would suddenly strain the PSU.

 

[aka1nas]

Try looping a version of 3dmark while doing Orthos. Your power supply probably can't handle the power demand from everything at once at those settings or your board simply doesn't like the higher FSB.

 

[Elfear]

The Xclio PSU's are actually pretty dumpy according to the XS forum (excellent source of info). The GreatPower series looks to be top quality but apparently the rest are tier 5.

That being said, it does seem strange that at 2.8GHz you are perfectly stable while at 2.9GHz (with no voltage changes) you freeze in games. What kind of hdd are you using? Sometimes if the board can't handle a high fsb it will get errors when trying to read or write to the hdd. The PCI-E lock should work but maybe the board just can't handle 415MHz on the fsb (not familiar with your board so that may not be the problem). Sata drives seem to be more susceptible to this problem than PATA drives.

I had the same problem with an MSI board I was using. Prime95 stable all day long but when I would jump into a game or access the hdd a lot it would freeze on me. Turned out to be the PCI locks not working correctly over a certain fsb.

I'd also run Orthos for at least 8 hours and see if you get any errors just to eliminate that as the potential culprit.

 

[Mallet]

Originally posted by: DoodieBody
http://www.newegg.com/Product/Product.asp?Item=N82E16817189005

It's received pretty stellar reviews...and it should be more than good enough to power my system. I thought about the PSU as well, but that almost exclusively leads to reboots, not freezes. And would the CPU really draw that much extra power going from 2.8 to 2.9 with no increase in volts?

If the PSU is overloaded, then yes it would most likely reboot not freeze. But an insuffiecient powersupply can also cause stability issues.

What is probably happening is the added load from the gpu is causing the voltage on the 12 volt rails to droop. This will negatively affect the stability of an overclock.

 

[Reznick]

I just started a post that mirrors this one (overlooked this), since I think some game are inherently more susceptable to OC than others regardless what your H/W configuration is.

What if you tried the games at the lowest resolution/graphics/audio settings? Might be painful on the eyes, but would they still crash? Just curious. Wouldn't power consumption be lower overall? Might help eliminate if this is a P/S issue, or some other issue.

 

[Markfw]

Originally posted by: DoodieBody
But it runs 100% stable, even during gaming at 2.8GHz. I wouldn't see how upping the fsb to 415, but not changing the voltage, would suddenly strain the PSU.

When it is running faster, it takes more juice. You can run faster, no games (no video stress) or slower but video stress. That almost guarantees that your PSU is weak.

 

[DoodieBody]

According to : http://www.hardforum.com/showpost.php?p=1027898523&postcount=7 it's been a good PSU.

http://www.tweaktown.com/reviews/829/4/page_4_final_thoughts/index.html also seemed to like it. Plenty of positive customer reviews at newegg as well.

Still, it's possible that power is the issue. I think I may have found an answer, though. I believe the northbridge handles some of the communication with PCI-e...so I thought, maybe that needs more juice. I upped my vmch and fsb termination, and I was able to run through 3dmark 06 (it would insta crash before). I didn't do any long term testing though, my northbridge was way too hot for my tastes. That may or may not have fixed it, I'm a little timid to run my NB that hot for hours. (it was hot enough that holding my finger on it for more than a couple seconds was very uncomfortable. Not quite burning, but much hotter than stock)

Default on vmch is 1.25. I'd like to up it to 1.35, unfortunately my board has a known issue where the 1.35 selection in the bios actually runs it a 1.45v...eep. Too much juice for me.

 

[DoodieBody]

While we're on the subject, just how hot should your NB get...with the higher volts it's still touchable, but definitely is very, very, hot...and would probably cause a burn after 30 seconds or so. I'm no temperature genius, I'm gonna say that's around 60C.