Prime95 runs stable for 4-5 hours, then computer reboots

[EricVPI]

When I was overclocking, prime95 used to give me some rounding errors. I increased voltages and thought I got it stable. Prime95 now gives no rounding errors, but will make my computer reboot after about 4-5 hours.

Here's what I have now:
Gigabyte P35 DS3R
E6850 3.0ghz @ 3.6 (400x9)
Vcore = 1.4
FSB voltage = +0.1v
PCI-E voltage +0.1v
MCH/G (whatever that is) voltage = +0.1v
Kingston DDR 800 voltage +0.1v (running at 800 mhz, but I added +0.1 because the default ram voltage is a bit too low).

I'm just using air cooling. Under smallFFT prime95 torture test, my cpu cores won't go any higher than 64C. It idles at around 37C. Anyone have any suggestions what I can do to try to make this more stable? I have NO problems playing games or anything else, but the anal part of me wants my system 100% stable.

 

[slicessoul]

If you lock PCI-E freq at 100, you don't need to set +0.1v for PCI-E voltage. Let it at normal.

Your kingston RAM is stated at what voltage ? if you set +0.1v means your RAM run at 1.9v. And what's the timing of your RAM ?

What's your PSU ?

Did your PC beeping before it shutdown ?

 

[yamahazr]

do you use stock cooler?upgrade your cpu cooler.it will definetely brings the temp down especially when load.less than 50c..64c is max temp for conroe,will result in shorter lifespan

 

[slicessoul]

do you use stock cooler?upgrade your cpu cooler.it will definetely brings the temp down especially when load.less than 50c..64c is max temp for conroe,will result in shorter lifespan

E6850 has thermal specs : 72°C. For max temp at 64°C, i think he's fine.

 

[EricVPI]

I have a 650W PSU. I forget which brand, but it not cheap and it had good reviews.

The Kingston ram specs SAY 1.95v but I've read some people have theirs default lower. The kingston ram is ddr800 but it is detected at ddr667 and has to be overclocked to reached spec. Other people have reported this too.

I am using the stock cooler because I think 64C is okay. If i need to continue upping voltages and it goes to 66C or higher I will get a better cooler. RAM timing according to specs is 5-5-5-15, but CPU-Z says it's 6-6-6-18. Does this sound like a RAM issue? Too high FSB issue? Or not enough vcore?

Does 1.4v vcore seem ok for a 400x9 overclock?

 

[slicessoul]

RAM timing according to specs is 5-5-5-15, but CPU-Z says it's 6-6-6-18

You need to set it in BIOS MIT manually. Press ctrl-F1 at BIOS main menu, go to MIT, set DRAM spd into manual and change the value one by one. Set RAM divider = 2 to get 800.

I am using the stock cooler because I think 64C is okay

What softwares do you use to check the CPU temp ?

Does 1.4v vcore seem ok for a 400x9 overclock?

Results may vary just don't exceed 1.5v. You need to test it by yourself. Try to reduce the voltage a little and torture the CPU again. If it's stable, reduce the voltage again. In the time you got error/unstability, that means, the previous voltage is the good one.

I have a 650W PSU. I forget which brand, but it not cheap and it had good reviews.

Check again it's brand.

 

[EricVPI]

PSU:

http://www.newegg.com/Product/Product.asp?Item=N82E16817171023

What softwares do you use to check the CPU temp ?

Coretemp. Under smallFFT prime95 @ 1.4v it will max out at about 64C for the cores. Prime95 stable for about 4-5 hours then usually my comp just reboots, instead of previous rounding errors. Tempted to up Vcore to 1.41, even on just stock cooling.

try to reduce the voltage a little and torture the CPU again.

Well it's not prime95 stable for at least 8 hours now so I don't think I should reduce CPU voltage yet. I'm trying to figure out what I need to change.

 

[EricVPI]

Upped vcore from 1.4 to 1.41v.

System crashed about 5 hours in. I used to get rounding errors when I was below 1.38v, now system just reboots. Something else is amiss. My computer is noticeably faster at 3.6 from 3.0 for many things. I am in agony here.

I don't mind the 64C max load temp, and I certainly doubt that temp is causing my comp to reboot. What else could it be? Insufficient +v to the FSB/MCH(G)? Ugh. My system has significant vdroop. Actually, HUGE vdroop. What could the problem be?

 

[cnumartyr]

Drop the multi 1 step and see if it does it again.

 

[cd14]

I find it that when my PC reboots it's because there's not enough voltage, either to the ram or CPU. When it blue screens, it's the ram. When my system doesn't boot on the first cycle, I need more voltage to the CPU. I have the 965-DS3 and P35-DQ6. They both act the same way.

Also, from my experience, there's no need to add voltage to the FSB or MCH(G) until you get close to 450 FSB on these boards. This is with a E6400 and Q6600.

 

[EricVPI]

Really? Before I added +0.1v to the FSB prime95 would error out in about 5 minutes. After I added +0.1 it changed to a few hours. All this was while at 1.38v.

I've changed my settings to auto and my now my vcore is at like 1.43. Max temp is now 65C under load but doesn't get any higher.


I've heard auto is a bad idea though. How bad is it?


CD14 my PC reboots only about 5-6 hours into prime95.

 

[EricVPI]

I have it on auto voltage and my core temperature is now LOWER 2C. Huh? (400x9)

 

[cd14]

Really? Before I added +0.1v to the FSB prime95 would error out in about 5 minutes. After I added +0.1 it changed to a few hours. All this was while at 1.38v.

I've changed my settings to auto and my now my vcore is at like 1.43. Max temp is now 65C under load but doesn't get any higher.


I've heard auto is a bad idea though. How bad is it?


CD14 my PC reboots only about 5-6 hours into prime95.

I've been up and down with these Gigabytes, just mainly with two E6400s and the Q6600. My first E6400 didn't fair as good as the second, it couldn't muster 3.2ghz. My Q6600 is a B3 and it's keep going as long as I give it juice, but the temps are way high. They all have varying overclocking capabilities.

I've had the same rebooting problem and it stems to the voltage issue...not enough. Either you'll have to lower the overclock or give it more juice. To give you another example, my machine would reboot when I ran CoreTemp. I gave the CPU more juice and the problem went away.

Try finding the limits of the board, ram and CPU one at a time. For example, set your multiplier lower (x8) and raise the RAM speed (multiplier) to see how far your ram will run while keeping everything else default (or close). Keep the stock timings and use "Blend." This will stress the ram more. I've had it pass the LRG FFTs for hours and fail the blend within 10 minutes. Once you find each components limits, you'll have a better idea on what to adjust.

Adjust to what you feel comfortable with.

 

[cd14]

I have it on auto voltage and my core temperature is now LOWER 2C. Huh? (400x9)

Your BIOS is adjusting the voltage dynamically based on the load. Perhaps you have more than enough on the manual settings. Or maybe your room is cooler?

 

[EricVPI]

Well I've heard auto-voltage is dangerous, but so far the temps seem ok. When I compensate for vdroop it seems autovoltage is clocking me around 1.43ish.

I'm going to run smallFFT prime95 overnight and see if my comp is still up in the morning on autovoltage. I hope this ain't dangerous...

 

[EricVPI]

Autovoltage seemed to work. Running small FFT stable now for 9+ hours.