E6400 Overclock Instability

[Ed989]

I have an E6400 overclocked to 3.2 Ghz on a Gigabyte DS3 board (400 x 8). It runs stable most of the time, but sometimes when gaming or doing other CPU intensive tasks, it reboots or locks up. When i run Prime95 it always fails, saying that too great of a rounding error has occurred. I don't think the problem is RAM, because it's running at stock speed and voltage (Corsair XMS2 DDR2 800 at 2.1V). I have been gradually increasing the CPU voltage, and it's currently at 1.28V. I have a Zalman CNPS7700-CU cooler on it. Any suggestions on what I should do to get it running more stable? I was thinking about trying to increase the fsb voltage a little bit, but i wanted to get some advice before doing that.

[chuck4456]

Run MEMTEST diagnostic for at least 20 minutes. Don't increase any voltages. Could we also have complete specs on your system?

[Ed989]

my system specs are:

Gigabyte DS3 w/ E6400 at 3.2 Ghz (400 x 8)
2 gigs Corsair XMS2 DDR2 800 (1 gig x 2)
512mb X1800XT
2 x 250 gig WD sata HDD's in RAID 0
500 watt Antec smartpower 2.0 ps

[Burn]

here what i say .. do what you want .. on your own risks ...

on a E6400, big jumps in voltage is around 2.8 - 3Ghz. So i say put that baby on 1.40v then use orthos , prime ..etc.. if its stable good , then lower the voltage by small increment until you loose stability. Always check your temps while doing so and STAY in front of your screen...

for that Corsair memory .. i have the same ... put that at stock voltage (1.9 not 2.1) i run them at 850mhz 1.9V 5-5-5-12.

all my voltage are set at AUTO in the BIOS but memory 1.9v and CPU 1.41v.

[Ed989]

Chuck,

I just ran memtest86 for 2 hours with no errors.....

[chuck4456]

No problem with the memory obviously. I'd give Mr.Ferrari's volume on overclocking a serious reading. The safest way to determine your problem is to go back to square one and take little steps upward. Hardware is expensive.

[Mr.Ferrari]

I think you mean redfallon's or tin's chuck.

Yeah I would try to add a bit more voltage (1.3v-1.34v) and see what happens. If that doesn't work, go for some more MCH voltage, set it to 1.65v and keep and eye on north bridge temps. Do not exceed 50c.

Also make sure your pci-e and pci buses are locked.

If spread spectrum is enabled, disable it.

[Ed989]

What should I lock the PCIe frequency to? I've heard that the standard is 100 mhz.

[Burn]

Also make sure your pci-e and pci buses are locked.

What should I lock the PCIe frequency to? I've heard that the standard is 100 mhz.

PCIe = 100Mhz
PCI =33.33

but i guess you should lock them for now ...

locking buses is discutable ... after months of perfect stability at 3.4Ghz with the buses locked i decided to put them back at AUTO and also put asus hyperpath 3 from disabled to AUTO ... i was like when i saw SANDRA memory bandwith test went from 65xxx to 68xxx ... so far i dont have issue .. even vista is happy with the OC

[Mr.Ferrari]

Thats just a old rule of thumb, depending on how high your overclock is. The pci-e and pci buses will usually raise along with the fsb of the processor overclock, and causing corruption in the devices on the bus.

For example, too high of a pci-e can cause you to lose LAN, get video anomolies, and loose connectivity to your hard drive.

Originally Posted by burn also put asus hyperpath 3 from disabled to AUTO ... i was like when i saw SANDRA memory bandwith test went from 65xxx to 68xxx ... so far i dont have issue .. even vista is happy with the OC

Actually, I think thats primarily due to the enabling of hyperpath 3. Hyperpath 3 basically allows speedier data transfer between the northbridge (memory control hub) and your ram.

However this has been known to cause FSB walls at higher overclocks due to tighter timings. As long as its stable for you, go with it.

[Ed989]

Alright, I have the PCIe locked at 100mhz, and the PCI at 33. I am currently at 1.31 V on the CPU. I have been running two simultaneous Prime95 test, and if either one gets an error, I've been going into the BIOS and increasing the voltage by the smallest increment it will allow. With both cores running at 100% load, temps are maxing out at about 49c right now. I figure I will continue with this strategy until either the system becomes stable or my CPU temps get too high.

[Ed989]

Now that I've exceeded 1.3 V on the CPU, CPU-Z no longer reports the voltage accurately. Any ideas why?

[Mr.Ferrari]

What voltage is it reading? What is actually set in bios?

[Ed989]

I have it set to 1.34375 V in BIOS, CPU-Z tells me 1.312 V

[LeftyAce]

Is that before it goes under load? What are the temps? It's possible the processor is scaling back to prevent overheating....

[Ed989]

Okay, I think I've got it. At 1.34 V the E6400 seems to be stable at 3.2 Ghz now. I ran 2 simultaneous threads of Prime95 with no errors for 2 hours. At lower voltages it was failing within 5 minutes. Maximum temp is 53c. Then I tested the system by playing FEAR, Rainbow Six Vegas, and Oblivion all at max settings, and not only were there no crashes or lockups, but performance is much better than before (much better framerates). Before, my framerates would dip noticeably when there was a lot going on (heavy action, etc) but not at all now; it just plows right on through.

[Mr.Ferrari]

It could be one of two things, either this is just classic intel vDroop (voltage droop). All intel motherboards/processors will do it under load usually.

Or this is just C1E and Speedstep controls lowering the processor multiplier and/or voltage while under low to no load.