I'll start this from the beginning:
I have an i7 920 and an EVGA x58 LE board. I OCed the 920 to 3.8GHz, set the DIMM voltage to 1.65, and set the Vcore to 1.3 (I know it's a bit high for the frequency, but my heat is under control so until I have a lot more time to fine tune my OC, this is ok.)
Anyway, I forgot to bump my VTT to 1.2V to match the increase in DIMM voltage. I was happily running the system for a long time, fully stability tested and folding some nice crunchy BugWUs. My version of stability tested, in this instance, is over 50 hours of systester (I run ubuntu) on all 8 threads. I read something about matching the VTT and DIMM voltage, remembered I'd forgotten, and decided to fix it. While I was in there, I tried to get my RAM clock up from it's meager 1066MHz (rated 1600). It's set at 2:8 and bclk is 190, but it still shows 1066 when I boot (although BIOS shows a "target" frequency of 1522MHz).
So I bumped it up to 2:10, just to see what would happen figuring at worst it might not boot. Which it didn't. So I hit the clear CMOS button, yadda ya, put all my settings back the way they were, only this time with the VTT voltage where it should've been. Booted.
Then I ran systester for a few minutes just to make sure I hadn't done something horribly stupid. There was no immediate lock up, so I figured this being my exact same OC from before it ought to be stable and I turned on the folding client. System crashed within a minute or two.
After going into BIOS a couple times, having the whole thing happen to me a couple times, and scratching my head more than a couple times, I finally found the one thing I'd missed: I set the board to NO VDROOP. I recall Brolloks review stating that the board had
And I didn't doubt it. Yet now I've been running the folding client for about a half hour with no lockup (after changing it to NO VDROOP). Could that really be the hole in my stability?
So now I'm considering things, and I have to wonder whether the observed lack of VDROP/VDROOP was because the setting "NO VDROOP" was selected, or without it? CPU-z doesn't work in wine (at least that's what I've found), so I cannot test this myself.
This post raises an awful lot of questions (at least for me), but the only one I'm really concerned with answering right now is whether my system should be stable. IDENTICAL settings, but with VTT bumped up 100mV. My expectation would be yes, but I'm just as surprised by the effect of not setting NO VDROOP as I am certain that I should be good to go after *upping* a voltage, which should make a system more stable.
Edit:
Speedstep is off. CPU Frequency Monitor is toggling between 1.6GHz at low load to 2.79GHz at peak load. BIOS still says CPU is at 3.8GHz. Folding is much slower than normal. Yet systester benchmarks match my previous findings for 3.8GHz, which far outshines even the 3.26GHz that the "Dummy OC" put me to.
Something bad is happening here, and I don't like it.
I have an i7 920 and an EVGA x58 LE board. I OCed the 920 to 3.8GHz, set the DIMM voltage to 1.65, and set the Vcore to 1.3 (I know it's a bit high for the frequency, but my heat is under control so until I have a lot more time to fine tune my OC, this is ok.)
Anyway, I forgot to bump my VTT to 1.2V to match the increase in DIMM voltage. I was happily running the system for a long time, fully stability tested and folding some nice crunchy BugWUs. My version of stability tested, in this instance, is over 50 hours of systester (I run ubuntu) on all 8 threads. I read something about matching the VTT and DIMM voltage, remembered I'd forgotten, and decided to fix it. While I was in there, I tried to get my RAM clock up from it's meager 1066MHz (rated 1600). It's set at 2:8 and bclk is 190, but it still shows 1066 when I boot (although BIOS shows a "target" frequency of 1522MHz).
So I bumped it up to 2:10, just to see what would happen figuring at worst it might not boot. Which it didn't. So I hit the clear CMOS button, yadda ya, put all my settings back the way they were, only this time with the VTT voltage where it should've been. Booted.
Then I ran systester for a few minutes just to make sure I hadn't done something horribly stupid. There was no immediate lock up, so I figured this being my exact same OC from before it ought to be stable and I turned on the folding client. System crashed within a minute or two.
After going into BIOS a couple times, having the whole thing happen to me a couple times, and scratching my head more than a couple times, I finally found the one thing I'd missed: I set the board to NO VDROOP. I recall Brolloks review stating that the board had
ZERO VDROP OR VDROOP !!!
And I didn't doubt it. Yet now I've been running the folding client for about a half hour with no lockup (after changing it to NO VDROOP). Could that really be the hole in my stability?
So now I'm considering things, and I have to wonder whether the observed lack of VDROP/VDROOP was because the setting "NO VDROOP" was selected, or without it? CPU-z doesn't work in wine (at least that's what I've found), so I cannot test this myself.
This post raises an awful lot of questions (at least for me), but the only one I'm really concerned with answering right now is whether my system should be stable. IDENTICAL settings, but with VTT bumped up 100mV. My expectation would be yes, but I'm just as surprised by the effect of not setting NO VDROOP as I am certain that I should be good to go after *upping* a voltage, which should make a system more stable.
Edit:
Speedstep is off. CPU Frequency Monitor is toggling between 1.6GHz at low load to 2.79GHz at peak load. BIOS still says CPU is at 3.8GHz. Folding is much slower than normal. Yet systester benchmarks match my previous findings for 3.8GHz, which far outshines even the 3.26GHz that the "Dummy OC" put me to.
Something bad is happening here, and I don't like it.
Last edited: