My first OC - some questions too

okay, so, maybe not completely my "first" OC as i've played around wth the ASUS OC utility and once bumped my system in at 8x400 with everything set to auto just to see if it would boot - it did, but i didn't leave it on very long...

so, okay then, this is my first time running an OC fully manually and testing stability.
now, the good stuff
my system is:
ASUS Maximus Formula SE w/ 0802 bios rev
Q6600 G0
EVGA 8800GTX
2x1GB Ballistix
corsiar HX520 PSU
2x 250GB perp HDD in raid 0
1x 250GB backup

it is air cooled - 4-120mm running front bottom to back top with one 90mm at the very top center of the case blowing down into the case (i'm leary about this choice... i was thinking it would introduce some turbulance and help distribute the air a bit more through my case... but that may have been over thinking things a little..)
the CPU has a monster tuniq tower 120 and all other cooling is stock (the waterloop has nothing on it)

i was able to get 9x334 (3.0Ghz) to run stable on al initial voltages - which seem a bit low to me (1.16Vcore?) and got absolutely no errors from prime95
but then i push it up to 9x367 (3.3Ghz) and stock wouldn't even run (as expected) but i had to push Vcore all the way up to 1.25 (shows as 1.22 on Probe) before i could get stable
even at that i've only run prime for about an hour and now F@H about an hour and a half, and ran several bench marks while pushing my graphics card
i guess i'm happy with 3.3GHz, and i dont think i go much further given that my load temp is about 63 when the room temp gets to its max...

but, so my question is is it normal to have to up the juice by such a large percentage for such a small gain?

next i pushed the vid card core+shader up in increments and benched it until i got errors (in this case a BSOD..) then pushed just the shader until it didn't seem to help my bench
and i stopped short on pushing my mem clock, a mixture of getting tired of running benches and not being sure of how to tell once i've crossed the line...
but so my GPU is running 612/1458/945 core/shader/mem - you guys think i can push any more out of that?
i did get a 14,165 score from 3DMark06 so no need to push it too far i suppose

here is a screen cap - let me know if you guys see anyplace i could gain some performance or stability (as in, i dont really know when to push the NB or SB voltages up.. i just kinda went with one step per three of the Vcore, which seems to have worked...)




sorry about the res on that... not certain of how to get it better.. (what site allows that size image?) let me know if you guys cant read it and i'll try to post another...
some of my other bios settings are as follows:
Strap: 400
twister: light
trans booster disabled - relax 1 (zero didn't seem to be cutting it)
PLL 1.54
FSB term 1.26
LoadLine Calibration: enabled
CPU GTL: 0.63x
NB GTL: 0.67x
and all the advanced CPU options disabled

also, should i stick with the 1:1 divider or push up the mem speed a bit?
if i stick with the 1:1 i'll be pushing the timings one notch down, but not certain about the timings on the next divider as the mem would be OCd at that point

You shouldn't really need to raise NB or SB until >400 FSB (the abit IP35 series generally starts needing it between 425 & 450). Once the CPU starts needing voltage, they can get to be hogs. Mine scales very nicely though. Up to 2.9 I'm stable at stock Vcore (1.325). After that, every 100MHz takes an extra .04v, all the way up to 3.5GHz / 1.565Vcore. I'm too chicken to try any higher.

I'd say keep going with your FSB and take your Vcore up to see how it goes. Keep an eye on temps and you're fine. Since you're Folding@Home, your temp limit may be a little lower than others. I keep mine under 60c when folding. Some say 65c. People who don't run 24/7 go all the way up to 70-75c but that's insane for a folding rig.

I'd find the limit of your CPU (or at least how high you want to push it) and then worry about getting your RAM up to speed. One step at a time.

Anyway...keep pushing, looks like you're going strong so far!

thanks

oddly enough, even after testing stable at the settings i had here and running F@H for a few hours i went later that day to run p95 overnight but the thing would go even one second without an error... so i shut the machine down and am going to change a few setting today. but, why would it loose stability like that?
also, after reading the OC tutorial in the sticky of this page i found that by default i had a few settings wrong that would cuase my PCI-e to "creep" up... so i'm going to go change those (pci-e clock set at 116 by default? normal?)
i want to keep my limit to 60
i cant really afford to replace parts very often, plus there are times that my ambient temp gets about 8F higher than what i was running here... and i've found that generally with this rig that one degree F room temp seems to translate into ~one degree C CPU temp (at 72F my load temp was 58C, at 76 it was 62...)

okay, so i went ahead and bumped up the volts on that one more nitch but at load i'm seeing temps of 67C plsu, even though there were no errors, one fo the prime tests stayed at test1 for about 45 minutes. i'm not certain but i think that is a sign of an instability...
so whatever, i dropped my bus back down to 334 and the volts to stock but went ahead and bumped the CPU volts a couple of notches to 1.20 to ensure a stable a run (hopefully) and i'm going to let prime run over night then drop them down tomorrow if all goes well

but i'm curious about something: when running prime cores 0 and 3 stay fully loaded - flat lined, but 1 and 2 never stay at 100%, more float around 95-98%
but F@H does push all cores at 100% constantly.. is this something i should worry about?
i think* it started once i OCed my rig - at least, i didn't note this behavior before...

Gripp said:

oddly enough, even after testing stable at the settings i had here and running F@H for a few hours i went later that day to run p95 overnight but the thing would go even one second without an error... so i shut the machine down and am going to change a few setting today. but, why would it loose stability like that?

Click to expand...

i'm still wondering what happened there?

plus i'm pretty bumbed that i cant even push a 3.3ghz OC on this machine
would, maybe, doing the pencil mod help this case? i'm a little scared of doing something like that though - is it as risky as it seems?

wow? no one around to help out?

Gripp said:

i'm still wondering what happened there?

plus i'm pretty bumbed that i cant even push a 3.3ghz OC on this machine
would, maybe, doing the pencil mod help this case? i'm a little scared of doing something like that though - is it as risky as it seems?

Click to expand...

I had a similar problem, I had 3.88ghz stable for 3 hours on orthos, then I stopped the test and said good enough for me, a week or 2 later, no settings changed and it wont go 10 mins without an error, same evap temp and everything, so im at a loss as to what happened too. I just bumped up the voltage some and its all good, but im on phase so I can get away with more voltage.

It may have something to do with the temps of the motherboard itself like the pwms or the nb and stuff like that on the board that gets hot, but seeing how you got so much airflow i cant see that being a problem with you.

And don't be afraid of the pencil mod, if its available for your board, a decrease in vdroop allows you to set a lower voltage in bios and still have stability because once it loads up the voltage wont drop off, just keep an eye on your temps and its no different really. Just be very careful when doing it as to not short out anything else (which isn't likely) and to not accidentally give a static charge to anything you may bump into.

thanks
i was thinking it could have been a vdroop issue - heat increases resistance is about the only thing i have going on that theory though... so wanted some input from the pros before putting something foreign on my board...

really i need to take more time pushing my rig step by step
i just kind of jumped from 2.4 to 3.0 to 3.3 and then back to 3.0...

i'm going to do the pencil mod this weekend

okay, haven't done the pencil mod yet but i did push some further tests
i found that something about 9x350 is the point before i hit some kind of wall
i can run 9x350 @ 1.2 (didn't try 1.8, just started with 1.2 and pushed it as far as i could since my temps are still very good at this voltage
but when trying to set 9x355 i kept pushing it up til 1.3something and was still getting errors so i backed down to 9x350 and left it at 1.21875 to ensure stability - i might tweak that later, but i really want to solve this wall issue first.
i found that while running @ 1.21875 my vcore is reported to be 1.1040 - a vdroop of 0.1085v - i dont know much about vdroop - so is that good/bad/average?
how much will i benifit from the pencil modbased on that?
are there other voltages or settings that i could change to combat this wall? maybe CPU PLL? or maybe the FSB termination voltage?

BIOS settings vs. what CPU-Z reports for voltage is not vdroop. Vdroop is going from idle to load. What does CPU-Z say at idle and load for you?

The vcore settings you see in the BIOS are just labels and may or may not correspond to actual voltages.

Try setting the vcore to auto, and let the mobo automatically increase voltage for you and see what happens. Make sure to keep an eye on temps, and take notes on what voltages CPU-Z is reporting. You should be able to get to 9x400 pretty easily. Later go back and set vcore manually and see how low you can get it at your target freq. If temps are good at that point your golden. If they're too high you should consider lowering your OC.

i've actually taken a turn for the super unstable...
i cant even run OCCT more than a couple more minutes until getting a code 1 on stock right now...
i was getting some ACPI compliant BSOD on windows startup too, seem to have gotten that fixed though - but now windows is laggy
i think maybe i've killed my board


edit:

jason4207 said:

BIOS settings vs. what CPU-Z reports for voltage is not vdroop. Vdroop is going from idle to load. What does CPU-Z say at idle and load for you?

The vcore settings you see in the BIOS are just labels and may or may not correspond to actual voltages.

Try setting the vcore to auto, and let the mobo automatically increase voltage for you and see what happens. Make sure to keep an eye on temps, and take notes on what voltages CPU-Z is reporting. You should be able to get to 9x400 pretty easily. Later go back and set vcore manually and see how low you can get it at your target freq. If temps are good at that point your golden. If they're too high you should consider lowering your OC.

Click to expand...

oh..
i thought it was a some carying amount of resistance that caused the actual voltages to differ from the volts set...
but, well, since this is the case - if anything i have noted my vcore going UP on load
i thought it was my ASUS essentially bracing itself for the load
maybe it is one of the ASUS processes that adjusts for this behinds the scenes?
i didn't check these values during the brief period that i had asus's stuff shut down
and probably wont kill those start up items again as my big problems seemed to have started when i did that..
so, with that said - does ASUS push the core up? or am just mistaken?
and even still, how then to test for what Vdroop is? what program etc. ?

speaking of which - as an update to what i wrote above - it was my RAM
i got 47 errors in memtest86 even though just a couple weeks ago it passed all
so i guess in one of the steps i took for this OC i fried my ram

The Voltage Damper option will sometimes boost load voltages above idle voltages. More power when you need it most!

Gripp said:

it is air cooled - 4-120mm running front bottom to back top with one 90mm at the very top center of the case blowing down into the case (i'm leary about this choice... i was thinking it would introduce some turbulance and help distribute the air a bit more through my case... but that may have been over thinking things a little..)

Click to expand...

Heat rises, so I'd think having that top fan as an exhaust would be a better choice.

jason4207 said:

Heat rises, so I'd think having that top fan as an exhaust would be a better choice.

Click to expand...

true
but wanted to make sure i had positive pressure in the case to help dust out..
i do think about switching that one though
really i need to run smoke tests
was just initially in a hurry to get the thing put together since 1-it had been so long since i had been in a computer and 2-wanted to test it for any defects while i could return things...

but, yeah
i tested both the sticks of memory with one known to be good and they both gave errors.... so i did fry my memory by just setting them to 840 or whatever it was
think i'm going to hold off on OCing my mem again until i get something that plays nicer with this mobo...