Q6600 What's a more stable overclock

[Angerisagift]

I looked around the boards and didn't see anything specifically addressing this.

Ok so this weekend I got the GA-P35-DS3L and the ASUS Arctic Square from frys, I've installed them and was able to push to 3.0 by simply locking the PCI-E at 100Mhz, setting my ram multiplier to 2.0 and pushing my FSB to 3.0ghz on stock voltage.

I've been able to push up to 3.2Ghz Orthos stable for at least an hour using both 4x800 and 9x355 with vcore around 1.3 (currently using auto setting)

I know my board is rated at 1333Mhz FSB so I would assume once I start backing down the voltages I would be able to get a lower vcore using the 9x355 since I'm not pushing the boards FSB as far beyond it's stock speed. but I could be wrong.

This is the first time I've done a major overclock so really I'm looking for some advice in what direction to go then I plan on refining that overclock to get it as stable as possible

turning back on C1E, EIST lowering vcore etc

I don't really care about synthetic benchmarks, the most intensive thing I run is Crysis, so I don't need a super high overclock, just enough to get me that little extra push, and from what I've heard 3.2Ghz sounds like the sweet spot for these chips.

Any advice is welcomed

By the way, I've not yet seen temps exceeding 56C using coretemp.

update: Current stable config is 9x355Mhz = 3195Mhz, C1E and EIST on, vcore at stock 1.2875, as low as 1.120 all other voltages also at default, when idle, highest temp I've seen is 57C stress testing with orthos, Crysis on high at 1440x900 stays around 46C. avg 35-45FPS ram timings tweaked to 4-5-5-12 as opposed to 5-5-5-15 at 800Mhz.

 

[Idontcare]

You could probably hit 9x367 without much more work if you wanted. I've got Q6600 at 9x367 on DS3L. That's 3.3GHz. Might not be worth it to you though if it takes more Vcore to do it.

Given your stated application usage and desire to not seek out the highest overclocks...a sentiment I share with my DS3L...I would recommend going with the lower FSB higher-multiplier combo and reduce the wear and tear on your motherboard.

You can recoup some of the lost memory bandwidth by tuning your memory timings if it bothers you to go with lower FSB.

Your temps are nice at that voltage and GHz, you'll enjoy that in your electric powerbill and AC cooling of your home this summer.

A note on C1E, EIST. etc...keep your Vcore setting to "auto" if you want to enjoy having C1E lower your Vcore when the chip is idle. Anything you do in the BIOS which causes the voltage to be manually set will subsequently prevent the C1E and EIST from reducing voltage at idle.

 

[Angerisagift]

would C1E and EIST essentially make up for lowering my vcore anyways, since they throttle down those settings when the system is under idle conditions?

 

[Idontcare]

Its a good question, do you have a Kill-a-watt unit to measure the differences?

If not then we can at least speculate. Lowering Vcore manually results in a lower Vcore at load (and at max multiplier) whereas leaving Vcore at stock and allowing C1E/EIST to reduce both mutliplier and Vcore at idle...which is lower on your household power bill?

Do you use your computer much? What is the ratio of idle time to non-idle time when the system is on?

I'd lean towards letting EIST do its thing and drop your Vcore at idle versus having a higher Vcore at idle just to have a somewhat lower Vcore at load. But obviously it depends on your expected load ratios.

 

[Angerisagift]

Well, I work and I'm married, so the most I use my PC is about an hour when I come for lunch, then maybe 2-4 hours at night, I might use it a bit in between, just web browsing etc, but it only gets loaded maybe 5 hours out of the day, maybe a bit more on weekends.

I'm going to try setting vcore at the stock 1.2875v and see how things work out, leaving EIST and C1E on

I also took your advice and tweaked the ram timings,

they were running at 5-5-5-15 at 800Mhz stock voltage

now I've got them running 4-5-5-12 at 710Mhz also stock voltage.

the modules are OCZ Gold XTC PC6400 sticks, let me know if you think I could adjust anything else.

 

[Idontcare]

We need to get some of the more knowledgeable posters in here to give you more assistance. I'd be happy with what you got there but I know there are folks on this board who could no doubt help you get even tighter timings and quite possibly a higher overclock.

Ha, I just got your NIC, anger-is-a-gift, clever

Next time I reboot my DS3L I will grab some pics of my BIOS screens and upload them here in case they have some more voltage/timing tweaks you can take advantage of. Even though I just set these systems up only 6 weeks ago, my memory is fading really fast.

 

[Angerisagift]

Sorry to Bump, still looking for additional input

 

[SystemPlue]

hmmm i would go with the higher multiplier...
I run my computer 24/7 ... so far it is really stable @ 9x333 1.264V

I mainly do work on the desktop while running World of Warcraft on the 2nd display..... I have ran Prime95 for 10hrs and it barely reaches 50degrees.... and even with WoW + Video + work apps core temp barely passes 40degrees

This is my first real overclock.... so i am no expert.... =(

on a side note my ram timing is still 5-5-5-18 and i have tried 9x400 1.4V ... it booted into windows.. but i ran Prime95 and went out for 30min... when i got back windows had rebooted itself =(

 

[MadScientist]

I would leave your ram timings loose, 5-5-5-15, and keep your ram at 800 mhz or lower, to take it out of the equation, until you know how high you can oc your CPU at temps you are comfortable with.

An Asus Artic Square HSF, not bad according to the review at Frosty Tech, but for the money I would have gone with one of big boys, like a Tuniq Tower 120. 57C full load seems a bit high for stock voltage but still fine. I can run my Q6600 stable at 3.6 Ghz at 1.3575 vcore, full load temps in the high 50's to low 60's with an el cheapo Coolermaster Hyper TX2 HSF. My VID is a little lower than yours, 1.2375V. I also have an Antec 900 case withh all fans on med setting.

For now I would disable C1E and EIST until you get a stable oc.

Use the Small FFts setting in Orthos when oc'ing your CPU and then Blend after you get your ram set to stable settings, use Memtest86+ V1.7 or V2.0. Running Orthos for 1 hour is ok for initial testing, 6 hours or more is better. I personally like to use Prime95 V25.5.

Are you stressing all 4 cores in Orthos? Open Orthos, Task Manager, Processes tab, right click on ORTHOS.exe, Set Affinity, and make sure all cores are checked.

 

[Idontcare]

Originally posted by: MadScientist
For now I would disable C1E and EIST until you get a stable oc.

I know this is common advice to would-be overclockers, but I have never understand the theory behind why this should make a difference.

Any ideas why disabling C1E/EIST would make an overclock more stable? These features should only be active when (1) the chip is idle, or (2) the chip is so hot it needs to be throttling.

Neither "corner" of the C1E/EIST feature set involves stability of a loaded CPU when the temps are below throttle setpoint.

I'm not disagreeing with the advice, I follow it too, but just curious why this is/was expected to help. Anyone know?

 

[Angerisagift]

Well I realized I wasn't stress testing correctly, installed prime95 25.5, was running 24.1 and it wasn't loading me at 100%, I get an error in 1 thread after about 30 minutes at 1.312v however if I leave Vcore at auto it's stable for at least an hour. is there any harm in leaving it at auto, since CPUZ is barely reading over 1.2 while loaded I'm doubting it's my vcore, any idea what I should be looking at next, I'm thining Dimm overvoltage but not sure,

what's the effect of just leaving my mobo V settings at auto, the DS3L or at least the F5 bios mine shipped with only has an option to set voltage in general to auto, you can't set individual components to auto.

 

[Idontcare]

Originally posted by: Angerisagift
however if I leave Vcore at auto it's stable for at least an hour. is there any harm in leaving it at auto, since CPUZ is barely reading over 1.2 while loaded I'm doubting it's my vcore

That's my advice, leave it to auto. Setting it to anything other than auto prevents C1E/EIST from lowering it when the computer is idle anyways, and leaving it at Auto lets the BIOS dial it up as needed even when overclocking on the DS3L.

The BIOS will allow Vcore to dial-up as needed until Vcore reaches Intel's max allowed VID which is 1.35V (IIRC...).

So your overclock will be rock solid at any speed that requires less than 1.35V peak during load (on DS3L at least) if you leave the Vcore set to "Auto" in the BIOS.

Meanwhile because you left it to auto you'll also get the power savings when the computer is not at full load.

I really enjoyed this feature of the DS3L. I wonder if more mobo's do this.

 

[lopri]

One of the reasons (actually the biggest reason, I guess) I was so impressed with BadAxe 2 was that it handled all the power-saving features flawlessly, with overclocked configuration. CPU frequency, voltages, and sleep/wake all work without a hitch - both on XP and Vista. There were times when video drivers caused issues with sleep/wake, but that looks to be solved for the most part, leaving it solely to motherboard engineering. My DFI P965-S has a bad habit of changing frequencies when waking up, and the 780i I have now runs every fan hooked to the max on wake-up. Both overclocked, of course.

 

[MadScientist]

Originally posted by: Idontcare

Originally posted by: MadScientist
For now I would disable C1E and EIST until you get a stable oc.

I know this is common advice to would-be overclockers, but I have never understand the theory behind why this should make a difference.

Any ideas why disabling C1E/EIST would make an overclock more stable? These features should only be active when (1) the chip is idle, or (2) the chip is so hot it needs to be throttling.

Neither "corner" of the C1E/EIST feature set involves stability of a loaded CPU when the temps are below throttle setpoint.

I'm not disagreeing with the advice, I follow it too, but just curious why this is/was expected to help. Anyone know?

It should not affect the stability of the overclock, but why take the chance, another possible variable to worry about. You can always enable them later and see if it does affect the stability. Here's a little more info about them: Link

 

[nullpointerus]

Originally posted by: Idontcare

Originally posted by: Angerisagift
however if I leave Vcore at auto it's stable for at least an hour. is there any harm in leaving it at auto, since CPUZ is barely reading over 1.2 while loaded I'm doubting it's my vcore

That's my advice, leave it to auto. Setting it to anything other than auto prevents C1E/EIST from lowering it when the computer is idle anyways, and leaving it at Auto lets the BIOS dial it up as needed even when overclocking on the DS3L.

The BIOS will allow Vcore to dial-up as needed until Vcore reaches Intel's max allowed VID which is 1.35V (IIRC...).

So your overclock will be rock solid at any speed that requires less than 1.35V peak during load (on DS3L at least) if you leave the Vcore set to "Auto" in the BIOS.

Meanwhile because you left it to auto you'll also get the power savings when the computer is not at full load.

I really enjoyed this feature of the DS3L. I wonder if more mobo's do this.

The DS3L tops out at 1.35v w/ auto. voltage -- are you sure about that?

IIRC, my DS3R did more than that, depending on the BIOS version, and also set +0.3v on the vDIMM (which is insane for my value RAM). I'd load SpeedFan (or check the BIOS's voltage readings after saving settings and rebooting) just to make sure.

Of course, the model numbers *are* different...

 

[Angerisagift]

Well I just ran almost 3 hours of Prime with voltage at auto no errors no warning highest temp was 61 about 20 minutes in, then after that it leveled off around 58 for the for the rest of the run.

currently running with memory tweaked from 5-5-5-15 to 4-5-5-12
still 9x355 multiplier

So my question is, "Is it safe for me to keep the BIOS voltage setting on auto"

and if the answer to the first question is yes, should it then be safe for me to turn EIST and C1E on to reduce heat and strain on the hardware during idle times which constitute a large portion of the day.

Also I was kind of curious if it would have any effect on longevity going from a 3.0 overclock to 3.2 overclock, I'm basing my assumption on manufacturers like Gigabyte, ABit and ASUS design their hardware to handle above the on the box rating, despite.

My line of thinking is this, Chip manufacturers like Intel don't use a different fab for every single SKU they sell, I could literally be buying a GX6850 with a locked multiplier as easily as I am a Q6600

then if my mobo is rated to run at 1333Mhz FSB, by running a 333x9 Overclock and underclocking my ram to run at 667 I'm technically running within perfectly acceptable parameters, (Assuming my line of thinking is correct)

or is not that big a difference that an extra 22Mhz on the FSB would even matter.

I'd love some different opinions, if it weren't for this board I wouldn't have even tried OC'ing this thing. and am very grateful for all the help, I work in IT supporting hundreds of computers everyday and I still find myself learning things I didn't know previously here.

so really, thanks guys.

 

[VirtualLarry]

Originally posted by: Idontcare

Originally posted by: MadScientist
For now I would disable C1E and EIST until you get a stable oc.

I know this is common advice to would-be overclockers, but I have never understand the theory behind why this should make a difference.

Any ideas why disabling C1E/EIST would make an overclock more stable? These features should only be active when (1) the chip is idle, or (2) the chip is so hot it needs to be throttling.

Neither "corner" of the C1E/EIST feature set involves stability of a loaded CPU when the temps are below throttle setpoint.

I'm not disagreeing with the advice, I follow it too, but just curious why this is/was expected to help. Anyone know?

It's because some motherboards ignore your set multiplier as the max multiplier, and instead will run the full range of possible multipliers. If you've dialed up a high FSB and a lower multi, then when the BIOS sets a high multi, along with the high FSB, you're screwed.

Other reasons have to do with voltage. Usually, when you OC, you increase the vcore. But some mobos go along with set voltage/frequency profiles (I have no idea where these are stored or located, wether in the BIOS or in the CPU itself), and those profiles are based on the default VID, so when the BIOS goes to set a particular frequency and voltage, it sets the voltage too low and you crash.

Other mobos go the other route, and allow you to enable SpeedStep and C1E without those perils. However, they don't reduce the voltage at all, if you've modified it, so those features really don't help all that much in terms of power savings. So you might as well just disable them anyways.

So in a nutshell - if the features "work" - they will causes crashes, so you might as well disable them.
If the features don't "work" - then they barely save any power at all, so you might as well disable them.

So if you are overclocking, disable SpeedStep and C1E. Always.

 

[VirtualLarry]

Originally posted by: Idontcare

Originally posted by: Angerisagift
however if I leave Vcore at auto it's stable for at least an hour. is there any harm in leaving it at auto, since CPUZ is barely reading over 1.2 while loaded I'm doubting it's my vcore

That's my advice, leave it to auto. Setting it to anything other than auto prevents C1E/EIST from lowering it when the computer is idle anyways, and leaving it at Auto lets the BIOS dial it up as needed even when overclocking on the DS3L.

The BIOS will allow Vcore to dial-up as needed until Vcore reaches Intel's max allowed VID which is 1.35V (IIRC...).

Are you sure it stops there? I've read that you should always set vcore manually when overclocking, otherwise the mobo will overvolt something fierce when overclocking. I've heard of things like 1.45v or 1.5v vcore.

 

[Angerisagift]

just finished a 2 hour run of prime no errors, with C1E and EIST on, though I suppose I might not see the effect of this while torture testing.

the highest reading I get from CPU-Z is 1.344V and it constantly jumps from 1.344 to 1.328, I'm stilling planning on doing an overnight stability test, but at the moment it looks pretty stable, though I really would like to know what I need to do with my voltages.

This computer needs to last at least 3-4 years. I went P35 in the event that the 45NM's get dirt cheap after a year or two, but for now I can't replace anything major, so I'm even willing to lower my FSB if it means better longevity.

 

[Idontcare]

Originally posted by: nullpointerus

Originally posted by: Idontcare

Originally posted by: Angerisagift
however if I leave Vcore at auto it's stable for at least an hour. is there any harm in leaving it at auto, since CPUZ is barely reading over 1.2 while loaded I'm doubting it's my vcore

That's my advice, leave it to auto. Setting it to anything other than auto prevents C1E/EIST from lowering it when the computer is idle anyways, and leaving it at Auto lets the BIOS dial it up as needed even when overclocking on the DS3L.

The BIOS will allow Vcore to dial-up as needed until Vcore reaches Intel's max allowed VID which is 1.35V (IIRC...).

So your overclock will be rock solid at any speed that requires less than 1.35V peak during load (on DS3L at least) if you leave the Vcore set to "Auto" in the BIOS.

Meanwhile because you left it to auto you'll also get the power savings when the computer is not at full load.

I really enjoyed this feature of the DS3L. I wonder if more mobo's do this.

The DS3L tops out at 1.35v w/ auto. voltage -- are you sure about that?

Quite positive but be sure and note the distinction that I am referrring to the max Vcore the BIOS will allow the Vcore to float up to when you have left the BIOS setting for Vcore to "auto".

You can of course manually set Vcore on the DS3L to sky-high values, 1.70V or higher to recollection.

 

[Idontcare]

Originally posted by: VirtualLarry

Originally posted by: Idontcare

Originally posted by: Angerisagift
however if I leave Vcore at auto it's stable for at least an hour. is there any harm in leaving it at auto, since CPUZ is barely reading over 1.2 while loaded I'm doubting it's my vcore

That's my advice, leave it to auto. Setting it to anything other than auto prevents C1E/EIST from lowering it when the computer is idle anyways, and leaving it at Auto lets the BIOS dial it up as needed even when overclocking on the DS3L.

The BIOS will allow Vcore to dial-up as needed until Vcore reaches Intel's max allowed VID which is 1.35V (IIRC...).

Are you sure it stops there? I've read that you should always set vcore manually when overclocking, otherwise the mobo will overvolt something fierce when overclocking. I've heard of things like 1.45v or 1.5v vcore.

Well I am sure my five specific DS3L's and five specific Q6600's operate this way, I can't be sure of anyone elses boards or experiences.

In my case this cutoff requires me to manually set the Vcore for 1 of 5 systems as that one Q6600 requires 1.40V to hit 3.3GHz, all the others are OK with the <1.35 setting the BIOS drives the Vcore to at 3.3GHz. In those 4 cases the loaded Vcore is <1.35 but >VID for the specific chip. If I set the GHz to stock (2.4GHz) the BIOS sets the Vcore to VID for all the chips (confirmed by CPU-Z).

It's actually a beautiful system with Gigabyte (don't know if other bopards have it), as it ensures you aren't needlessly wasting Vcore on your chips (they are all Prime95 small FFT stable >24hrs). But like I said the Bios won't float the Vcore above Intel's upper spec of 1.35V (the highest VID allowed by Intel's binning process).

So if you are striving for an overclock which requires the CPU to suck up more Vcore than 1.35V then you have no choice but to go manual and juice up the Vcore in the BIOS. But once you do that, on DS3L's at least, then your power-savings no longer work in terms of having C1E/EIST reduce Vcore when the system is at idle.

 

[Angerisagift]

I'm going to leave it to auto for now, because atm I can't run prime for more than a minute or two without an error, that's with C1E and EIST off. might trying a few more small increments though

 

[nullpointerus]

Originally posted by: Idontcare

Originally posted by: nullpointerus

Originally posted by: Idontcare

Originally posted by: Angerisagift
however if I leave Vcore at auto it's stable for at least an hour. is there any harm in leaving it at auto, since CPUZ is barely reading over 1.2 while loaded I'm doubting it's my vcore

That's my advice, leave it to auto. Setting it to anything other than auto prevents C1E/EIST from lowering it when the computer is idle anyways, and leaving it at Auto lets the BIOS dial it up as needed even when overclocking on the DS3L.

The BIOS will allow Vcore to dial-up as needed until Vcore reaches Intel's max allowed VID which is 1.35V (IIRC...).

So your overclock will be rock solid at any speed that requires less than 1.35V peak during load (on DS3L at least) if you leave the Vcore set to "Auto" in the BIOS.

Meanwhile because you left it to auto you'll also get the power savings when the computer is not at full load.

I really enjoyed this feature of the DS3L. I wonder if more mobo's do this.

The DS3L tops out at 1.35v w/ auto. voltage -- are you sure about that?

Quite positive but be sure and note the distinction that I am referrring to the max Vcore the BIOS will allow the Vcore to float up to when you have left the BIOS setting for Vcore to "auto".

You can of course manually set Vcore on the DS3L to sky-high values, 1.70V or higher to recollection.

Last night I did a simple test with my DS3R.

When I set vcore to auto, EIST/C1E to on, and overclocked, I saw voltage upwards of 1.390v in CPU-Z. Factoring in vdrop and vdroop, that's well over the BIOS's 1.350v setting. It's more like the BIOS's 1.450v setting. And my RAM voltage was far above 2.1v.

It's good to hear that the DS3L works properly...
... but IMO it's unsafe to extrapolate that for other Gigabyte boards.

Anyway, the point I was trying to make turned out to be moot to the OP...

the highest reading I get from CPU-Z is 1.344V and it constantly jumps from 1.344 to 1.328, I'm stilling planning on doing an overnight stability test, but at the moment it looks pretty stable, though I really would like to know what I need to do with my voltages.

OK... Here's what I've learned about voltage.

1. The BIOS vcore setting is always higher than what CPU-Z reports.

2. The BIOS vcore setting is higher than what you will see in CPU-Z at idle. The difference between the BIOS and CPU-Z idle voltage is called vdrop, and it's typically upwards of 0.020v, probably closer to 0.030v.

3. Additionally, the actual voltage being supplied to the CPU at any given time depends on how heavily the CPU is loaded. So there is a different voltage at idle than there is at load. The difference between CPU-Z's reported voltage at idle vs. load is called vdroop, and it's typically upwards of 0.020v, probably closer to 0.030v.

4. A minor point: CPU-Z reports to three decimal places, but in reality the results are very coarsely-grained, so you will see a sudden "jump" between 1.296 and 1.312 instead of any numbers in between. It sounds wierd, but that's the way it works.


As I understand it, the goal of stress-testing is to make voltage drop as low as possible while making heat rise as high as possible, for the given settings. The rationale is that, if the processor passes these tests, then it should be fine for normal tasks.


So... when I was running 3.0 GHz w/ 1.5v... this is what I saw:

-- BIOS vCore: 1.50000v (manually set)
-- CPU-Z vCore @ idle: 1.452/1.476v (fluctuating)
-- CPU-Z vCore @ load: 1.440v (constant)


Now... when I dropped that 1.50000v setting to 1.47500v... this is what I saw:

-- BIOS vCore: 1.47500v (manually set)
-- CPU-Z vCore @ idle: 1.452/1.440v (fluctuating)
-- CPU-Z vCore @ load: 1.424v (constant)


So now if we look at your CPU-Z report voltages of 1.328/1.344v (fluctuating), are they idle or load? Meaning: did you get these numbers from CPU-Z while the system was doing nothing (idle), or while you were using Prime95's Small FFT stress test (load)?

It's hard to guess at what's going on...

Can you test CPU-Z voltages at idle vs. at full load?


I'm hoping that (a) we can confirm what Idontcare says your motherboard and (b) you'll end up with a better understanding of what's going on here.

 

[VirtualLarry]

That's really interesting, the different auto vcore behavior between the DS3L and the DS3R. I have a couple of DS3Rs, but I haven't tested auto voltage on them ever.

 

[frythecpuofbender]

My Q6600 G0 is running sweet and fast at 4x800 with Ram:FSB at a nice 1:1 ratio and 3,2Ghz. I could go higher but why? The hot summer months will come and I'm tired of the days when a CPU would heat up the room, hitting 60°C. Now when I'm at 100% all four cores (Maya rendering) it's hitting 45° and i like it that way.