Overclocking my Q6600 soon

[Nagisan]

Just thought I would post to get some ideas of what this can do, it is the G0 stepping, and I run water cooling.

Now, here is the cool part, I also have an Antec 900 case (very bada-- case IMO), my Q6600 idle temps are 25C. How is that for an idle temp on a quad? Good? Bad? Before you respond, take into account that the computer room it is sitting in has an ambient temp of around 27C, so basically my water cooling is keeping my stock Q6600 below ambient room temps. The highest I have seen it (havnt gone 100% load for hours, but have played games for hours) is 31C.

That being said, what would you guys estimate an achievable overclock is just based on the temps I am getting fully stock in a fairly high ambient room temp?

The ram I am getting here soon is DDR2 1000, I plan on probably setting it to 800 so it has more overhead for the overclock, good or bad idea?

Another question I have is what is a good voltage for a high O/C on a Q6600?

Last, the setting that underclocks your processor when its got a low load on it, is it ok to keep that on when O/Cing or should that be disabled, and what exactly would it be called in the BIOS?

Just trying to get all the info together in one spot before I get started.

EDIT: Ok, I went ahead and threw a small overclock on it, from 2.4 up to 3Ghz (25% O/C) and turned the voltage to 1.35, ran prime95 (the multi-core version) for ~10 minutes (under max heat mode), came back and it was sitting at a nice 33C, room temperature is at ~26.6C (80 F). I feel confident I will be able to hit a pretty decent overclock with it. Yes I know a 10 min test isnt much, but I figure in all practical usage, 10 minutes of 100% cpu usage on all 4 cores isnt pretty practical when the main usage of the computer is gaming.

EDIT2: As reference, my old case (crappy 4 year old cooler master) ran an AMD 4200+ X2 at 40C+ idle (around 50C load), with the same water cooling, and the same block (different mounting on it). Goes to show you what a case with good ventilation can do, even when you have water cooling.

Edited August 31, 2008 by mwpeck

[gwai lo]

Okay, there are a few things I'd like to point out in the your second paragraph, don't take them too personally. First off, even though no one should care about idle temperatures, people do for some reason. Why shouldn't you care? Well, unless your chip doesn't play nice with subzero temperatures (it's happened before on XS), then the lowest temperature the chip is going to reach is not important. It is the highest temperature that defines when the chip will throttle itself and then when it will shut down. Since you couldn't have a cold bug issue, we can go ahead and not care about your idle temperatures.

Next, the nature of the digital thermal sensors used by these new chips from Intel translates to them not being very accurate in load temperatures, but even more so with idle temperatures. The temperatures you're looking at are most likely found using this formula: Tjunction Max - Delta to Tjunction Max (read from the on die registers) = Temperature. Tjunction max is individually calibrated per chip at the factory *and* we don't have numbers released for 65nm chips, unlike the 45nm chips. The register reads the delta from tjunction max as an offset of zero, with the 45nm chips it "bottoms out" at 20?C under Tjunction max, some people have done testing (here) so we know tj max is still pretty far away from your 25?C.

Next, there is NO WAY for your chip to run under ambient temperatures unless you have a chiller. If you didn't have a chiller, then that requires evaporative cooling methods with a coolant with a wet bulb point much lower than the ambient temperature. Water's dew point is relatively close to ambient temperatures, so it does not possess that property. Then we consider thermodynamics concepts and we'll realize that your loop cannot be cooled to sub ambient temperature. The transfer of heat from the water to the air is done via the fans and the radiator, however it is the temperature of the air in the end that is one of the determinants in the delta between water entering the radiator and the water leaving as it is the medium that is carrying away the heat. If your water was running cooler than your ambient temperature, it will absorb heat from the environment until the two reach an equilibrium point. It's like your prcocessor running cooler than the temperature of your water, it's possible..but it won't stay like that for long. I'm not sure if your entire second paragraph was some attempt of showing off or not, but either way it doesn't really add much to your thread. Idle temperatures are useless and it's unlikely that you're running cooler than ambient.

Next up, to actually answer your question. No on can tell you what you can overclock, every system is different, each chip has minor defects as do northbridge chipsets, RAM chips, the motherboards themselves, your power regulators, etc etc. So, you're going to have to try it yourself. With that said, you should be able to pull of 3.6 GHz with your water (no idea what board you have..if it's an Nvidia based board then probably not).

500?9 would be quite a feat, 400?9 is pretty good for most Q6600's, however you've already got the RAM so there's nothing you can do about it now. DDR2-1000 won't hurt, that's for sure.

A good voltage? Again goes back to each system being different, you'll have to do it yourself. The voltage ceiling for the 65nm chips seems to be around 1.6v though, you'll want to take this reading at idle via CPU-Z or something. Why idle? Same concept as temperatures, it's the highest value and also most likely the only value that will affect longevity.

You'll want to disable Speedstep and C1E until you've found a stable point. It's just one more factor to remove from the equation as the feature drops the vcore, in addition to the multiplier, which could lead to instability.

I've noticed most of these threads end up with "just try it".

[Nagisan]

Thanks for the reply, I wasnt trying to show off, I just found it interesting in how cool it was running. For reference, the individual cores are running between 28-35C when its idle (at 25C). I guess what I need to ask is how is 33?C for 10 minutes of full load in a 27?C room?

I understand theres always differences in hardware, no two setups no matter how similar are guarenteed to get the same overclock, what I was wondering is how much overclock room does that leave having a load temp not a whole lot higher than ambient temp?

[gwai lo]

33?C leaves you plenty of room from a temperature standpoint, just wanted to point out that it's not possible to run sub ambient so your temperature reading is probably inaccurate.

Anyway, I guess the short answer to your question is.....temperature is not a factor in your overclock right now.

edit: Oh yeah, just curious but what are you running in your loop?

[Nagisan]

Would it just be an inaccurate sensor or the program? I sort of doubt the program (everest) is wrong because it shows the same temps for each individual core as core temp does. Under load the hottest individual core is around 45?C, the coolest can be down around 35?C.

[+shift. MVC]

And to answer your first paragraph again, don't look at absolute temperatures. Look at the term coined as "delta to Tjunction max" which is directly from the on die registers.

And to further elaborate, Core 2 Duo chips (Conroe and onward) use what is called "tjunction maximum" to determine absolute temperatures.

The formula is roughly as follows:

Tjunction Max (a number, which Intel has not released for 65nm chips, but for 45nm dual core chips it is 100C) - Delta to Tjunction max (and from this point onward will be called delta) = CPU Temperature

Now what is the problem with this formula, and why should you be looking at the "delta" and not the "CPU temperature" ?

Well to make an extremely long story short, since the "delta" is the only actual number that the CPU is providing you with, it is in fact the "most" accurate value that is provided to the CPU owner. In order to calculate absolute temperature (CPU temperature; refer to formula above) one must know the Tjunction maximum (which should be provided from Intel, and at this point is not available for 65nm chips).

Now what is this "delta" and "tjunction max" anyway, and why is it important?

Tjunction maximum is the temperature at which the CPU sends out the "PROCHOT#" signal and tells the CPU fans to go to 100% in an attempt to cool your CPU. Now what is important to understand is that, this tjunction maximum is not available for 65nm CPU users (and this will be the third time that I state this fact). So this translates to the following, since we don't know the Tjunction maximum, we will be forced to look at the "delta" which is the distance (in degrees C) TO the tjunction maximum. And IN THEORY, one should be be able to run a CPU at 1C (delta to tjunction max, or distance from tjunction max) 24 hours a day 7 days a week 365 days a year, for 3 years; and if your CPU fails (assuming you have not overclocked) Intel will give you a new chip (this analogy was taken from a user at HardForums.com)

Now with that in mind, remember that the Tjunction max is the temperature at which PROCHOT# signal is sent, and does NOT meant that your CPU will spontaneously combust at that temperature, any further from the Tjunction max, and your CPU will shutdown to save itself.

And also remember, that the only relatively accurate value that the CPU provides to the user, is the delta to tjunction max. The actual value of Tjunction maximum for each individual chip is (at this point in time) unknown.

So to make an extremely long story short in a small equation:

A (Tjunction max) - B (Delta or distance FROM Tjunction max) = C (CPU Temperature; The absolute temperature)

Now, if the CPU is providing us with B. And Intel has provided the very rough number of 100C (for 45nm C2D chips).

Then that must mean that if A = 100, and B = Let's say 55 in this situation.

Then that must mean that A - B = C which means 100 - 55 = 45.

BUT, here is the catch! For each individual processor, the "A" is different (maybe one or two degrees, maybe by one hundredth of a degree; we will never know).

So if A is unknown, and B is known, and in order to find C we need to know A. Then really, all we know is that as long as "B" remains ~10-20C, then my CPU will not reach the point at which PROCHOT# signal is sent. :)

It's like using the Pythagorean theorem or Sine , Cosine and Tangent to find angles and lengths of a side in a right triangle. You don't want to use calculated values to find other values. You want to use the numbers that are given to you, and that you know for sure 100% are correct. And in this case, you don't want to use A to find C, because we are not sure of what A is in the first place, so just make sure B never gets lower than 10C and you will be fine. ;)

Edited August 31, 2008 by shift.

[Dessimat0r]

I'm at 3.5ghz, stable on a Q6600, so they can be pushed rather far :) Just remember to tweak with GTLREF in your BIOS, mine's at around 70-68% at the moment.

[Nagisan]

Thanks for the replies, I was at first going to stick it on 3Ghz and just let it go at that, but my average FPS in CS:S's benchmark (I know not the best thing to use) went up about 20FPS just by going from 2.4Ghz to 3Ghz.

That being said I think I will likely do what I have done on all my AMD chips (this is my first intel chip :D), and thats, overclock till it doesnt post (or becomes unstable) then back it off until its stable. That seems to be the general answer and is likely the best way to go about overclocking, push it till it errors, then lower it until its nice and smooth.

[gwai lo]

Would it just be an inaccurate sensor or the program? I sort of doubt the program (everest) is wrong because it shows the same temps for each individual core as core temp does. Under load the hottest individual core is around 45?C, the coolest can be down around 35?C.

Okay now this is kind of confusing. What is this 25?C reading from? Your cores sound about right as they're all above ambient, is the 25?C the Tcase or something?

and lol I shift, I read that exact same thread with the analo:p. :p

Also, mwpeck check out this database, it's the one i've found with the most entries:

http://www.hardforum.com/showthread.php?t=1138241

It's too bad XS doesn't have something similar, but they do have this. You'll just have to work harder for your information.

http://www.xtremesystems.org/forums/showth...highlight=q6600

[Nagisan]

I'm not exactly sure what the 25?C is reading from, its just labeled(in everest) as "CPU", the 25?C doesnt show up in Core Temp, but core temp just shows the internal core temps. I have a feeling the 25?C is the temp if you were to touch the outside casing of the processor(probably the Tcase).

Like I said, with the same cooler on my 4200, the "CPU" reading was never lower than 40?C.

[+shift. MVC]

Use CoreTemp or RealTemp, but instead of looking at CPU core temperature, look at distance from Tjunction max.

[Nagisan]

Been running Prime95 on all 4 cores for the past 30 mins and heres my temps (also bumped my O/C up to 3.2Ghz for now).

Will be running it overnight to see what my temps are in the morning, right now it is 25?C ambient temp.

[+shift. MVC]

Looks good. Notice how if you change the Tjunction max, the distance from it stays the same. That's what I mean by using numbers that you know is "fact" and not "calculated values"

[Nagisan]

Ok, it ran overnight and stayed at the same ?C to Tjunction max, I overclocked to 3.6ghz and I after re-adjusting the voltages, I get a hard lock after about a minute of stress testing. Hard locks are something I havnt seen in my previous overclocks, what from overclocking can cause a hard lock? I figured maybe it cannot handle that high FSB (bought this as a budget board way back in the workstation I put it in), but I want to check to be sure.

[Nagisan]

Sorry for the double post (cant edit my last one), got it to 3.4Ghz stable, I'm pretty happy with that as its still 400Mhz above the previously highest processor I had (when I had my FX, it ran at 3Ghz), and its quad to boot. That being said if anyone knows what would have caused the hard locks at higher speeds, let me know, I dont plan on pushing it higher unless I can find for certain whats causing the locks. Core temps at this O/C vary from 38?C to 46?C, rather low IMO, but then again I am running water cooling.

Well, like I said, if anyone has any thoughts on the hard locks, let me know, otherwise I'm happy sitting at 3.4Ghz.......I have a feeling the hard locks are from the MOBO FSB, its an nVidia 650i chipset which has a rated max FSB of 1333 Mhz, I'm running it stable at 1512 Mhz.

[gwai lo]

Try dropping the multiplier and seeing what kind of FSB you can do. This doesn't guarantee that you can just bump the multiplier and vcore back up and be fine at that frequency, but it gives you an idea of what your chipset and RAM can do.

[Nagisan]

I actually locked my RAM at 800Mhz so I could take that outta the equation, didnt think of lowering my multiplier, I'll test that out as soon as I get a chance.

[Nagisan]

Ok, FSB of 1600 (or 400, w/e number you wanna go by) didnt freeze with a multiplier of 8x, but it did fail a prime95 test after about 8 minutes, FSB of 1556 (or 389) seemed to not freeze and it lasted at least 10 minutes (when I stopped it) in prime95.

Would that mean my CPU's limit is somewhere between 3.4 and 3.5 Ghz (runs fine on 9x at 3.4 ghz but not at 3.5 ghz), or is there another setting or something I could change that would fix it (like possibly raising the voltage a little more)?

EDIT: Just tried 3.5 with a higher voltage and it froze after about 11 minutes of prime95, hard lock, forced to reboot it. Think somewhere between 3.4 and 3.5 is just the limit for my proc? Either way, I'm happy with what I was able to get.

Edited August 31, 2008 by mwpeck

[gwai lo]

Not necessarily, it only gives you an idea. Overclocking is a weird process that isn't as logical as people would like it to be. Finding your max overclock will require tweaking both sides.

[Nagisan]

I'm the kind of person who likes overclocking, but doesnt like spending hours getting the settings tweaked just right, think I'll just stick with my 3.4 and be done with it, decent enough O/C for me.

[toadeater]

I'm the kind of person who likes overclocking, but doesnt like spending hours getting the settings tweaked just right, think I'll just stick with my 3.4 and be done with it, decent enough O/C for me.

3.4 isn't bad for a q6600. There's not much to be gained going to 3.6. Tweak your RAM timings and tRD as much as you can.

You can also try Set Affinity for a slight performance boost. It can assign single-threaded programs to core 3 & 4 so that they don't utilize the first two cores which most Windows software uses. This way you can get use out of all four cores. Assign all the stuff that runs in the background like AVs, firewalls, hardware utilities to cores 3 & 4. You could also limit programs to a core(s) so that they can't lock up your PC because of 100% CPU usage.

http://www.geocities.com/edgemeal_software...inity/index.htm

[Nagisan]

3.4 isn't bad for a q6600. There's not much to be gained going to 3.6. Tweak your RAM timings and tRD as much as you can.

You can also try Set Affinity for a slight performance boost. It can assign single-threaded programs to core 3 & 4 so that they don't utilize the first two cores which most Windows software uses. This way you can get use out of all four cores. Assign all the stuff that runs in the background like AVs, firewalls, hardware utilities to cores 3 & 4. You could also limit programs to a core(s) so that they can't lock up your PC because of 100% CPU usage.

http://www.geocities.com/edgemeal_software...inity/index.htm

I'll think about setting the affinity, as for my ram, it was running at (auto detected) 5-5-5-18, it was rated at 4-4-4-15, so I tuned them down to their rated speeds (lower timings) and its running fine.

Quick question about Affinities.......if I set a program (lets say my internet browser) to Affinity 4, if I reboot (or just close and reopen it) will it default to Affinity 4?

[Dessimat0r]

As I said, GTLREF really helps. I wouldn't be at 3.5ghz without it.

[MiPeNiS]

stumbled across this thread after inheriting a pc. looks like a stock cooler, no name that sticks out. overall I am getting idle temps in the 50's and under some load I am in the 60's and games I can see 70 and what looks to be 75 max. Is this too hot? (I am new to overclocking) also please keep in mind I am running a 280 gtx as well.

Edited September 3, 2008 by MiPeNiS

[gwai lo]

What are your ambient temperatures? What stepping Q6600?

At first glance, it seems warmer than average, but it's no where near too hot.