Quick question before I attempt to OC

[Xee]

Well, I finished my new build a couple of weeks ago and want to try my hand at overclocking (never done it before). First off, here's what I got:

Case: Antec 900
Processor: Intel Core 2 Quad Q9450
Mobo: Asus Rampage Formula
Memory: G.SKILL 4GB
PSU: PC Power & Cooling S75QB 750W ATX12V
Cooling: Thermalright Ultra-120 Extreme CPU Heatsink Lapped + 120mm Scythe Minebea
OS: Windows Vista Ultimate 64 Bit

Before I get started though, I just had a quick question concerning my idle temps. I downloaded and installed RealTemp and it's currently showing temps of:

37°, 37°, 28°, 29°

Is it normal for their to be that much of a difference between the first and last two cores? Should I not pay attention to these temps until I run a stress test to see if the difference still exists? I'm beginning to wonder if my heatsink isn't seated properly.

 

[MightyDiomed]

Your temps are exactly where they should be. Don't worry about all the cores reading differently, it really comes down to the different sensors for each core.

 

[jimbo6846]

Well, I finished my new build a couple of weeks ago and want to try my hand at overclocking (never done it before). First off, here's what I got:

Case: Antec 900
Processor: Intel Core 2 Quad Q9450
Mobo: Asus Rampage Formula
Memory: G.SKILL 4GB
PSU: PC Power & Cooling S75QB 750W ATX12V
Cooling: Thermalright Ultra-120 Extreme CPU Heatsink Lapped + 120mm Scythe Minebea
OS: Windows Vista Ultimate 64 Bit

Before I get started though, I just had a quick question concerning my idle temps. I downloaded and installed RealTemp and it's currently showing temps of:

37°, 37°, 28°, 29°

Is it normal for their to be that much of a difference between the first and last two cores? Should I not pay attention to these temps until I run a stress test to see if the difference still exists? I'm beginning to wonder if my heatsink isn't seated properly.

Different temps are normal, but you have a rather high gap between cores 0/1 and 2/3. I see that you lapped your TRUE, but not the CPU, correct? It's possible that the IHS of your Q9450 is slightly uneven, and the side with cores 0/1 is slightly further away from the base of the heatsink. Did you test the "even-ness" of your CPU?

 

[Xee]

Did you test the "even-ness" of your CPU?

Nope. I didn't even lap the heatsink myself; I had it done from the store I bought it from. I really hope that's not the case though.

 

[jimbo6846]

Nope. I didn't even lap the heatsink myself; I had it done from the store I bought it from. I really hope that's not the case though.

Don't stress too much about it. Your temps are actually pretty decent, you have plenty of OC'ing headroom, even with the 2 cores starting out at 37*C. You could try re-seating the TRUE (make sure you use new thermal compound if you do, always reapply the compound when removing/reseating the HSF). If you want to check the even-ness of both your HSF and CPU, set a razorblade on them and shine a light through it. You'll be able to see the gap if there is one. I do not recommend lapping your CPU, or having it lapped by someone else. It will void the warranty.

Even if the temps you have now are the best you're gonna get, you're still fine for a decent OC.

 

[Zero82z]

Here's a great post from unclewebb (the guy who wrote Real Temp) that basically sums up the whole Core 2 temperature situation:

I used to believe that but according to the information that Intel released at the August IDF, that really isn't true either. They showed that at the point they calibrate these sensors there is a certain amount of error but for the 45nm Dual and Quad Cores that most of us use, they didn't specifically say how big that error is. Plus or minus 5C at the calibration point is a reasonable assumption based on the information they've released for the 45nm Atom line as mentioned above.

They also explained "slope error" which means that the data coming from the on chip digital thermal sensors is changing at a different rate than the core temperature is changing at. A sensor might move 40 positions when the actual core temperature has only changed 30 degrees or maybe the core temperature has changed by 50 degrees.

What you end up with is a temperature curve that is basically floating at both ends. Intel might have tried to clarify things by announcing TjMax for the 45nm processors but due to calibration error, TjMax is far from an exact number. When combined with the "slope error" these sensors have, an uncalibrated reported temperature of 50C is little more than a wild guess.

If your sensors are not sticking and you take the time to calibrate RealTemp to your CPU then I think your reported temperatures will be more accurate. 100% accurate? I doubt it and even if they were 100% accurate, there's no easy way to prove that one way or the other. Knowing the nature of these sensors, you have to assume that it isn't. Other software that doesn't give you any method to correct for "slope error" is going to be even less accurate.

Intel hasn't provided enough information about these sensors for any software to accurately translate DTS data into core temperatures and the bigger problem is that these sensors simply aren't good enough for that purpose. I guess that's why they've been telling us for years not to use them to try and report accurate core temperatures.

In short, your idle temps are completely meaningless since the DTSes aren't designed to report temperatures that low. For all you know, the actual temperatures could be 50C or 15C on every core, because the readings simply aren't accurate enough to rely on. As for the temperature gap between your cores, that's perfectly normal, and nothing to be concerned about.

 

[BillParrish]

In short, your idle temps are completely meaningless since the DTSes aren't designed to report temperatures that low. For all you know, the actual temperatures could be 50C or 15C on every core, because the readings simply aren't accurate enough to rely on. As for the temperature gap between your cores, that's perfectly normal, and nothing to be concerned about.

I tend to agree, and keep in mind windows was running stuff (even if all you were doing was coretemp) and it normally does stuff on the first two cores first so that could easily account for 5C of the delta.

Its if you see a big difference at load between the cores that you start wondering about the heatsink mount.

 

[Xee]

Thanks for all the help. Looks like I got to put aside some time this weekend and see what I can get.

 

[Zero82z]

Its if you see a big difference at load between the cores that you start wondering about the heatsink mount.

Not even. Big variations between the temperatures of different cores are generally not a sign of improper mounting, but merely that the chips simply don't run at the same temperature. Intel quad-core CPUs frequently have differences of ~10C between cores, and it's no cause for concern.

 

[Riccochet]

The contact between the IHS and cores also comes into play, but is less likely a culprit for un-even temps across cores on a quad. Like others have said, when you slap two dualcores together one may run hotter than the other. One may overclock better than the other. It's the shittier dual-core that limits your overclocking potential.

 

[BillParrish]

Not even. Big variations between the temperatures of different cores are generally not a sign of improper mounting, but merely that the chips simply don't run at the same temperature. Intel quad-core CPUs frequently have differences of ~10C between cores, and it's no cause for concern.

Well then my question would be, When would mounting issues be a concern ?

 

[Riccochet]

Well then my question would be, When would mounting issues be a concern ?

IMO, it's always a concern. You never know if there's an air bubble in there until you pop the HSF off. Remounting the HSF is the easiest way to rule it out. If there are still temperature variations between cores then the problem lies with the operating differences between core pairs or sensors not reporting properly at lower temps. IHS contact with the cores would be another issue, but there's no good way to figure that out without popping the top on the chip, and nobody wants to do that anymore. AMD's were a little more forgiving when popping the IHS off.

 

[jimbo6846]

Not even. Big variations between the temperatures of different cores are generally not a sign of improper mounting, but merely that the chips simply don't run at the same temperature. Intel quad-core CPUs frequently have differences of ~10C between cores, and it's no cause for concern.

While true some of the time, it's also true some of the time that remounting the HSF will improve temps and lessen the gap between the core temps. It's also true, some of the time, that surfaces of either the HSF, IHS, or both, are not completely flat and you could improve your temps by lapping them, in order to improve contact between them.

 

[Zero82z]

Well then my question would be, When would mounting issues be a concern ?

Only if all the temperatures are out of whack. Improper mounting would cause the entire CPU to run significantly hotter than it should, which is pretty easy to spot. However, variations between different cores at normal temperature levels are nearly always caused by the differences between the cores themselves rather than any sort of mounting issues.

 

[jimbo6846]

Only if all the temperatures are out of whack. Improper mounting would cause the entire CPU to run significantly hotter than it should, which is pretty easy to spot. However, variations between different cores at normal temperature levels are nearly always caused by the differences between the cores themselves rather than any sort of mounting issues.

An improper mount could case wider variations between the cores. I've seen it happen, and I've seen it improved by remounting the HSF. That isn't to say it's the only cause, but it is one of the causes.