I recently purchased an OCZ Vendetta for my Q6600 to replace my stock HSF so that I could run it at 3.0Ghz while keeping the heat relatively low.
Currently, it's running at 40-45C idle and 55-60C load which i believe is safe however not the jaw-dropping results others appear to be getting. I believe i used plenty of thermal compound however it is a little hot in the room where i keep the computer being that it's summertime. Thinking about installing an AC in there though.
Speed Step can also help, although with really High OC's, Speed Step can make it unstable from voltage fluctuations.
I have mine OC to 3.2ghz. I use the Tcase for my main readings. At idle, my Tcase temp in 80-85F ambient room temps, is usually around 33-35C. With the ambient temps in 75-78F I've seen my Tcase temp around 28C.
Using prime95 for putting a load on the CPU, the tests it does will vary in temps. The highest Tcase temp I've had so far on my P35 MB, is 61C, and my room temp was 85F.
The temps i'm implying are the temps i got from Real Temp 2.70. This is what i'm running. I'm assuming that the ambient room temperature in my room may be a bit high which may have an effect on the cpu temperatures. As far as sensors go, i'm assuming the sensors would be the sensors which Real Temp is checking to provide me with data regarding the temperature of the CPU Cores.
I believe i have decent airflow running through the system, 1 fan in front blowing in, 1 fan on the side and 2 fans in the back blowing out. i have the fan from the OCZ Vendetta facing towards the fans blowing out.
I think what i need to focus on right now is getting the room temperature down which will help in my situation.
Okay, you are then going by what would be called tjunction temps. To take the time to explain further, basically a Quad would have 5 sensors.
Tcase - IHS sensor (used by the Bios for health monitor, by speedfan shown as CPU or temp#, or by MB utility app that shows cpu/system temp)
Tjuctions - sensors on the die for the cores themselves, or known as DTS (digital thermal sensors).
Any program, (realtemp, speedfan, coretemp, hd monitor, ect) depend on a tjmax to determine a temp per core. This tjmax is basically a guess by the software maker, and is why you would get different readings from different apps. Some of the programs allow you to offset the tjmax. It is interesting to have a program that gives temps per core, but to me, its not going to be reliable. Even using real temp explains that idle temps will not be accurate:
You should be able to see that message when you click on test sensors.
Now there was suppose to be an article for IDF I believe to explain more on why Intel didn't publish the tjmax, and to better understand how we could get better or more actuate temps on the cores.
To complicate this even more, I went from a MSI P6N Platium MB (NV 650i) to a DFI P35 DK T2RS MB, and the temps are actually acting different, except the Tcase sensor. Normally Idle temps on the Q6600 for the Tcase were about the same, 30C. The cores however, seemed better on the P6N to use a tjmax of 95C, while now I have to use tjmax of 100C on my DFI P35 MB. If I used 95C (which realtemp specifies by default) my core temp readings are lower then my Tcase, which in my mind is wrong.
In the past, you could take the same cpu, and put it into different MB with different temp readings. That was because there wasn't a standard to determine the cpu's temp. Now that Intel wants to perhaps make a standard, even with my Q6600 in 2 different MB, the core readings are different when I was thinking the tjmax 95C was more acceptable.
So, in conclusion, this would be the main reason why I go by the Tcase temp. Also, Intel does publish the thermal spec of all their cpu's. The Q6600's thermal spec is 71C. Its much more simple to go by 1 temp, then try to make sense of 4 temps to understand if your CPU is too hot. Since I over clock some (I not going to go to extremes) what is important to me is not to get to close to 68-70C for the Tcase temp.
Now to explain that HS less important vs Ambient temps, I had an opportunity to do a stress test with an ambient temp of 75F.
My system was being stress test on Prime95, small FFT. The 800000 iterations are what cause the most higher temps on the cores. So after an hours of reg stressing on the small FFTs, I restarted it to run the 800000 iterations over again. This does 11 tests, then the tests seem to get easier, which in turn lowers temps. Here is an example of my previous stress test that is not 800000 iterations:
As you can see the blend test doesn't quite heat the CPU up. The ambient temp was around 80F.
Here is the way I measure my ambient temp:
As you can see its about 75F.
Now this is the small FTT test (FPU max stress) on the 800000 iterations:
So its not exactly all about how good another HS maybe, but how well you can give it cool air.
Oh BTW, look at the fan speed. Not even at its max of 2500rpm, which Dragon seems to be annoyed by its loud fan.
I did install an AC into the room and it dropped my temps significantly as reported by Real Temp software. 35-40C idle and 45-50C load. While i was playing Bloodymare Requiem, it never went higher than 52C and while i was monitoring the temps, it stayed pretty much around 43-47C.
Again, thanx for taking the time to post this information.