This is not an overclocking system, I just want maximum longevity.

CPU: Stock 3.4GHz
Turbo and HT: Disabled.
Cooler: Noctua NH-U9B SE2
Ambient temperature is 24.5°C (~76°F).

Getting 60°C at 100% load (Prime95 large FFT). Tried it with 2 fans both on Noctua silent adaptors, and then with 1 fan on the motherboards dynamic fan control. Temps were pretty much identical and barely any difference in noise level.

Is 60°C good, or should I try and lower it?

Why did you disable Hyperthreading and turbo? you should have gone with 2500K if you didn't want HT, would save your $100.

On my 2500K i usually get around 50C on load now that i lowered the fans, 60C doesn't seem high since you using a smaller heatsink and i dont know the rest of the system (other fans and components inside the computer). Nothing to worry about the temp, i would check on summer also to see if its not going up much.

It is summer!

The case is a Silverstone FT02 with the three big fans (original non-AP) set to low and the top fan removed. There is a GTX 285 in there at the moment, soon to be replaced by an HD 7970. I haven't tested cpu temps with the GPU at full power - it shouldn't affect the CPU much anyway, thanks to the case layout.

I had a look at the following test, and I guess my temps aren't that unusual.

http://www.behardware.com/articles/824- ... olers.html

The Noctua NH-U9B SE2 is just not a very high performer. I have a Hyper 212+ that I could swap in, but I like the Noctua build quality and mounting system better.

By the way, I'm using the ultra low noise adaptors (blue), I forgot there were two different types.

Hi, 60°C under Prime 95 sound pretty reasonable to me for 2600k. I tested one, at stock with Turbo and HT on, so 8 threads at 3.5GHz running Folding@Home, 100% but probably not as heavy as Prime95. Had it under a Cooler Master Hyper TX3 which is a 3 heatpipe 92mm fan tower cooler not far in size from your Noctua, was getting temps of 66°C with single stock fan ~1000rpm. I believe your stock fans with ULNA run this sort of speed. Also got temps of 67°C with a pair of Arctic Cooling F9 PWM fans at 650~700rpm, that was quieter, I think at the lowest fan speeds two fans help much more than at higher speeds.


This was open motherboard sitting on table, the two fan speeds listed are the two F9s

Seb

60c at load is not bad at all and perfectly safe. It doesn`t seem too unusual considering the relatively small size of the cooler, the slow speed of the fan and lack of a case exhaust. The hyper 212 could lower your temps a little but I don`t think it`s worth the trouble as they are fine to begin with.

+1


Also:

You should turn HT back on IMO, Its a shame not to have it on when you have a i7 2600K. Thats like having a i5 2500.

About Turbo, even with turbo disabled you are still getting speed changes only your max speed is 3.4 GHZ. Still fast for any takst, but having it on will give you an extra bump at no cost. It is perfectly safe. The K models are graded to do 4.5+ GHz, so you shouldn't worry about the defalut 3.8 single thread turbo at all.

I disabled HT because I've heard reports of it affecting games (one example is that it allegedly causes stutter in BF3). I haven't done any in-depth testing myself, but since I'm mostly interested in older single and dual threaded programs I disabled it as a preventative measure. I rarely have use for four threads let alone eight, and I never saw any improvement with HT on. My benchmarks showed no difference.

As for turbo, I'm not worried about it damaging anything. But I've been experimenting with it enabled, and it's operation appears far too erratic. What happens during gaming when the CPU is adding or dropping 100-300MHz on a whim?

Without turbo, c-states and EIST, the CPU is always at 3.4GHz. (although I don't really have any issue with c-states/EIST)

You are getting a 35.5C temp rise with a newer improved version of this heatsink... with an additional fan:
http://www.frostytech.com/articleview.c ... 270&page=5

Something is just not right. This cooler should do much better than that. Why don't you try reseating it with the Arctic Silver thermal compound that doesn't require a break in period.

The temperatures are definitely higher than they should be - I have temps in the upper 60s under load with a very small Scythe Kozuti on a 2500K. Yours should be much lower with that Noctua.

Does this look like a PC case to you ?
http://www.frostytech.com/testmethod_mk2.cfm

I don't think so - and that is exactly why can't you just take values from any site not testing coolers in PC case and call his temperatures higher or "not right". First of all his CPU fan(s) is(are) running at 1000 RPM (ULNA adapter) instead of 1600 RPM. He doesn't have the outtake fan above the CPU cooler either. Considering the low airflow Noctua fans produce, combined with the fact that his fan speed is lowered using the adaptors makes his temperatures just about right.

I guess you are going to say my 40-44C idle temperature with my i7-2600K is "not right" either - not taking in account that all my fans run at 700RPM. Real world experience is that people are getting around 60-65C with i7-2600K with silenced computers (lower airflow, lower CPU fan speed).

To be fair, with the FT02's positive air pressure, there is probably just as much air moving through that grille as when the outtake fan was installed. But yeah, that article doesn't seem particularly relevant.

I have a 2600k in a general purpose/web browsing system, so I ran the same Prime stress test:

CPU: Stock 3.4GHz
Turbo/HT: Disabled/Enabled
Integrated Graphics: On, but not under any load.
Cooler: Hyper 212+ with AC F12 Pro fan on mobo auto control.
Ambient temperature is 23.5°C
Case: Lian Li PC-A05

54°C max.

I have a FT03 which is very similar in airflow to his FT02 except i don't have the intake bellow (the fan didn't fit in there with the Mega Shadow). As stated before, very much depends on the fan speed, and when he uses ULNA he has only 1000RPM 9cm fan blowing through his heatsink, which means not lower performance.

Then we have the difference between what frostytech measures and what can you measure. Frostytech measures surface temperature, but the applications on your PC report Core temperatures. There can be huge difference between these two values, and the only way to measure the same value is to drill a hole for a temperature sensor in CPU IHS, which none of us will do .

And then there can be a huge difference between each piece of CPU - one CPU can overclock to 5GHz, another has problems with 4.4GHz. One can have higher temperatures than others, easily even by 4-5C.

In the end you have so much variables that you just can't say "there is something wrong with that temperature" unless you have exactly the same setup and have more than 10C temperature difference, or unless he has 75C+ core temperatures.

That still seems like too great a difference. Two fans on slow should move at least as much air as one fan on high. I would actually expect them to move more... especially back to back in a tower cooler. Also look how effective the heat sink is at a lower 85W wattage. The 65 watt 2600k just doesn't generate a lot of heat... even running full bore.

Why would you not try reseating the heatsink, and doing so with a thermal compound that does not require a burn in?

I would not expect that much difference. Where exactly is the heat going to leak off to? What magnitude of difference would you expect?

Well the Silverstone's case has very good airflow and Intel cpu's have so good power management that the CPU doesn't warm up that much when idling.

My i7 920@3.2GHz is idling at 28C with Noctua NH-U12P with the golfball exhaust fan mounted on the heatsink, at 5V. With Prime95 temps are around 60C. Those 3x 18cm fans are around 600rpm.

I wouldn't worry about temps as long as they stay under 65-70C. If you want lower temps you may want to try watercooling. For example Corsair sells ready to install 120 kits where you can mount the rad&fan to the exhaust hole in case.

Well, for start you are comparing numbers measured using calibrated tools (frostytech) with numbers measured by sensors in CPU, which are not meant to be 100% accurate. For example, how do you explain a 4-5C temperature difference between different cores on a completely idle system ? Explanation is simple and i already said it - temperature sensors in CPU are not meant to be 100% accurate. And let's think about it a bit more - imagine his CPU sensors are all off by 5C compared to real temperature (possible situation) - and you are at 30C over ambient instead of 35C. Add to the fact that his system can have a hot spot around the CPU cooler area, or his motherboard, memory and VRM emits extra heat not accounted for in frostytech review, and you are again a bit closer to frostytech number.

And then we got to that 65W number. i7-2600K is a 95W TDP CPU, and at full load it easily reaches this number. Or how else do you explain that my system goes from 95-105W idle power usage (from socket) to ~200-210W power usage when running OCCT Linpack ? With PSU efficiency taken in account, we talk about 86W in idle and 182W in load - that is about 96W if i calculate correctly. Not far from the 95W TDP, and we can account for the bit of extra power with VRM and bit increased system power usage.

Simply put - depending on your system cooling, ambient temperatures and other factors, temperatures bellow 70C are fine for stock i7-2600K system at full load.