questions about my intel 2500k setup

[Anarchist420]

First off, my room temp is about 61F (~16C) now (it's cold outside, but the heat isn't on). 2nd, I have a cooler a little bit better than the stock one and I used a little bit of AC MX-2; I probably should've used more. 3rd, my cooler is running at full speed (it's 92mm hydrowave bearing and full speed is ~2.4krpm). Finally, I have EIST off and I set the ratio from auto to 33. I have turbo disabled.

My CPU is running at 25C (everything, not just the cores) which is about 9C more than the room temp with winamp playing, IE, core temp, and AIDA 64 running (with ~11W power reported under core temp). The reported core temps are always changing, they go from about 25-39C (I'd say they've averaged around 33C over the past 2 hours in a room with constant temp of 61F).

Now my questions (I know that idle temp doesn't matter to some people, but it does to me):
1. Is coretemp reporting the right power usage?

2. Are my temps about right for ~11W usage reported by core temp?

3. If I were under a load where coretemp were to report 90W usage, then would my temps still be okay? I got it up to about 48W (everything, not just the cores) using the software mode in PCSX2 and the max core temp was 49C and the minimum was about 41; the cores averaged about 46C during that 48W session).

4. Is it normal for the whole CPU (everything, not just the cores) to be 9C hotter than room temp under light load (as described under question 1) and 1.22V?

I'll really appreciate any answer(s), because I have no clue how good/bad these temps are given the conditions. I'm asking because I want to make sure before I take everything out and take the HSF off, apply more AC MX-2, put the HSF back on, and the put everything back into the case and plug it all in again.

 

[dma0991]

1. No, unless you're measuring with a multimeter and something like a Kill A Watt device to measure global power consumption.
2. Hard to say. Too many variables involved and temp vs power consumption is totally nullified when you're comparing a chilly winter versus a hot summer as ambient temps would affect your idle and load temps.
3. It is a 95W TDP processor, 90W is not an issue and it should not even be near 90W as the 95W TDP accounts for the IGP as well.
4. It is an electronic device that will emit heat constantly. I will definitely be surprised if it would have the same temps as my ambient temp.

Your setup is good enough and needs no changing or probably slap on a Noctua NH-D14 and take advantage of the cold ambient temps to push the chip to 5GHz. With TIM, less is more and you shouldn't put anything more than a grain of rice.

 

[Zap]

1) IDK
2) IDK
3) 49ºC under full load is really low.
4) Probably. If you want lower idle temperatures, you have to re-enable EIST. With a Core i5-2500K, I don't know of any reason why you would want EIST disabled. You say idle temperatures are important to you, so enable EIST.

 

[Anarchist420]

Thanks to both of you🙂 I was worried I didn't put enough AC MX-2 on.

I actually didn't want EIST on because I wanted it to be at full speed all the time. I just didn't want the clocks constantly changing, something about me I guess.

 

[Wuzup101]

Thanks to both of you🙂 I was worried I didn't put enough AC MX-2 on.

I actually didn't want EIST on because I wanted it to be at full speed all the time. I just didn't want the clocks constantly changing, something about me I guess.

You won't notice the clocks changing - I would enable EIST....

Edit: and turbo... enable turbo... these technologies are there for a reason...

 

[Zap]

I actually didn't want EIST on because I wanted it to be at full speed all the time. I just didn't want the clocks constantly changing, something about me I guess.

There is absolutely no reason to not have EIST and Turbo enabled. There was in previous CPU generations, but usually only when overclocking really high. With Sandy Bridge there are no issues.

 

[JumpingJack]

I will take a crack at answering your questions.

1. When Intel introduced Nehalem and dynamic clocking (Turbo Boost -- how I hate marketing crud) the Intel processors have had the ability to read/monitor current. Intel's dynamic clocking is controlled by the PCU (power control unit), a microcontroller within the processor that measures both power (of which, you must know current) and temperatures such that it can dynamically adjust the clock speeds.

Per VRM spec 11.1, http://www.intel.com/Assets/en_US/PDF/designguide/321736.pdf, the VRM is required to send a signal, Imon, the produces a voltage proportional to the current supplied to the processor. From that the processor can ascertain power. This is how Core Temp (or AIDA 64 aka Everest) calculate power, i.e. power = current * Vcore. In the spec you will see in figure 5.1 the diagram for Imon, the software simply reads the MSR register to obtain the current that the processor records. Reading through the specification some more, you will find motherboards must use VRMs that report Imon to within +/- 15% accuracy, but at low currents that accuracy balloons to +/- 50% at 15% of Imax (see page 39).

This does not necessarily mean that the reported power has this much error in it, rather it means that is the max allowed error for the VRM to be spec compliant. However, it is an indication that the accuracy can be off by quite a bit. In short, it really depends on your motherboard and the quality of the VRM to report Imon.

I have profiled power consumption of SB CPUs by measuring the amps pulled along the 12 volt rail (as well as all the other voltage rails when the CPU is under load). My experience is quite positive on the reported power by AIDA 64, it is always in good agreement (I use mid to high end Asus boards in most all my builds and they typically have pretty good VRMs).

2. Your CPU temps look fine. The absolute temperature is a function of the efficiency of your HSF, thermal contact to the integrated heat spreader, and the temperature differential to ambient at any give power load. When considering CPU temps, the only real worry about temperature is at full load and how far away (or close to) it is Tjmax (maximum junction temperature). Intel CPUs typically have Tjmax between 90 C to 105 C, so 30 or 35 C is no sweat. Even if you hit 65 or 70 C, no big deal you are well below the thermal limit of your CPU.

3. See my response to 2 above. The reason you must cool your CPU is an interesting one, but beyond the scope of this discussion/forum. It requires some study in solid state physics, however, I will briefly describe the end purpose. Silicon, a semiconductor, has unique properties. But lets start with a metal -- in metals, conductivity usually goes down as temperature goes up... this is due to phonon scattering which in simplest terms can be described as interaction/disruption of the traveling electrons with the vibration of the atomic cores about their lattice positions within the material. Silicon behaves differently, it is nonlinear in the conductivity response to thermals and has a very sharp transition to strongly conductive at a certain temperature. This is because at a certain temperature electrons take on enough thermal energy to hop across the band gap and populate the conduction band. When enough electrons are in the condition band, silicon will behave kinda like a metal... i.e. rather than being a semiconductor it is like a copper wire or you can think of it as infinite leakage. This is catastrophic, and in CMOS devices it is called thermal runaway. When you hit a thermal runaway condition, current goes up drastically at some fixed voltage and as such so does power -- i.e. transistors no longer switch on or off it is just a straight up short, and it is a catch-22 situation, the more current it flows the more power it produces, the more power it produces the hotter it gets, the hotter it gets (unlike a metal) the more conductive, the more conductive the more current... and so on and so forth, to the point that it simply would melt all the solder and PCB (remember this video: http://www.youtube.com/watch?v=QIsr1R1qy1Y -- in this case, Intel had thermal protection circuits to shut off power at a certain temperature, at the time AMD lacked the thermal cascade protection and you can see it melt, it is quite spectacular).

4. Yes it is normal, 9C is not a huge differential. If you use a larger HSF and faster fan speeds, the temp would likely be lower, but only by a few degrees. If it were 50 C delta I would be worried, or even 20C, but 9C nahhh, don't bother. If it bothers you, you can try re-seating your HSF -- use only a dab of TIM (thermal interface material), it does not take much, and you can either mount and let pressure spread it, or pre-spread it yourself, but the most common mistake is putting too much TIM and that ruins the thermal contact of the HSF to the IHS.

Hope this helps, you got good answers before me as well....

Jack