Calibrating My CPU Core Temps With Real Temp
Helped Me Worry Less And Love My PC More
I’ve been running Core Temp for many years now on my Core 2 Quad that I bought in mid-2009 – and through this, by looking at the temperatures I’ve ditched the stock fan, grabbed an aftermarket fan from my local MicroCenter on a last minute decision (which I read later wasn’t suggested for a Core 2 Quad… at all) and now am running one of the best rated Zalman air coolers – that would fit in my case/motherboard/RAM config. [ED; It was much more expensive than the price listed at NewEgg – but I luckily got it free when the MicroCenter employee accused me of stealing something from a motherboard box when I was looking at the clearance to see if the fan would fit. No one wants a CPU fan/heatsink that doesn’t fit – and the specs on sizes aren’t given for the area around your CPU on most boards… Of course, I wasn’t stealing so the manager ended up sending me away with the fan free because he could see I shopped there instead of online for most of my builds and buys.]
So it took a while, but I got the new air-cooler in. As one NewEgg reviewer rightfully noted, the way the heatsink lines up with the brackets is a bit annoying in some cases. Of course, I’m one of those lucky ones. And it really didn’t drop my temperatures at all. I actually broke aftermarket fan #2 because it annoyed me, and ran the “stock” Intel cooler for a while until I was basically given my current CPU fan. I used three types of thermal grease, and after digging around for the Arctic Silver in a box I just ran out to Radio Shack about a year and a half ago when I got annoyed with seeing high temperatures after a long World of Warcraft raid on a chilly night. The idle temperature, with this cooler, should have been sitting around 25 considering it’s in the 45nm CPU line from Intel and the ambient temperature, in Celsius, was around 16 or 17 degrees [ED; I note later where someone has tested direct CPU core temperatures with a thermocouple and given guidelines on base temps]. I’ve read quite a few times the CPU could be a hot one – but the temperatures I was getting didn’t make sense.
And frankly, at one time, my temps were low – when I first built the machine in the earlier part of 2009.
As a word of advice: Intel tests their chips against stock coolers, the ones Intel makes and ships. If your CPU is “hot”, without overclocking, and you are still rocking the so-called stock cooler – get an RMA. As much as you hear that “stock coolers suck” online – they are pretty much approved by the engineers who made the central processing unit in your PC. Intel will ask you to RMA if they suspect the temperatures are too high- though the temperature of the cores themselves on CPU’s before the i7, i5 and i3 are ignored by Intel 100% of the time and ~95% of the time on those mentioned. If you are concerned, contact them, they will help you. However, they are concerned about the “main” CPU temperature – or the T-Case temperature – not the cores. In fact – Intel doesn’t even actually measure the core temperatures on the chip – in fact software that gives you a temperature derives this value from it’s distance from the “Junction temperature” or “Distance from TJ Max” (wikipedia basic info / Guide to Understanding Intel Temperatures – very informative, helpful, in-depth article).
So we have SpeedFan, Core Temp, HWMonitor (from the CPU-Z/CPUID folks) and the acclaimed Real Temp by the true overclocking engineers/gurus/nerds. Which to choose? Honestly, I liked Core Temp because the icons in the task bar and the window itself was pretty – it integrated with Windows 7’s taskbar nicely (coloring the bar on load or temperatures) and – why not use it? They all were giving me the same readings.
But I was watching the temperatures again at idle and they were just too damn high. And the machine is in a room that’s closed off, has a window air unit that rocks – and the unit even has to be turned off at times (we don’t use central in this room, it’s inefficient where the room is in the house). Needless to say, even with it being 101 degrees Fahrenheit outside (38.333° C) it was much cooler in here (closer to 17° C most of the day – and night). As a qualified forum user has tested – I shouldn’t be seeing temperatures as high as 56° C on one of 4 cores with 1-2% CPU usage. While another was in the 40’s. For a few months I had the machine overclocked to over 3 Ghz because it actually didn’t raise the temperature and it was still crash-free (my longest uptime in Windows 7 is 26 days, 44 minutes and 58 seconds – while overclocked in every regard – just last October to November, 2011).
And the breaking point was having a buddy come over to geek out over Android phones and install ROMs and when he saw Core Temp running in the tray – almost the first thing he noticed in the entire room – he was concerned. I had been for a long time, but it seemed there was nothing I could do. I know how to force airflow through a case optimally. I’ve tested that flow with smoke to ensure it’s travelling as intended and have cleared out small animals from the case when I could. I’d even gone as far at one point to toss out totally functioning case fans and buy new ones that promised to push more air (and, I’ll admit, the light up brighter than the last ones, so that was cool). I replaced an exhaust vent with an exhaust fan, I reversed it, I installed more fans, less – both directions – all configurations, I stripped a few screws in the late night hours because everyone seemed to be telling me “those temperatures aren’t right”.
I started to wonder if I needed to put the case on it’s side to make a better contact with the CPU heatsink and the CPU itself. Which after due diligence in using some Google-Fu it was entirely a possibility – no reason not to except for reasons from people who just were making stuff up in forums – just don’t put a monitor on the now-flat tower and keep those case fans in mind, etc. In my configuration, it might have helped for other reasons actually (look at the back of your PC – do you have an exhaust fan below your power supply, and it’s own exhaust fan…).
And you know what – everyone seems to have been wrong. I’ve run the machine at full load and hardware never kicked in and shut it down (the motherboard or the CPU itself). Only software like Core Temp made it reboot (using Multipar, though never with Prime95). I never had a lockup or blue screen – that wasn’t caused by my general tinkering with the BIOS settings such as clock speeds and multipliers that weren’t kosher. Though, I can save 10 BIOS profiles to the CMOS, the motherboard detects an OC error (such as a boot loop, etc) and will boot you into setup and then ask you to load defaults, adjust, etc… so I’ve had fun with that. Most of my crashes have been bad software (since I started logging in November, 2010, when Windows 7 decided to barf on loading a system file that clearly existed and I was busy working to fix it).
As I mentioned above – those core temperatures don’t really mean anything – because they aren’t even actual readings of temperatures. Yes, they are derived values – but it turns out the sensors on the cores – well, Intel, the largest chip maker in the world, just didn’t use the greatest parts. Probably because it doesn’t matter. As mentioned above, they are concerned with “hot” CPU’s, and they’ll replace them. But they are worried about the heat of CPU casing.
So how did I save my sanity? How did I stop the constant nagging reminder that my CPU naughty-parts were running at close to 150 degrees Fahrenheit just by turning it on (while the chip’s case was running at less than room temperature according to the BIOS?). Computers are as fickle as modern cars (which, of course, run on computers) and in the past you could literally make an IDE cable go bad forever because of the strangest occurrences. In my last quest to check my CPU/heatsink mounting and do some routine case cleaning I somehow rebooted a machine without any hard drives detected. C’est la vie they say.
I finally decided to dig deeper, and ditch Core Temp for Real Temp. As I said earlier in the post I didn’t care which I ran since they both gave me the same readings. The difference is that with Real Temp you can run a heat-up and cool-down test with Prime95 and calibrate your baseline temperatures at full load (distance to TJ-Max – not the discount store) to see what your true results should be at 50% – as well as then look at how you should adjust your idle temperatures due to what is aptly called: a stuck sensor. Once you can find the core that reports the sanest temperature, re-run the test again after adjusting the settings of Real Temp itself for the TJ Max values (adding or subtracting from 100 in Core 2 Quad cases for example) to get a better perspective of how poorly the idle temperatures are being reported. Idle temps seem to be what bothers most people – and it was bothering me until yesterday.
Essentially the quick and dirty of this is;
- Install Real Temp and Prime 95
- Set or make sure your PC’s CPU clock and multiplier are at the stock settings, double check your voltage is matching Intel’s spec (or go a tad higher – you may want to Google your CPU model if you get a freeze in the test).
- If you have never overclocked, ignore this – if you never touched the voltage – you probably want to ignore this as well. In fact many lock-ups during testing can be due to too little voltage to the CPU, not a damaging effect as opposed to too much voltage
- Run the Sensors Test
- Adjust TJ Max values in Real Temp settings (bottom right button in the main window)
- Possibly – re-run the test again to see if one of the sensors is just plain stuck/broken and should be ignored – and to ensure you’ve adjusted in a way that makes sense – are all the temps within a degree or less of each other across the cores?
- Re-run the test now that you feel as if you have adjusted your TJ Max settings in Real Temp accordingly – this will help you determine your idle temperatures. Only after making the first run (or two, or however many you’d want to test with) and setting the TJ Max will you likely see that at low usage or no usage you’re CPU is not cooling any more.
- Note; read through the Real Temp documentation linked after this list to see what you should be looking for as far as idle values. If you’ve mounted your heatsink correctly, and your idle temperatures are close to 25% usage – or just not dropping after a certain point, one is unusually high, use the notes from rge on the XtremeSystems forum as a guide. Ambient case temperature and cooling types determine what you should be looking at at idle – more than anything else.
In reality there are a few myths still out there about the Core line of Intel processors – that all four cores will vary in temperature at idle and at full load. Maybe with affinity set in Windows or certain programs using just a core – but not with an even spread of tasks. If your values are all different when you are at 0% cpu usage (small window on this machine!) a sensor is stuck. And only by testing it are you going to be able to calibrate for the broken sensor. And the other myth is that core temperatures actually matter. In reality, they kind of don’t after reading around the interwebs. Don’t get me wrong – you might want to watch them, or you might want to shut down a machine before it gets unsafe, or just see if your liquid nitrogen setup is working as expected – the core temperatures matter.
But really – there is a good chance you are fine. For those of you with rotating house fans and window air units built into the side of your case – remove them, calibrate, and see where you come up. You might be shocked to see you’ve been mislead by noobnish the whole time because you didn’t actually know how the chip worked.
I can’t even claim this is “news” – I just accepted the lie for so long I started to believe it. My Intel™ CPU lied to me. I was too busy bash scripting and other miscellaneous PC hobby/work stuff that I stopped even caring until Ty mentioned it again. So don’t walk, run, and download Real Temp right away.
Step by Step, With Charts and Pictures