false temps?

hi first of all been reading these forums alot as of late, its nice to see im not the only one with the 6920g(i love this thing btw) lol

ive been reading a lot on under-volting on here and temps and i finally decided to dive into it. my only problem seems to be is whether or not im getting accurate temp readings from my programs, it seems every one i use has a different opinion, ive tried real temp, hw monitor, the temps in rmclock and even some other ones i no longer have that i found around the web. the thing that scares me tho is realtemps readings(which i get the feeling are wrong but not when my hands are on my acer)

realtemp comes of with 49c-54c idle and 89c-94c at max load on orthos
but what throws me off is my rmclock and even hwmonitor come up with 34 idle 56-60 max load, which i prey those will be right lol

what bothers me tho most of all is this is after under-volting.

has anyone had innacuracy with one more then the other? i need to get the temp down cause i cant even play games anymore due to the computer randomly shutting off after 20-30 mins of gameplay(wow,aion,crysis)

im almost tempted to going as far as strapping my fan to the bottom of my desk and assentialy turning my desk into an airhockey/cooling pad for my acer.

 

Wow! You just gave me a great idea for my next desk. I have to find a small air hockey table ...

 

If your computer is shutting off due to it overheating, then the load temps of 94*C are much more likely to be correct.

The differences in temps may be due to the different methods of calculating the temp used by different programs

 

loxel: You didn't mention what CPU your laptop has.

Intel CPUs contain a reverse thermometer that counts down towards zero as the CPU gets hotter. A direct reading from that sensor is reported as Distance to TJMax in RealTemp. When that number counts down to zero, your CPU will start to thermal throttle and slow down to try and control its temperature. The thermal shut down temperature is still 25C beyond the thermal throttling point so this should not happen until the CPU hits 125C to 130C depending on the CPU model.

Here's the formula that all software should be using:

Reported Temperature = TJMax - Digital Sensor Reading

TJMax for the newer 45nm CPUs like a T8100 is 105C and TJMax for the original 65nm Core 2 mobile CPUs is usually 100C. If I had your CPU model number I could look it up for you here.

Processor Spec Finder

It's possible that the cooler is not making good contact with your CPU. You might have to pull your laptop apart to make sure the heatsink is snug. Might as well replace the thermal interface material while you're in there and see if you can shim something up so the heatsink isn't flopping around.

RMClock was released before the 45nm Core 2 CPUs so doesn't fully support them and HWMonitor is likely wrong. If RealTemp is wrong, I'll buy you a beer.

Edit: Here's an example of HWMonitor using a TJMax of 100C for a T8100. By using a TJMax that is 5C lower than the actual TJMax, the core temperature is also reported 5C lower than what it actually is. 39C vs 44C

 

The critical think when checking temperatures is to look at the trend.

Set your monitor prog to run at startup.

Check the minimum temp it captures when it starts

Is that minimum room temperature or a few degrees above? If not readings are clearly wrong

Does it stick at one precise value or move in big steps. If so it is going to be wrong and reading fan speed or something and will clearly be wrong

If it starts at room temperature, trends smoothly, bears a logical relationship to fan speed increases etc THEN start worrying about the temperature it shows

 

This is usually true but you need to keep in mind that the sensors Intel uses are not calibrated or very accurate at idle. The core temperature sensors are calibrated to trigger thermal throttling and thermal shut down and tend to lose accuracy and start reading either a little high or a little low at idle.

Some Intel sensors also have a problem with getting stuck at idle. They can be very accurate when the CPU is warmer than the sticking point but completely useless when the core temperature goes below the sticking point.

At least Intel documents TJMax for their mobile CPUs so once you know that, it is easy to use the formula I posted to determine the core temperature. It's obvious that HWMonitor is ignoring the Intel documentation.

 

Yep. Seem to started a bit of a buzz on this topic with loads of people saying their readings are different /non-credible and others saying they are confident theirs are right. Of course everyone may be right to some degree (pun) because every system has its own tolerances and quirks and the temperature of the CPU /fan speed even running the same benchmark on the same model is bound to vary from unit to unit and with ambient temperature etc.

I am using a well cooled non acer desktop system where SIW reports one core 5c hotter than the other (a virtual impossibility given the thermal conductivity of each core to the massive cooler) and it also shows a ghost 'AUX' temperature of 26c at all times. . . so I tend to ask myself if any reported reading is credible at boot and if behaves credibly therafter. In the above example i am happy that core temps right within 5c are useful for checking over time . . and I am going to remember not to include the AUX temp in any later investigation of heat problems.

I have a healthy suspicion of new and un-tested car fuel gauges too after a long cold walk a few years back but that is just me

 

You've discovered the other problem that Intel sensors have. Intel does not guarantee that TJMax is set equally for both cores. They like to refer to this specification as TJ target. It's a number they shoot for but there is no guarantee what percentage of their CPUs hit the target TJ number they specify.

I find Prime 95 - Small FFTs to be the best tool for testing. It equally loads both cores. If two cores are reporting different temperatures during this test then it's usually a good sign that TJMax is not consistent.

I got very lucky with the T8100 I recently purchased. The reported core temperatures track each other very well from idle to full load and when first started up in the morning, after doing a resume from stand by, the reported core temperature is a couple of degrees higher than the room temperature which is also very reasonable and fairly accurate.

Most Intel sensors are not nearly as good. People put too much faith in the numbers some of these sensors put out. They also argue too much over what the data means that these sensors sometimes display. Many sensors are not even close to accurate and are not worth arguing about.

 

Yep I think we agree on all of that I'm just trying to share a bit of sunlight around on this 'dark' issue.

My other love is desktop gaming systems and my infra red thermometer gives me the final view on the heatsink/ cooler and airflow temperatures . . as a 'credible' hot chip in the sensor reports and yet a cool heatsink/cooler indicates either sensor inaccuracy (see above) or mechanical or thermal compound failure.

 

i want to say thank guys to all the quick replies since i first posted you shed alot of light on my problem i will make some progress tests with each program to see which is the liar, im starting to believe realtemp more and more considering the actual reaction of my computer to the heat and the steadiness of its temp climb,

Sorry for noobing it up here im more of a hardware modder ( soldering irons are my friend) i noticed that tjmax thing and its good 2 know what its about now , i was worried it was a "gonna catch on fire" countdown which scared me into shutting orthos down a few times early when approaching it

i did a vent clean recently but i should certainly take a look at those heat sinks which i have yet to have a chance to do a deep clean, i will probably do it when i plan out my internal xbox controller receiver mod ( im sick of it and its long dangly chord ) i need to open my acer to see what space i have to work with anyways.

also sorry for lack of info my cpu is intel core 2 duo t5750 2.0Ghz,4gb ram with geforce 9500m GS.

p.s. my computer started out with vista and had throttling issues but when i got win 7 they seemed to go away, at least for games as far as i know. has anyone else had this happen? not complaining lol just interested.

 

Intel lists the maximum safe junction temperature of a T5750 at 100C.

Intel® Core?2 Duo Mobile Processor T5750 - SLA4D

Make sure RealTemp is using that number and it should report some fairly accurate temperatures. The sensors on these CPUs tend to be significantly better than the later 45nm CPUs.

If you are using ThrottleStop you can add this to the ThrottleStop.ini configuration file:

TJMax=100

and then this program will change the DTS column to the core temperature.

Some of the Acer laptops use chipset clock modulation which can slow a CPU down internally. Run ThrottleStop with the Log File option enabled while gaming and it will show you if this is a problem. Both types of clock modulation should be steady at 100.0%.