Q9450 58C Idle, Whats wrong ?

I don't spread the MX2, just a decently sized dab in the middle of the proc works well..push the HSF down, give it a couple of small twists a few degrees both directions, push the clips in or screw down and go. I've found good spread and seating this way...with AS5 I would spread evenly tho...I don't think it matters as much with MX-2 tho.

Also that is a 45nm chip right? It's a C1 stepping, Could have bugged temp sensors...how warm is the HSF under load?

I use the same Zerotherm HS and it blows through the heatsink as you would expect. If for some reason is it blowing away from the heatsink you can remove it and adjust (although I doubt that it would even work blowing away)

I have a Q6700 @ 3.3ghz with the same HS and Arctic 5 poorly applied. My hottest core is 37 @ idle, my coolest core is 31 (I said it was poorly applied)

Could it be a faulty temperature sensor? Does the HSF feel warm when it's idle? If not then I would say temps are being reported wrong.

I don't think you have a problem.
I've had the same issue for the last couple months and it's related to some 45nm cpu and motherboard combinations. After looking into it (that's +40 systems), i can only conclude that some boards read the internal sensor and approximate it wrongly.
Last week i've build an Asus Rampage Formula with a Q9650. Since the Q9650 is new (not QX9650) the bios read the temp wrong too. After a bios update, the temp dropped 30°C. But that's one case, most boards i've used (mostly Asus & Gigabyte) just read a very high temp even though the heatsink feel cold or lukewarm. I've had cases in OCCT where the temp would rise up to 83°C and the heatsink would feel just lukewarm (and system would be stable). I don't have this with 65nm cpu's.

i just built a q9450 build for my brother this week and with the stock cooler i was getting 32c on all cores at idle, make sure you have seated the hsf correctly, double check it just incase anyway, and if so, reapply thermal paste sparingly.

If the HSF is only lukewarm then windows is probably just reporting temps wrong. Nothing to worry about unless the HSF is too hot to touch under full load.

I'm experiencing the same issue, but with E7200 and P5W DH Deluxe. Real Temp reads 70C under heavy load, but heatsink is simply cold. However, motherbord's diode reads 45C which I think is closer to the real temperature. I thinks it's caused by wrongly calibrated sensors inside CPUs.

What is your BIOS temp say?

Take a razor and check the flatness of your hsf and ihs. If your cooler uses push pins, make sure nothing is bending or warping under pressure. The way you are doing it though seems correct to me, so I'm thinking it's probablly faulty temp. sensors too.

here's mine running at 3.68 GHZ with noctua cooler

I am jumping on the faulty sensor train too. That isn't possible, unless you aren't making good contact between the CPU and cooler.

I get 51C idle on Core Temp and 30C from EasyTune6. Speed fan shows both values, Temp2 is what easy tune shows and Core 0 & 1 what Core Teamp show. Motherboard bios doesn't show core temps, just CPU temp (is the temperature sensor still under teh CPU like it used to, dunno).

Seeing as you are on Gigabyte motherboard as well, I'd say those temp are just reported wrong. (and yes my cooler has great contact and not too much AS5 in between). I got funny temps with E4300 and ASUS P965 board too.

edit: looked my old screenshots, seems E4300 idle temps weer 42C More funny is that this E7200 idled 56C on my previous board. That is the same 51C @ 1600MHz (as TJ Max was 105 then in Core Temp) as now @ 2000MHz. So most likely the sensors in the CPU are just way off.

Maybe the sensor is near a particularly hot part of something?

WhiteLotus said:

Maybe the sensor is near a particularly hot part of something?

Click to expand...

That wouldn't explain why some get so much less for their idle temp. But the sensors aren't made for idle temperature measuring, just accurate at the shutdown temp territory.

Looking for the limits for this processor in previous board I did see a nice 105C in the bios and a pretty text for "CPU over temperature!" So if core temp are 20C higher, that would have been 125C. Maybe that's the shutdown temperature, but it got in the way of my OC

Anyways I don't let the 50C idle bother me anymore as I know it's not true, my GPU even idles lower than that (47C). If someone has managed somehow to get 20C higher idle temperatures with some bios setting or motherboard change on 45nm Intels that would be nice to know.

RealTemp 2.75 uses the TjMax that Intel told us was true at IDF.

http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip

For sensors like yours that were reading too high before, Intel's new information doesn't help one bit. It only makes things worse. The new version of RealTemp has increased the calibration factor range up to plus or minus 9.9 and changed the formula used compared to previous versions to specifically help with CPUs that read too high.

It was nice of Intel to finally admit that these sensors suffer from "slope error". This means that the farther your real temperature moves away from TjMax, the larger the amount of error will become. This error can go either way. Some CPUs will read too low at idle while others will read way too high. The really bad CPUs will have one of each type of sensor on the same CPU so the combined error can be huge at idle. A difference of 15C or 16C from two sensors only millimeters apart at idle is more than just a little silly.

For case airflow you generally want cool air entering your case at the bottom front and flowing towards and out the back of your case. All fans including your CPU should be in this general direction.

Don't take what is being reported as CPU temperature as 100% accurate. It might be but if it hasn't been individually calibrated then you have no idea how accurate it really is. On my board, this sensor reports about 7C or 8C too high at low temperatures.

Intel's presentation also slipped in the fact that some of their 45nm core sensors "bottom out" or get stuck at lower temperatures. This can happen in the low 50C range that you are at and it can effect multiple sensors on multiple cores.

When you are at 2000 MHz and 1.120 volts, how is your CPU load in Task Manager. In XP, the CPU Usage for me is usually between 0% and 1%. If something is loading your CPU then try to kill it in the Task Manager when testing. I don't think your sensors are stuck yet. It might be a task running in the background that is contributing some heat to your CPU.

What's your room temperature near your computer? Try doing some more testing with the new version of RealTemp and see if you can get your idle temps down to a more believable range.

unclewebb: That wasn't ment for me, but I'll answer as well anyways. With -9.9 in RealTemp I get 39C idle @2022MHz and 1.168V. That seems a lot more believable than 51C idle. 23.1C room temp and case on the floor, so it's even cooler there.

Orthos load with 3200MHz (1.152 with vdroop) goes from 63C to 54C.

Both cores give sensor movement of 4 and report identical temps. Idle load is between 0-2% and wolf2009 idle load (0-3%) is seen in the CoreTemp screenie, so it's not that.

Should try undervolting and underclocking sometime to see how near (far) room temp I can get. Don't think my CPU is bottoming out at 50C temp just showing higher, as in 4Ghz clocks idle (2532MHz) was about 56C.

wolf2009@
possible process bugged or virus (cpu usage run at 100% ).Check this with "windows task manager" (run:taskmgr.exe or CTRL+Alt+Del) .

OnBoard: Thanks for the info. I opened my eyes and wolf2009's CPU load is right where it should be when testing so it's not that. I was helping one user who had CPU Usage at around 25% because one core was at 100% when he thought he was idle. Hard to calibrate under those conditions.

The reason I originally chose TjMax=95C is because using TjMax=100C implies that there is a huge, sh*t load of "slope error" in some of these 45nm sensors. I was being too kind to Intel.

I re-tested a 45nm E8400 - C0 and a 65nm Q6600 - G0 and I set both of them to TjMax=100C. I took both of them up to about 80C and a little beyond at times where sensor error is minimal and measured the IHS with the IR thermometer again. When using TjMax=100C, the hottest measured spot on the IHS is always 5C less than the RealTemp reported temperature. I spent half an hour and moved the IR thermometer around and around at various temperatures and that was as close as I could ever get. The CPU covers a very small area of the IHS and heat dissipates very rapidly at these temperatures that a 5C gradient between IHS temperature and core temperature is not impossible to believe.

These pictures were taken about 2 seconds apart:




Intel showed that TjMax is not a fixed number and does vary by a few degrees but they didn't release how much it can vary by. It might be plus or minus 2 or 3 degrees or more. There's no way I can buy a large enough sample of processors and test them all to come up with an average range of error. Intel knows these details but they have decided not to release any additional information that is necessary to write an accurate core temperature monitoring program.

TjMax=100C might be an absolute maximum and the average processor might be a few degrees less than that. As I said on XS, releasing TjMax for 45nm processors was nice of them but so much additional information was left out that enthusiasts are no further ahead. They mentioned slope error but didn't show how large this amount can grow to. With the two Quads on this page, it can obviously grow to be quite large.

The 65nm Q6600 and other 65nm Dual Core CPUs I tested do not have the slope error that my E8400 has. Core0 and core1 on my Q6600 are very consistent and track each other closely from idle to TjMax. At low volts and low MHz and with no correction, they report a core temperature of about 6C over room temperature at idle. Considering how little heat they are putting out at this setting and how much heat a large air cooler can dissipate with the fan on max, this seems reasonable. 45nm CPUs put out less heat at the same settings compared to a 65nm CPU so there is no way that they should be idling at a temperature way higher than a 65nm processor.

All I can do is increase the amount of slope correction that RealTemp allows you to do maybe from a maximum of -9.9 to a max of -19.9. This will give you some more flexibility and should let you guys get your 45nm Quad temps down to a much more believable range. I wish Intel gave me more info to work with but they haven't. I'll post a link here when the next beta is ready.

It will be interesting to see how RealTemp tracks your CPU reported temperature.