W
wolf2009
Guest
^^ Thanks for your hard work unclewebb !
-
Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
- Joined
- Jul 28, 2008
- Messages
- 3,300 (0.54/day)
| System Name | Primary|Secondary|Poweredge r410|Dell XPS|SteamDeck |
|---|---|
| Processor | i7 11700k|i7 9700k|2 x E5620 |i5 5500U|Zen 2 4c/8t |
| Memory | 32GB DDR4|16GB DDR4|16GB DDR4|32GB ECC DDR3|8GB DDR4|16GB LPDDR5 |
| Video Card(s) | RX 7800xt|RX 6700xt |On-Board|On-Board|8 RDNA 2 CUs |
| Storage | 2TB m.2|512GB SSD+1TB SSD|2x256GBSSD 2x2TBGB|256GB sata|512GB nvme |
| Display(s) | 50" 4k TV | Dell 27" |22" |3.3"|7" |
| VR HMD | Samsung Odyssey+ | Oculus Quest 2 |
| Software | Windows 11 Pro|Windows 10 Pro|Windows 10 Home| Server 2012 r2|Windows 10 Pro |
It is possible the sensor is reporting wrong or your cooler it is not making enough contact with CPU refit the CPU cooler. 50+ C you can fell with hand.
unclewebb
ThrottleStop & RealTemp Author
- Joined
- Jun 1, 2008
- Messages
- 8,362 (1.35/day)
Killing a tube of thermal paste and applying your heatsink in 101 different ways isn't going to correct for sensor slope error. It's possible that the IHS and cores aren't making perfect contact but far more likely is sensor error.
"Slope error" is when the digital thermal sensors on your CPU move at a different rate than what the core temperature changes at. Your actual core temperature might decrease by 35C but the digital thermal sensor might only change by 20 positions or might go crazy and change by 50 positions. Intel has told us that this is a fact of life with the sensors on the 45nm processors. All a user can do is use temperature monitoring software that lets you try and correct for this.
For seriously borked sensors, the only software at the moment that gives you any hope is RealTemp. Here's the latest beta version for you guys to test out.
http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip
I increased the maximum negative calibration to -19.9 so it will be possible to get your temps to where they should be during the calibration procedure.
Other changes include the ability to see the Distance to TjMax in the System Tray. If you don't agree with correcting for slope error then just watch this number. As long as you leave yourself a couple of degrees of headroom away from TjMax then your processor won't thermal throttle. The more headroom you leave, the more you should be able to overclock so cooler is always better.
I've also adjusted TjMax for a lot of the 65nm processors. Intel has still released zero info about 65nm so I'm left guessing. My most recent guesses for 65nm are based on what Intel disclosed at the developer's forum for 45nm as well as testing 45nm and comparing that to early 65nm B2 processors. The long assumed TjMax=85C for the early Core Desktop processors is impossible if TjMax=100C for the E8400. The measured difference is 10C so the B2 and L2 and Pentium M0 have all been adjusted to 90C instead of the original 85C.
Re-calibrate with your case open and CPU fan on max at low MHz and low voltage and compare that to your air temperature near your computer. After that, let's see some testing to see how RealTemp compares to your CPU sensor. The CPU sensor is not perfect and it's calibration can get changed depending on the bios version but it doesn't seem to have the same gross errors that the on chip thermal sensors do.
"Slope error" is when the digital thermal sensors on your CPU move at a different rate than what the core temperature changes at. Your actual core temperature might decrease by 35C but the digital thermal sensor might only change by 20 positions or might go crazy and change by 50 positions. Intel has told us that this is a fact of life with the sensors on the 45nm processors. All a user can do is use temperature monitoring software that lets you try and correct for this.
For seriously borked sensors, the only software at the moment that gives you any hope is RealTemp. Here's the latest beta version for you guys to test out.
http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip
I increased the maximum negative calibration to -19.9 so it will be possible to get your temps to where they should be during the calibration procedure.
Other changes include the ability to see the Distance to TjMax in the System Tray. If you don't agree with correcting for slope error then just watch this number. As long as you leave yourself a couple of degrees of headroom away from TjMax then your processor won't thermal throttle. The more headroom you leave, the more you should be able to overclock so cooler is always better.
I've also adjusted TjMax for a lot of the 65nm processors. Intel has still released zero info about 65nm so I'm left guessing. My most recent guesses for 65nm are based on what Intel disclosed at the developer's forum for 45nm as well as testing 45nm and comparing that to early 65nm B2 processors. The long assumed TjMax=85C for the early Core Desktop processors is impossible if TjMax=100C for the E8400. The measured difference is 10C so the B2 and L2 and Pentium M0 have all been adjusted to 90C instead of the original 85C.
Re-calibrate with your case open and CPU fan on max at low MHz and low voltage and compare that to your air temperature near your computer. After that, let's see some testing to see how RealTemp compares to your CPU sensor. The CPU sensor is not perfect and it's calibration can get changed depending on the bios version but it doesn't seem to have the same gross errors that the on chip thermal sensors do.
Last edited: