• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

RealTemp General Discussion

Joined
Sep 8, 2005
Messages
836 (0.12/day)
Location
Romania
System Name Rocket
Processor Ryzen 3600X
Motherboard ASRock B450 PRO4
Cooling Noctua NH-D15
Memory HyperX Predator Black 16GB DDR4 3200MHz CL16
Video Card(s) MSI GeForce GTX 1060 GAMING X 6GB
Storage ADATA SX8200 PRO 512GB + Intel 535 Series 120GB + WD 6400AAKS
Display(s) Benq EW2420
Case Antec P182
Power Supply Antec Signature 650
Software Win 10
If you change TJMax value you're fooling yourself. :)
The only value that matter for your CPU thermal behavior is distance to TJMax. Keep it >20 in daily use and be happy.
If you have a 10C jump at idle you can take a look at those pesky push-pins. Maybe one is loose.
What's your core temps after 15 minutes load with prime95 Small FFTs?
 
Joined
Nov 18, 2009
Messages
34 (0.01/day)
ok, ill check.im not sure i should run prime, the comp is at a constant 52-51c under 26-28% load.

edit...btw the room temp is now 63f

and i checked the push pins and they seem fine all 4 feel tight and there all sticking out the back of the mobo

edit ok i rebooted the system and the bios say 24c then when the os booted the temps where at 36c and i push 60%load and it didnt go over 47c
 
Last edited:

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
RealTemp 3.59.3
http://www.fileden.com/files/2008/3/3/1794507/RealTempBeta.zip

The temperature sensors that Intel uses in their CPUs work the opposite of a thermometer. As the CPU heats up, the sensor counts down toward zero. When it reaches zero, the CPU will start to thermal throttle and slow down. The number that RealTemp reports as Distance to TJMax is a direct reading of these sensors.

If you know that your CPU core temperature will be at 100C when this sensor counts down to zero then you can approximate from that point. If the sensor reads 10 that means you are about 10 degrees away from the throttling point so you must be at (100-10) 90C. If the sensor reads 30 then you must be at a core temperature of (100-30) 70C.

Core Temperature = TJMax - Distance to TJMax

The problem is that Intel never released any documentation about what the correct value is for TJMax. I think RealTemp 3.00 might have taken a wild guess and used a value of 95C. Long after Core 2 CPUs were first released, Intel finally released some information at one of Intel's IDF conferences. They stated that for a Q9000 series CPU, 100C is the correct value. I believe that number but they later released information for the earlier 65nm CPUs that seem a long ways off from actual TJMax.



In the fine print they call this number TJ Target. The actual TJMax of a CPU can vary from one Q9550 to another and it can also vary from one core to the next on the same CPU. Intel did not release any further information like how much actual TJMax might vary or any other details about their calibration procedures. They did say that trying to get accurate core temperatures out of these sensors based on the above formula is not a good idea. These sensors were designed to control thermal throttling and thermal shutdown and that's it. They're not documented for any other purpose than that.

100% accurate core temperatures from idle to full load are not possible when that is based on data coming from these sensors. If no one knows what TJMax really is, then you can't compare your sensor readings to your friend's sensor readings. There is so much error in the 45nm sensors that this is pointless.

About all these sensors are good for is to compare your CPU from one day to the next. They seem fairly consistent and repeatable that way. If you change the air flow in your computer or if you change a heatsink and your full load temperatures are 5C better than before then you've done good.

The CPU temperature sensor that the bios reports is coming from a different sensor than the core sensors. There's no guarantee that sensor is 100% accurate either so comparing one inaccurate sensor to another one also isn't worth worrying about. The core sensors are located on the hottest spot on the CPU and the bios based temperature sensor is located in the middle of the CPU. At full Prime95 load, there can be a significant difference (20C) between these two sensors even if both are properly calibrated but neither sensor is.
 
Joined
Aug 13, 2009
Messages
3,187 (0.60/day)
Location
Czech republic
Processor Ryzen 5800X
Motherboard Asus TUF-Gaming B550-Plus
Cooling Noctua NH-U14S
Memory 32GB G.Skill Trident Z Neo F4-3600C16D-32GTZNC
Video Card(s) Sapphire Radeon Rx 580 Nitro+ 8GB
Storage HP EX950 512GB + Samsung 970 PRO 1TB
Display(s) HP Z Display Z24i G2
Case Fractal Design Define R6 Black
Audio Device(s) Creative Sound Blaster AE-5
Power Supply Seasonic PRIME Ultra 650W Gold
Mouse Roccat Kone AIMO Remastered
Software Windows 10 x64
The last stable version on the web is still 3.40. Are you planning to update that? Considering the beta is already 3.50+...
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
The most stable version is the one I posted just above your post. Donations to project RealTemp have been averaging $0.00 the last few months so I've come to the conclusion that this program is not very important to the hundreds of thousands of people that use and download it so I don't put too much effort into keeping the main site updated. If I have some free time later this summer I will try to update it but I have another project that I've been working on for the last 6 months so I haven't gotten around to it.

You can check out ThrottleStop if you're interested.
http://www.fileden.com/files/2008/3/3/1794507/ThrottleStop.zip

Beyond zero money, the next biggest problem is zero feedback from users. RealTemp 3.40 has a major bug as soon as you click on the Reset button if you have an ATI GPU but less than 1 person per 100,000 downloads takes 30 seconds out of their day to notify me of this. I have to assume that no one gives a crap about this project anymore or is interested in seeing it improve so I've decided to put my free time into other projects. Sorry about that.
 

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
58,413 (8.21/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570-F (BIOS Modded)
Cooling Alphacool Apex UV - Alphacool Eisblock XPX Aurora + EK Quantum ARGB 3090 w/ active backplate
Memory 2x32GB DDR4 3600 Corsair Vengeance RGB @3866 C18-22-22-22-42 TRFC704 (1.4V Hynix MJR - SoC 1.15V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144)
Case Fractal Design R6
Audio Device(s) Logitech G560 | Corsair Void pro RGB |Blue Yeti mic
Power Supply Fractal Ion+ 2 860W (Platinum) (This thing is God-tier. Silent and TINY)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE ( Sexy white keycaps)
VR HMD Oculus Rift S + Quest 2
Software Windows 11 pro x64 (Yes, it's genuinely a good OS) OpenRGB - ditch the branded bloatware!
Benchmark Scores Nyooom.
The most stable version is the one I posted just above your post. Donations to project RealTemp have been averaging $0.00 the last few months so I've come to the conclusion that this program is not very important to the hundreds of thousands of people that use and download it so I don't put too much effort into keeping the main site updated. If I have some free time later this summer I will try to update it but I have another project that I've been working on for the last 6 months so I haven't gotten around to it.

You can check out ThrottleStop if you're interested.
http://www.fileden.com/files/2008/3/3/1794507/ThrottleStop.zip

Beyond zero money, the next biggest problem is zero feedback from users. RealTemp 3.40 has a major bug as soon as you click on the Reset button if you have an ATI GPU but less than 1 person per 100,000 downloads takes 30 seconds out of their day to notify me of this. I have to assume that no one gives a crap about this project anymore or is interested in seeing it improve so I've decided to put my free time into other projects. Sorry about that.

i'd donate if you worked in AMD support :p
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
I really liked the last AMD x2-555BE I used for a friend's build that magically and easily turned into a Quad. Unfortunately I don't have enough time for all my Intel projects so I won't be starting any AMD projects. I don't mind working for free but my projects turn into negative numbers when I need to start buying hardware. Translation: My wife found all the Intel CPU boxes hiding in the closet downstairs.

Lucky the recent QX9650 didn't come in a box. :laugh:
 
Joined
May 22, 2010
Messages
4 (0.00/day)
Dear unclewebb.
Please, try to comment this screenshot:


From this: http://www.fcenter.ru/cgi-bin/sitemanager/redirecturl.cgi?urlid=6433

The ambient temps = 30.2-30.6 C! How is the RealTemp GT can show minimum temps lower then ambients?

Thanks!
 
Joined
Aug 13, 2009
Messages
3,187 (0.60/day)
Location
Czech republic
Processor Ryzen 5800X
Motherboard Asus TUF-Gaming B550-Plus
Cooling Noctua NH-U14S
Memory 32GB G.Skill Trident Z Neo F4-3600C16D-32GTZNC
Video Card(s) Sapphire Radeon Rx 580 Nitro+ 8GB
Storage HP EX950 512GB + Samsung 970 PRO 1TB
Display(s) HP Z Display Z24i G2
Case Fractal Design Define R6 Black
Audio Device(s) Creative Sound Blaster AE-5
Power Supply Seasonic PRIME Ultra 650W Gold
Mouse Roccat Kone AIMO Remastered
Software Windows 10 x64
As a reaction on last posts, do you need any help with testing?
I have E8400 CPU and P45 board, which is well covered though, I assume (there is Core i3 on my gf's desk too)
Anyway I do have some time I could help with.
 
Joined
Sep 8, 2005
Messages
836 (0.12/day)
Location
Romania
System Name Rocket
Processor Ryzen 3600X
Motherboard ASRock B450 PRO4
Cooling Noctua NH-D15
Memory HyperX Predator Black 16GB DDR4 3200MHz CL16
Video Card(s) MSI GeForce GTX 1060 GAMING X 6GB
Storage ADATA SX8200 PRO 512GB + Intel 535 Series 120GB + WD 6400AAKS
Display(s) Benq EW2420
Case Antec P182
Power Supply Antec Signature 650
Software Win 10
The ambient temps = 30.2-30.6 C! How is the RealTemp GT can show minimum temps lower then ambients?
Don't ask Kevin, ask Intel why they come back at crappy sensors from Wolfdale/Yorkfiled era.
Any monitoring program will show you same temps as RealTemp because now TJMax (TJTarget in fact) is written in MSR and core temp formula is simple: TJMax-DTS= reported temperature.
 

Cheezeman

New Member
Joined
Jul 24, 2010
Messages
2 (0.00/day)
i'd donate if you worked in AMD support :p

Hello all -

First of all, Mussels your avatar cracked me the hell up!! :roll:

I have some questions about my recent build regarding CPU temps and could use some help. The build is:

Intel Core i5 750 2.66GHz @ 3.6GHz 1.1875v
CoolerMaster Hyper 212 Plus CPU cooler
ASUS P7P55D Deluxe motherboard
4GB (2x2) OCZ Gold memory
2x nVidia 460 GTX 1GB in SLI w/ 3D vision
Win7 64-bit



I believe I've captured all the info needed in these screenshots but please let me know if I need something else. The first screenshot is at stock settings. The second screenshot is overclocked to 3.6GHz @ 1.8v.


STOCK





3.6GHz @ 1.1875v.






(all of the following refers to the STOCK settings screenshot, the OC was taken on a different day)

OK so reading the RealTemp website calibration portion, it says to leave the CPU settings at default (for the new i5/i7s because of the auto-throttling) and that the temps I should be seeing should be about 8* C above what my room temperature is for my cpu cooler type (decent single fan air cooler). However as you can see in the screencap, the minimum temperatures is 21* C - which is 70* F. Now unfortunately I don't have a way to take my room temp, and it was pretty cool in the room the day, but let's say at coldest it was maybe 68* F (20* C) room temperature. So do I need to go enter a +8/+9 calibration into RealTemp or are the new CPUs really putting off that little heat at idle when throttled down? BIOS reporting somewhere around 22*-24* I believe.

The second part of the calibration involves the sensor cool down test, but I honestly have no idea what I'm supposed to do there... :confused: I take it the test is supposed to tell you how to further calibrate the sensors so they are normalized against each other, but I don't really understand how to interpret these results.. Could someone help me?

After I know I have accurate temp readings, my final question is what is the real max temp for the i5 as displayed in RealTemp? I know Intel lists Tcase as 74* C, but RealTemp shows Tjunction (at 100* C!!), and I don't really know which I'm supposed to be adhering too. What is the difference between Tcase and Tjunction, and what is the max for Tjunction for an i5 750? I can't seem to find a straight answer to this.

As you can see in the overclock screenshot, the max temps 'only' get to ~65* C (without possible calibration yet) but the minimum temps are drastically higher, even with the throttling enabled and it has self-throttled to just 1.6GHZ. But I'm not sure if I should be alarmed of 74* (intel spec) or 100* (realtemp) and that I might be harming my CPU already if my calibration is off.

I appreciate any help!
 
Last edited:

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
Sergey Lepilov: Nothing I can do about the poor quality sensors that Intel is using on their new Gulftown CPUs. The sensors on the Core i7-900 series were excellent. The new sensors are not.

Intel only designs and calibrates their sensors for accurate thermal throttling and thermal shutdown control. They were never intended to be used for accurate temperature monitoring, especially at idle. These sensors have what is called slope error. That means the further you get away from the calibration point, the more these sensors will wander and start reading either temperatures that are too high or in your case, temperatures that are too low.

If I had some new hardware and a little bit of help from Intel it would be easy to calibrate these sensors to get reasonably accurate core temperatures out of them. At least these sensors don't have the severe sticking problems like many 45nm Core 2 sensors have.

Octopuss: Thanks for your offer to help. I don't have anything that needs testing at the moment but I do appreciate the offer.

Cheezeman: All of the testing I've done has been with Core 2 CPUs. I can't afford to start an extensive testing program for the Core i CPUs so I won't be bothering. The information in the documentation about calibrating Core 2 based CPUs is probably not 100% relevant to calibrating Core i CPUs. The Core i CPUs also have a feature where cores are being mostly turned off at idle and can spend time with virtually no voltage going through them. This behavior might be partly responsible for some of the differences you see from core to core in your idle temperatures but most of the difference is likely the slope error that I mentioned before where sensors move at different rates compared to how the actual core temperature is changing.

The maximum Tjunction temperature for your CPU is 99C. If your CPU ever reaches that temperature it will start to thermal throttle and slow down so it probably won't go beyond that temperature as long as your heatsink doesn't fall off. Even if your fan fails, the CPU is designed to automatically slow down to prevent a melt down.

It is impossible for software to accurately measure your Tcase temperature so the Intel rating is meaningless for most users. The only way to accurately measure Tcase is by taking a Dremel and cutting a groove into the top of your CPU and running a calibrated thermocouple to the geometric center. That's fine if you're in the lab but not practical for the end user to do to their own CPU. As long as you are operating your CPU so you are not thermal throttling, then you are operating your CPU within the Intel design spec. RealTemp shows the word OK if you have not triggered any thermal throttling. Your CPU says OK, OK, OK, OK so everything is OK. No worries about your core temperatures.

It would be a good idea to scroll up to post #682 and download the latest version of RealTemp that I posted.

The second screenshot is overclocked to 3.6GHz @ 1.8v.

I sure hope that 1.8v number is a misprint or else you'll have far bigger things to worry about than your core temperature. :)

It's not worth trying to calibrate these sensors. You can use RealTemp to make the numbers line up and look nicer but there are too many unknowns. Your calibration might not be any more accurate than what you've got now.
 
Joined
May 22, 2010
Messages
4 (0.00/day)
unclewebb
Thanks!

So you mean, that tests coolers on the Core i7-920 more exact and it is possible to trust then the tests coolers on the Core i7-980X?

And what do you think about the necessity of alignment IHS of the Core i7? This is the ansew of Intel:

Datasheet should describe the material used for the Integrated Heat Spreader (IHS), if not, it is nickel plated copper.The actual construction (IHS thickness) does vary from processor family to processor family. This is done to ensure a uniform height for the top of the IHS in the socket to facilitate the reuse of common thermal solutions from generation to generation. I would have to ask package designers the exact thickness, but Xbit could probably figure it out by taking a device apart and measuring it. The IHS’s are not perfectly flat, the flatness spec should be documented in the Mechanical section of the Datasheet.Note that once the processor is mounted in the socket, the compression will slightly change the flatness.Polishing (or lapping, as it is called) the IHS is a common practice among extreme over-clockers. The purpose of lapping is to improve the flatness and make the surface smoother, it is important to also lap the heatsink being used. Again, this should be done with the device mounted in a socket. A flatter, smoother surface will increase the contact surface area between the IHS and the heatsink, resulting in better thermal performance.The improvement in heat dissipation will be fairly small under normal conditions, but for extreme over-clockers it is worth the effort.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
For your heatsink testing, I would stick with the Core i7-920. The 45nm sensors that Intel used in those CPUs were excellent. I can't remember seeing a single screen shot of an i7-920 with bad sensors but all of the i7-980X have sensors that do not track the change in core temperature 100% accurately. The amount of error increases as the CPU cools. Core 0 on the i7-920 seems to be the most accurately calibrated by Intel. They have never said what their calibration procedure is but the data from that core always looks very accurate and most importantly, repeatable so your results should not change very much from one test to the next.

Some users have lapped their CPUs but unless there is something very wrong with the IHS, they usually don't see much of an improvement in reduced core temperature. I think the difference in core temperature between core 0 and core 3 in the i7-920 is mostly sensor error and the way Intel calibrates their CPUs. Lapping the CPU usually won't change the difference between those two cores.
 

dnd

Joined
Aug 1, 2010
Messages
5 (0.00/day)
stupid question...occasionally realtemp gt will sound the alarm (just for a moment) and then be quiet again. if i look at the temperatures they all seem to be fine (like 50 below tjmax), so is this a bug? interaction with some other software (eg everest)? this is with 3.40 btw.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
What temperature do you have the alarm set to? What CPU do you have? Are you using both a CPU and a GPU alarm? Does CPU-Z show your CPU as a retail processor or an ES processor? What GPU do you have?

Generally speaking, I haven't heard of any complaints about the alarm feature but it's a feature that I never use so maybe there is some obscure bug that I don't know about. Other software should be OK and they shouldn't be causing RealTemp to signal a false alarm. The GPU sensors seem more prone to reporting blips of false data.

Let me know a few more details. If it is bad data coming from your sensors then it might not be anything that I can fix. Run a RealTemp log file so when an alarm goes off, you can look in the log file and see if a sensor is reporting some bad data.

Ultimately, if the alarm is not reliable for you then I guess you'll have to turn this feature off and not use it. :(
 

dnd

Joined
Aug 1, 2010
Messages
5 (0.00/day)
it's a gulftown, temp is set to 80 deg so it is nowhere near the alarm temp at 55 deg max. it's a retail proc, having said that it is oc'd and errored out on a run today, so it could just be a manifestation of instability. GPU is a 480 gtx so that is probably buggy as well lol.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
I'll have a look at the code when I have a chance and see if I can figure out anything that I might have done wrong. I don't own a Gulftown and haven't had any feedback good or bad so I'm not sure what works and what doesn't.
 

dnd

Joined
Aug 1, 2010
Messages
5 (0.00/day)
one other thing i've noticed with i7 turbo is that it looks like turbo mode is throttling the first core, but only momentarily...normally all 6 sit at a 31x multi, but occasionally i'll see the first core drop to 25 but only for an instant, and then it jumps back up again. this is on a rampage 3 extreme.
 

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
What other monitoring software are you using? There are some other programs out there that don't use the system timers in a friendly manner which can screw up the results that i7 Turbo displays. At one time, Everest used to cause problems for the timers that i7 Turbo uses but I haven't kept up recently with what third party software does what. When i7 Turbo is run by itself, it usually gives very consistent results.

i7 Turbo GT
Version 1.1
http://www.mediafire.com/?w5lpqqpy6wwq8wi

The Asus P6T series had some problems with turbo throttling but I haven't heard of any problems with the rampage 3 extreme. When you have actual turbo throttling at full load, it looks more like this.

http://img17.imageshack.us/i/asusp6tddlxw3520throttl.jpg/

When the CPU is fully loaded, the average multiplier on each thread will drop down fairly equally.

ThrottleStop and i7 Turbo use the same timers and same method to determine the multiplier on each thread.
 

dnd

Joined
Aug 1, 2010
Messages
5 (0.00/day)
i have got everest running so perhaps it is making a mess.

EDIT: is it normal to have a bit of variation between the cores? right now under prime95 i have 50, 52, 45, 47, 54, 49.
 
Last edited:

unclewebb

ThrottleStop & RealTemp Author
Joined
Jun 1, 2008
Messages
7,265 (1.26/day)
The sensors in the Gulftown CPUs are very poor at reporting accurate core temperatures. Intel might also do something like offset TJMax slightly from one core to the next but they've never documented that or anything else about the calibration procedure they use.

Your Gulftown temps look typical. They're usually all over the place and I have no idea which one is right.
 
Joined
Aug 13, 2009
Messages
3,187 (0.60/day)
Location
Czech republic
Processor Ryzen 5800X
Motherboard Asus TUF-Gaming B550-Plus
Cooling Noctua NH-U14S
Memory 32GB G.Skill Trident Z Neo F4-3600C16D-32GTZNC
Video Card(s) Sapphire Radeon Rx 580 Nitro+ 8GB
Storage HP EX950 512GB + Samsung 970 PRO 1TB
Display(s) HP Z Display Z24i G2
Case Fractal Design Define R6 Black
Audio Device(s) Creative Sound Blaster AE-5
Power Supply Seasonic PRIME Ultra 650W Gold
Mouse Roccat Kone AIMO Remastered
Software Windows 10 x64
Joined
Sep 8, 2005
Messages
836 (0.12/day)
Location
Romania
System Name Rocket
Processor Ryzen 3600X
Motherboard ASRock B450 PRO4
Cooling Noctua NH-D15
Memory HyperX Predator Black 16GB DDR4 3200MHz CL16
Video Card(s) MSI GeForce GTX 1060 GAMING X 6GB
Storage ADATA SX8200 PRO 512GB + Intel 535 Series 120GB + WD 6400AAKS
Display(s) Benq EW2420
Case Antec P182
Power Supply Antec Signature 650
Software Win 10
Here you go.
Usually Kevin posts only final versions on TPU.
 
Top