Real Temp - New temp program for Intel Core processors

Quote:

---

Originally Posted by Bullet92

Usually 95°C.

---

Always 95°C. ;)

My theory based on testing is that PROCHOT# or the point where a processor starts to thermal throttle can vary by a degree or two from one processor to the next but TjMax is fixed for all E8400 CPUs. I realize that there are plenty of people that disagree with that statement but I haven't seen any testing yet to back up their theories.

JohnZS: I don't put too much faith in any software that tells me my processor is "Too hot" if it doesn't even know what the correct core temperature really is. OCCT obviously means well but it's using the wrong TjMax and is reporting the core temperature 10C too high just like Everest and CoreTemp do.

If you are Prime stable at 63C and don't plan to overclock any further then you're fine. Running as cool as possible improves stability but if you're already stable and happy with your overclock then temps aren't really important. Do you have any real world software that you run that gets your cores up anywhere near 63C? Most users don't. Prime is great for stability testing but is a little overkill compared to most gaming or video / music encoding a user ever does.

You can be Prime stable at over 90C if you're not overclocking too much and your processor won't start to thermal throttle and slow down until about 93C so at 63C you still have lots and lots of head room before you'll ever hurt anything.

http://img214.imageshack.us/img214/1381/test1smllf3.png

Hey uncle, I did the comparison, water vs air.
My CPU loop is a 355 w/XSPC res top, Dtek v2, MCP320 w/ultra kaze 3000.
The air cooler is a Thermalright Ultra 120 w/Panaflow high speed, FBA12G12H1BX

edit- oops! my sig. Q6600@ 3.4 Underwater, P5E-VM HDMI, 4GB OCZ 5,5,5,15 EVGA 8800GT, P.C.P.&C 610w

in the bios, 1.10 vcore, auto multi as it wouldn't boot vista w/ the bios set to 6, speedstep, CIE enabled, 266 fsb . CPUz read 1.048 volts, 6 x 267.3 for 1603MHz.
I ran the air and and water temps in F because one of the thermos I had to verify didn't do C.
Air temp was steady 77f (25c) for both runs. I had 3 thermometers, one is on the PC, a sensor in front of the rad, this is the same unit I use for water temp , Lian Li 3 fan control. (sucks for fans but the temp sensors are good!)
Two others within inches of the PC and when I ran the TR I set one infront of the intake air inside the PC, both side panels are off.
The water temp bounced from 77-79f so I'm thinking 25.5-26c, the sensor for the cpu loop in the res right after the chip. You're right on the money with the low and slow approach. There is very little heat
here's the screenie, water on the left-
http://aloader.com/files/65/air77water77or79compare.jpg

What are your thoughts? What would you use for calbration in these two instances?

Thanks WoodButcher for your numbers.

Based on my testing of my Q6600 - G0 with a Tuniq air cooler, I would try using calibration factors so that your reported core temperatures are about 6C higher than your air temperature.

With your numbers that would mean approximately:
( 1.5, 1.5, 2.0, 1.3 )

That should bring your idle temps up to about 31C which is 6C above your 25C air temp.

Using these same factors let's see if your reported temps when using water would make any sense.

Your water screen shot shows ( 23C, 23C, 22C, 25C )

When using the above calibration factors, that would be reported as ( 29C, 29C, 30C, 30C ) or an average core temperature of 29.5C. If your water temperature is about 25.5C then your core temperature is being reported about 4C higher than your water temperature which seems pretty reasonable to me.

We have to keep in mind that this calibrating procedure is far from an exact science. There is a margin of error in all of the temperature data that is gathered. Any uncalibrated, consumer grade digital thermometer will never be 100% accurate. The other problem is that the data coming from the Intel digital thermal sensors is integer data and we don't have any documentation which shows their accuracy, repeatability or even how they round off the data before outputting it. Would an actual temperature of 24.7C get rounded up to 25C (DTS=70) or does it get chopped off and output as 24C (DTS=71)?

Even with all of the maybes, I think the end result after this calibration are reported temperature numbers that are closer to the actual temperature. With something like an IR thermometer you might get another degree closer to the truth but I don't think it's worth it. The Fluke 62 that I've used is rated at +/- 1C at these temperatures so you might not be any better off even if you threw some more money at trying to get the exact core temperature of your Quad.

A quick RealTemp style calibration is enough for most users to get in the general ball park.

Thanks to you!:up:
I know you put a great deal of time into RT and many others have helped w/ the testing on air. With rge and I both doing the air / water comparison folks on water will be able to have some idea where to start. If a few more folk run this comparison it would be a worthy footnote to the calibrations page. Until now I felt like working off a guesstimate, I'll grant you any temp program given the inacuracies and lack of hard date is a guess but Realtemp because of you is an Xtremely well educated guess.

Now,,, jump in the car and go someplace others can call you a "dirty old man" you deserve some bikini watching! If the other half asks why you're taking binoculars tell her,,,:rolleyes:
Birdwatching!
redbreasted harlots, blonde topped boobies and white tailed hens! works for me!:yepp:

Well i seee no point on hunting sub 1°C accuracy. I need temperature monitoring just to keep an eye on my CPU. If it's running at 35°C all the time and some day it's constantly hitting 50°C i'd probably go and check if cooler is attached right or if fan is rotating or maybe dust crapified everything. 49, 50 or 51, does it matter? I'll notice the 35 to ~50 jump...

If you are using Excel it isn't too hard to import the RealTemp log file.

Start up Excel (2003 for this example) and select File -> Open and change the view to All Files. Point it to the RealTempLog.txt file and click on open.

It will give you some options and all you have to do is select Fixed Width and Next a couple of times and you're finished. Remember to save it as a standard Excel worksheet when you're done.

That's not a perfect solution but it will work. I'll try to create a comma separated file option (.csv) maybe tomorrow so it will be a little easier to directly import into Excel.

RejZoR: Your absolutely right. 1C here or there isn't anything to get too concerned about. It's when the errors start to approach 10C or 15C like some software does, then I get concerned.

When you look at all the possible measurement errors and the nature of these sensors, etc. I think the temps we are getting from RealTemp are excellent. As the programmer of this I know I am more than just a little biased but as a person that has used RT and tested with RT for several months I'm happy and confident when I grab this program to do some temperature testing.

I thought I was done but if someone needs RealTemp to output some Excel friendly (.csv) files then I'll do it. By the way, it's done!

http://www.fileden.com/files/2008/3/...alTempBeta.zip

It's not user friendly at the moment but I think there is enough room to squeeze in this option into the Settings window. For the moment you will need to go into the RealTemp.ini file and add this:

CSV=1

Now the log file will be saved as RealTempLog.csv which can be directly imported into Excel and saved as a worksheet. How's that for productivity on a Friday night? Obviously I need to get a life!

Sweet! csv is exactly what I need as well! thanks andy and unclewebb! :up:

Thank you unclewebb for your great program! :clap::clap::clap:
I have a little question: can you update the TjMax for Q6600-G0 step to 100°c?
Or you think that 95°c is the correct value?
I know that i can edit, but i don't understand the difference from your program and others... and i don't know who to believe :confused:

Sorry for my eng, and thank you again. :clap::clap::clap:

I believe 95c for q6600 G0. If you think it's not right simply change it yourself.

mcoffey: I found just enough room in the Settings GUI to squeeze in a CSV option so you can change it on the fly. I'll get that updated in the next few hours.

Quote:

---

psychok9: I have a little question: can you update the TjMax for Q6600-G0 step to 100°c?

---

I could do that; but then it would be wrong. The long standing belief that TjMax=100C for a Q6600-G0 is wrong. I have the same processor in my computer at the moment and I have tested it with an IR thermometer and TjMax=100C is somewhere between highly unlikely and impossible.

If you can provide me with some official Intel documentation which states TjMax for the Q6600 then I'll think about changing it for you. RealTemp is about doing things differently and not blindly following other temperature programs just because they are/were popular. Have you done any calibration testing as outlined in the docs like other users here have done? Give that a try first. loonym and I don't believe for a second that TjMax=100C for our Q6600 - GO.

Quote:

---

Originally Posted by unclewebb

I thought I was done but if someone needs RealTemp to output some Excel friendly (.csv) files then I'll do it. By the way, it's done!

http://www.fileden.com/files/2008/3/...alTempBeta.zip

It's not user friendly at the moment but I think there is enough room to squeeze in this option into the Settings window. For the moment you will need to go into the RealTemp.ini file and add this:

CSV=1

Now the log file will be saved as RealTempLog.csv which can be directly imported into Excel and saved as a worksheet. How's that for productivity on a Friday night? Obviously I need to get a life!

---

:clap: :up:

btw the start minimized still do NOT work with vista at startup.

RealTelstar: RejZoR and another user mentioned that it was working in Vista now. Here's some Vista info:
http://www.xtremesystems.org/forums/...postcount=1751

RejZoR says he is using the Task Scheduler in Vista and not the Startup folder to automatically assign the Administrator rights that RealTemp needs to read the temperature data from the CPU in an OS friendly way.

Does Start Minimized work when you try running RealTemp from the Desktop or does it never, ever work?

Quote:

---

Originally Posted by unclewebb

RealTelstar: RejZoR and another user mentioned that it was working in Vista now. Here's some Vista info:
http://www.xtremesystems.org/forums/...postcount=1751

RejZoR says he is using the Task Scheduler in Vista and not the Startup folder to automatically assign the Administrator rights that RealTemp needs to read the temperature data from the CPU in an OS friendly way.

---

Thats the difference, I put it in the startup folder. I have UAC disabled, so i think i should just use the task scheduler.

Quote:

---

Does Start Minimized work when you try running RealTemp from the Desktop or does it never, ever work?

---

Yes, it does.

Thanks RealTelstar for clearing that up. As a perfectionist, it's the little things that drive me crazy! :wth:

Now I can update the documentation explaining this for other Vista users.

Speaking of the log file, how often does it write to disk? Seems as of now, it writes the buffer to disk something like every hour + on shutdown.

So now if the system blue-screens, the buffer is lost. It would make sense to write to disk more often, like every minute or 10 seconds.

I think if you have logging enabled it writes to the log file every 5 seconds by default?

Quote:

---

Originally Posted by loonym

I think if you have logging enabled it writes to the log file every 5 seconds by default?

---

Yes but the data isn't actually written to disk every log interval. It's stored in a memory buffer and gets written to the file once every while... at least on vista.

I see, interesting, thanks. I really have never tried the log feature.

edit: is regular bluescreening an issue for you? j/k lol

majestik: You're right. RealTemp only writes the log file to disk once per hour or when it exits. I did this because I hate programs that are constantly thrashing and fragmenting my hard drive. Any suggestions about how often you would like the log data written to disk? I get a little carried away when it comes to efficiency sometimes. How about once a minute? Would that be a reasonable compromise?

If someone has the Logging Interval at one second then I really don't want RT writing to disk every one second. On an unstable system, RealTemp would end up getting the blame for screwing up someone's hard disk while writing to disk during a blue screen.

You could have a list item option for how long. Start it at 5 secs, up to 1/hour, with logical intervals. That way RT could not be blamed for the BSoD, but yet still offer some configuration to how the log is setup.

Edit:
Scratch that, have a .ini option so that it's not something that the average user would be able to change. It would default to 1/hour if the option didn't exist in the ini, but if it did exist it could be set to x seconds.

Yeah .ini value for sure.