Real Temp - New temp program for Intel Core processors

[unclewebb]

Here's the official release of RC5.

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

A couple of minor fixes so the MHz will continue to be updated correctly during a CPU Cool Down Test. Also included the MHz in the XS Bench window so you can collect these little windows like football or hockey cards. You can check back a month later to make sure your CPU is not degrading.

1610 is the number to shoot for with your Core i7 at 4GHz.

[MrHydes]

i have real temp 2.5 and beta 2.9 the new one when cpu is idleing doesn't
matter if its running 3.3GHz default speed or 4.2 always shows 30ºc

unclewebb can you explain why?

and congratulations for the popularity of your work
thank you

[gymenii]

Thanks for the update.

[unclewebb]

Mr Hydes: At idle with good cooling, there isn't much change in temperature even when there are some very big changes in MHz. There are also lots of sticking sensors with the 45nm chips. With version 2.90 do a CPU Cool Down Test and post your results. That usually tells me everything I need to know. Try running it at your default MHz and core voltage.

[randomizer]

Ooo... Realtemp with serial. What does that unlock?

[unclewebb]

Ooo... Realtemp with serial. What does that unlock?

I wish those guys would send me some of the money they're scamming by using RealTemp's good name.
Maybe with a serial your temps will drop a couple of degrees. Cheaper than buying a better cooler.

[sakis_the_fraud]

Congratulations for your excellent work on the programm unclewebb!

I suggest you to stop updating the versions with RC1,2,3 etc because at the realtemp window they appear the same, as a 2.90 version

So you can't tell the difference between realtemp 2.90 RC2 and realtemp 2.90 RC5. Keep this in mind for the next update.

[PCTwin]

Hi unclewebb. I've been running my qx6700@3.6 for a while now with no probs but rc5 seems to not recognise my multi setting. I don't know exactly wich version it went wrong on, but rc1 is fine.

[bowman]

Any way you could implement temperature graphing? I got the idea when I got up this morning and saw when it hit the lowest temperature.. I'd love to have a graph of each core's temperature along the timeline. Perhaps with load graphs snatched from the task manager above it?

Just, you know, if it's feasible it'd be really neat.

[burebista]

So you can't tell the difference between realtemp 2.90 RC2 and realtemp 2.90 RC5.

With a little detective work you can.

[PCTwin]

Silly me. I've just caught up with the last three pages. I still don't get it. I have disabled c1e and all that in bios and I don't use software to oc. So does this mean all 4 cores are not @ 3.6.

[rge]

Silly me. I've just caught up with the last three pages. I still don't get it. I have disabled c1e and all that in bios and I don't use software to oc. So does this mean all 4 cores are not @ 3.6.

Do you have EIST disabled as well?

[unclewebb]

PCTWin: I've decided to let RealTemp report the average multiplier. I realized that some users weren't going to like seeing the truth because it's not what they're used to seeing. We've all been led to believe that the multiplier is a nice fixed number but I found out during testing that depending on how you have your computer set up, the multiplier might be jumping up and down like a yo-yo. RealTemp displays the average multiplier over a 1 second time interval. If it does not show a stable number then your multiplier is not stable.

On my motherboard, EIST (aka. Enhanced Intel SpeedStep®) is mostly in control of this. Depending on your bios and 101 other things, you might need to go into the Windows Power Options and adjust your Minimum processor state in Vista. Make sure that EIST is disabled and your Minimum processor state is set to 100% and then your multiplier should be steady at 8.0. If you want your Core 2 multiplier steady at 6.0 when your computer is idle then enable EIST and set your Minimum processor state to whatever percentage it takes to get the multi steady at 6.0. When using Vista, my Q6600 needs a setting of 66% (6.0 / 9.0) or lower. In Windows XP you might need to set your Power Options -> Power Scheme to Portable / Laptop to get your multi steady at 6.0.

sakis_the_fraud: I wasn't originally planning for all these RC versions. The last official release has been out for 5 months without an update so I wanted to make sure that version 2.90 Final was a good one with lots of new features and not too many bugs. I think we've reached that point. I'm just waiting to hear back from one user that had problems with MHz but I'm pretty sure that only happened when he was doing a CPU Cool Down Test which used to temporarily suspend updating the MHz. If all goes well, the official 2.90 will be released in a day or two. Next time I go off on an RC tangent I'll include the RC label in the title bar so it's clearer.

bowman: My good friend burebista has been twisting my arm for months trying to get me to write a RivaTuner plug-in for RealTemp. He says if I did this then RivaTuner could be used to handle the graphing. I haven't written a plug in before. If I don't get too confused then I'll probably do that next. If I get confused then maybe I'll just write my own graphing code. For me, that would be easier. I hate running SpeedFan just to see a temperature graph.

If I have enough time I might do both in the New Year. I like RivaTuner and I'm interested in that project but I'd also like some more advanced graphing abilities. I think it might be easier to calibrate RealTemp if you could see what your temperature curve for each core looked like. Being able to visually see the effects of adjustments to TJMax and the Calibration settings would be great.

[PCTwin]

EIST is also disabled. It just seems that any setting other than the default multi doesn't add up. I'm now on 10x multi and it's showing correct speed, but if I move the multi either up or down the cpu speed isn't right.

[unclewebb]

At full load, does CPU-Z and RealTemp show the same FSB MHz and the same multiplier no matter what multiplier you're using?

I think some motherboard and bios versions simply do not handle EIST correctly. Turning EIST off in the bios may not actually turn it off when you are in Windows. Try using RMClock 2.30 to check for that. I had problems with a newer version so I've stuck with version 2.30.



On my board, I can't toggle EIST with this program but it does correctly read if EIST is enabled or not. If the problem you are having with RealTemp is because of your bios, then I will provide a switch in RealTemp that you can use so it reports the multiplier the way it used to. If the problem is because of RealTemp, then I will provide a switch in RealTemp that you can use so it reports the multiplier the way it used to.

Either way I'll get things fixed up for you. It will be interesting to see what the bug really is. I'm hoping to blame this one on the Abit bios. Post a few pics of CPU-Z vs RealTemp while running Prime with your multi locked at 10 and above or below 10. Your Extreme processor might be working a little differently than what I'm used to seeing and what I've tested on so far.

Edit: I just went into my bios. EIST was disabled but as soon as I manually set a multiplier, the EIST option disappeared and when I booted up, it showed that EIST was enabled. There doesn't seem to be a way to turn it off when a multiplier is entered manually.

I get this when the multiplier is set to Auto in the bios and EIST is disabled.



It might also be an Intel design spec that whenever a multi is manually entered that EIST must be on.

[X.T.R.E.M.E_ICE]

Is there a reason why the readings don't go lower then 28c ?

[PCTwin]

First test results. Will update my sig soon. Water cooled cpu, 18c ambient but I dont know the ambient of the water. Got to fit a sensor soon. Reset all bios settings to auto and ran a sensor test. C1E, EIST enabled, will do some more tests tomorrow.

[PCTwin]

My calibration and sensor test results.

[unclewebb]

Is there a reason why the readings don't go lower then 28c ?

Most of the 45nm sensors stop moving by that point. That's just the way they are.

Theoretically, if the sensors didn't get stuck, they should be able to read down to -27C.

The CPU Cool Down Test might show your sticking point.

PCTwin: changing TJMax to 90C for your B3 Quad recently looks like a good decision. Your 25C idle temps in an 18C room don't look like they are too far off. It will be interesting to see when you get a reading of your water temp.

[PCTwin]

Unclewebb. Just a quickie before bedtime, what do you think of the calibration settings.

[unclewebb]

PCTwin: When I first started this project I assumed that TJMax was a fixed value across all 4 cores. It's definitely not and Intel agrees with that. 90C is a TJ Target that Intel was aiming for but due to manufacturing tolerances, actual TJ Max can be slightly higher.

Now that I'm not afraid to change TJMax, why not try this calibration instead. For TJMax use 90, 90, 93, 91 and now you won't need any Idle Calibration for cores 0, 1 and 3. Core 2 on 65nm Quads frequently reads lower than the rest. The slope of the curve is slightly different than the other ones so I'd still use an Idle Calibration setting of about 0.6.

The end result isn't going to be much different than what you have. It might be a degree or two more accurate but given the quality and limitations of these sensors, it's not that important to have a perfect calibration. At least you're in the ball park. It's the 45nm Quads that need some fancy math to try to decipher what their sensors are saying.

When done, try running Prime95 Small FFTs with the log file set to 1 second. Start and stop Prime a few times every 30 seconds so the temperatures can cycle up and down. Afterwards have a look at the log file and you'll probably see your 4 cores tracking each other pretty closely from idle to full load.

One thing I was going to mention since your screen shot shows the problem is that with the old Visual C++ compiler I'm using, there's a bug and no way to change the heading colors when you adjust the GUI colors. There's a few minor things like that I'd like to try and fix in the future. That's the sacrifice I had to make to get the buttons with the rounded corners that I like.

[rge]

Had to break in RC5 Realtemp XS bench and update my sig

[PCTwin]

Results @3.4g 8xmulti 426 fsb. Cpu-z and Realtemp dont show the same clock loaded or idle. EIST,C1E disabled in bios. With rm, EIST shaws as enabled even though disabled in bios. Changed my calibration, set tj to 90 90 94 91. There's no cpu function in control panel power options.

[PCTwin]

Calibration and sensor test result. Will do 10 multi soon.

[unclewebb]

rge: Congratulations for setting the XS Bench record.

I'm going to use that number in the XS Bench window to give users something to shoot for.

PCTwin: Interesting results. I'm going to play around with multipliers today to see if this method of calculating the average multiplier is defective.

Can you try a couple of tests when you get the chance. How about running your computer at about 3400 MHz with a 10X multiplier and then try to run at that same speed using an 8X multiplier. Use CPU-Z to confirm your MHz. At both settings run a SuperPI bench and maybe an XS Bench as well. Try to have your memory at similar speed and timings for both runs. I just want to make sure that your CPU is running at the same speed. Your bench scores should be very similar if you are and will be quite a bit different if you're not.

If this average multiplier method does not work on a QX processor then I'll scrap it. Thanks for your help.

Edit: I was also wondering what program you're using to load your CPU? Can you use Prime95 Small FFTs?