Real Temp - New temp program for Intel Core processors

[unclewebb]

Here's a test I use to check how software deals with odd ball multipliers and situations that you don't read about in any book.

A Quad CPU has two Dual Core CPUs inside of it. Using MSR 0x199 you can trick it so that one of the Dual Cores uses one multiplier while the other Dual Core uses a completely different multiplier, even at full load running Prime95.

For my example I set core0 and core 1 to use a 9.0X multi and core2 and core3 to use a 6.0X multi. Right click on CPU-Z and you can get it to read any individual core.

It correctly shows one side of the CPU running at 3000 MHz and the other side at 2000 MHz so RealTemp averages that out and reports 2500 MHz. RealTemp also averages the 9.0 and 6.0 multipliers and displays 7.5.

The average multi read from the MSR is shown as 7.5 and the calculated multi also shows the correct 7.500. Not a lot of tools can handle this situation correctly. Hell, I'm not sure if any tools besides the ones in this screen shot can handle this situation correctly.

That's why I'm having a hard time understanding what's going on with your CPU. These tools are usually pretty good. When they display something odd, it's usually because something odd is going on.



Thanks for showing up burebista. It's past my bedtime so you can take over trying to figure out what's going on.

Edit: Run 4 instances of CPU-Z and set each instance to monitor a different core. Maybe in your case two cores are running at 8.0 and two cores are running at 6.0 so the overall average really is 7.0. That's all I can think of at the moment. Everest 5.0 can also be used to read the individual cores. I need another clue to this mystery!

Edit: Thanks for the signed driver burebista. RMClock works in Windows 7 x64 after replacing that file.

[burebista]

Edit: Thanks for the signed driver burebista. RMClock works in Windows 7 x64 after replacing that file.

No problem Kevin. It's just RivaTuner/Precision signed driver.

[RyanH]

Ok first of all thank you as well burebista I can get my Rmclock to work now and just to clarify I am running Vista 64 bit SP1. Ok first things first, yes uncle you're right it's showing a 6 multiplier on both core 2 and 3 and an 8 multiplier on core 0 and 1 so that's why the Realtemp and turbo are showing a 7 multiplier because it's averaging it out to 7 just like you said. How would I go about changing the multiplier on core 2 and 3 then? In my bios it only shows one FSB and multiplier setting theres no option to change 2 separate cores. No I am not running the latest bios on my motherboard and after looking through the lengthy thread for my motherboard in these forums it's said that the new bios versions have increased 45nm capability among other things because my motherboard was released back in 2007 before the newer penryn 45nm was released. However I did not bother to change it because my OS recognizes all 4 cores and as you can see all these programs are showing 450FSB on all cores including core 2 and 3 and the default multiplier/FSB for this processor is 8/333mhz so obviously the bios is compatible at some level as it raised the FSB on all 4 cores but for some reason opted to have a different multi on core2/3 even though I have it set at 8 in the Bios settings. Is it possible the C1E or EIST is changing this? Another question I have is do most motherboard bios usually let you set different multipliers and FSB for the different dies on Quadcores?

As you can see I don't have a processor power management option in neither my advanced power settings or RMclock so I can't change my minimum processor state and I can't see wether my C1E or EIST is still on according to RMclock. Maybe once I update my bios this option will appear?

Also, I have my Vcore set at 1.3325 in bios but it shows roughly 1.29-1.3 when I boot up and then it goes down again to about 1.26 again once i start prime95. I believe this is attributed to Intel spec Vdrop and Vdroop.
If I do turn off C1E and EIST this is still suppossed to happen correct?

So it appears the best course of action would be to flash my bios to the latest version would you agree with me?




Edit* I just wanted to add that both the FSB and multipliers for all 4 cores always stayed the same both at idle and while running prime95 small ffts, so core 0/1 were always at 8/450mhz and core 2/3 at 6/450mhz. However the FSB does go up or down .1 to 1mhz regularly but I'm guessing that's normal.

You said you were able to trick your quad into using 2 different multipliers using the MSR tool is there anyway to correct my situation using that tool? I guess it's supossed to be the same multiplier on all the cores at default though.

[k4vz0024]

I have a problem with Real temp 3, after Blue screen of Death (BSOD). Every time when I run Real temp3 I got massage Driver not Loaded Try Run as Administrator. Before the computer crashed it’s working fine. How can I fix that in Vista?

[unclewebb]

I thought your first comment would be something like this.

"Thank-you RealTemp for showing me that my computer is borked!"

No bios lets you individually set the multipliers for each core but when a bios is working properly, when you set the multiplier, it gets set equally for all 4 cores. Your present bios has a problem.

If the bios you are using was written before the 45nm Penryn chips even existed then you shouldn't be surprised that it doesn't properly support them. You need to update your bios.

If you enable EIST and C1E in the bios then RMClock and the Power Options might show some more options. EIST and C1E have no effect on your multiplier problem at full load.

The vDrop and vDroop you are seeing is normal. In the thread for my motherboard I was able to find a "pencil mod trick" where you write on some components on your motherboard to slightly change the resistance which reduces or gets rid of this issue. I know it's part of the Intel design and there is a good reason for some vDroop but I don't like it so I fixed it.

However the FSB does go up or down .1 to 1mhz regularly but I'm guessing that's normal.

You probably didn't see very much FSB fluctuating when using RealTemp. CPU-Z likes showing that random variation but clock generator chips are a lot steadier than that. CrystalCPUID seems to correctly report the FSB too.

Update your bios and if you can't get all 4 cores using the same multiplier then I'll add a line of code to try to fix this problem for you. If you use my MSR Tool then you can examine MSR 0x199. Type 0x199 into the MSR Number box and then click on Read MSR. Core 0 and core 1 will show an 8 in the third last digit and core 2 and core 3 will show a 6 in the third last digit. Go into the Copy To section and click on the Write button that is beside Core 2. Change the third last digit from 6 to 8 and then click on the Write MSR button and all 4 cores should now run at 3600 MHz.

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

If you enter numbers into this tool randomly and randomly click on the Write MSR button then your computer might crash or lock up but after a re-boot everything will be back to normal. Use at your own risk. If you follow my directions exactly, you won't have a problem.

Depending on various settings within Windows and whether you are idle or at full load, MSR 0x199 might change so don't plan to use this as a permanent solution.

[unclewebb]

k4vz0024: When you re-boot your computer a second time, does RealTemp work properly?

When you have a Blue Screen of Death (BSOD) it can screw up Windows or programs or drivers that programs depend on.

I would try RealTemp 3.20 to see if that makes any difference.
http://www.fileden.com/files/2008/3/...alTempBeta.zip

The WinRing0 driver that RealTemp depends on is an open source, third party driver provided by http://openlibsys.org/
RealTemp 3.20 uses an updated version of this driver so maybe that will help.

If the WinRing0 library has a problem you might have to use a program like Autoruns to delete the start up entry for it.
Once that is deleted, re-boot and try reinstalling RealTemp. I've never had this issue before so try the above and see if it works.

If you don't like Autoruns you could try using Regedit and delete this key and then reboot.
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Servic es\WinRing0_1_2_0

[k4vz0024]

I notice that, when I re-boot my computer and the system start up, than few minutes I can not run RealTemp and I got ”Driver not Loaded Try Run as Administrator”. However when is the system loaded completely (the hard disk not so busy so much) then some time take than I can run RealTemp and it’s work properly.

The Blue Screen of Death I think not caused by Real Temp

[unclewebb]

I suggest that you try using the Task Scheduler. It can be used to start any program automatically when you start your computer. You can use it to add a small time delay to give your computer time to boot up before RealTemp starts. That might help.

Here is a good tutorial about adding new tasks to the Task Scheduler:

http://www.vistax64.com/tutorials/13...eate-task.html

The Task Scheduler was introduced in Vista and is available in Windows 7 too.

Create Basic Task... is very easy to understand and to add a new task.
After you create a task you can go back and double click on it and edit the details.

You might also want to try disabling SuperFetch.
http://www.howtogeek.com/howto/windo...windows-vista/

On a computer with a slow hard drive it can make your computer feel very slow and take a long time to fully boot up and settle down.
Vista has a few more great ideas that can be disabled too.

[cheungtsw]

Where can I find Real Temp 3.20 ?
Thx.

[SimpleTECH]

Where can I find Real Temp 3.20 ?
Thx.

Lol, unclewebb has posted his link several times already.

Anyways, here ya go: http://www.fileden.com/files/2008/3/...alTempBeta.zip

[radaja]

looks like i finally got a cpu with decent temp sensors.there stuck at 27c/29c.
but thats alot better than my last one,they were stuck at 27c/39c.

[darkzone]

Looks like RT 3.20 works fine on i7. I have to mention again that CPU is under Scythe Ninja 2 fanless yet. Maybe when we'll have here 45°C outdoor I'll put a fan over, but I'm not sure. Yet

[kees1869]

Just tried 3.20 and when running a sensor test it crashes when it tries to quit/shutdown prime95. My whole system just locks up/hangs and I have to hard reset my pc. Anyone else have this problem? (I'm using Prime95 25.9)

[unclewebb]

I was using an older version of Prime95 when I originally tested that so I'll go try it again with version 25.9.
What operating system are you running and what CPU and what MHz and etc. Give me some details so I can see if there is anything on my side that might be causing this problem. What temperature do you max out at when testing?

Edit: I just ran a Cool Down Test using Vista 32 bit and Prime95 25.9 and there weren't any problems.



At the end of Test #8, it displayed "Stopping Prime 95", Prime95 stopped and disappeared from the screen. The only thing left was the green Prime95 system tray icon but when I moused over it, that disappeared too. Test #9 started after that and then Test #10 and then it was done.



I can't think of anything I changed recently that controls the shut down of Prime95 but I'll definitely have a look at the code. This is the first time I've heard of this issue. I'll also try and check Windows 7 x64 later tonight.

Edit: In Windows 7 beta Build 7000, the test ran fine as well. The only difference was that the system tray icon disappeared automatically like it is supposed to do.
Thanks MS for finally fixing that.



If you have any more issues let me know.

[OdinEidolon]

i have just bought a e6400 to toy with. TJ is set at 100C for Coretemp and 85C for realtemp (or viceversa, cant remember, i'm not at home now). which one is the right one? Mind that with the 85C TJ i get 40C under Prime95, with a 7V HDT-1283. Isnt it too low? thanks in advance!

[unclewebb]

OdinEidolon: When you get home post some more details. A screen shot showing Core Temp, RealTemp and CPU-Z would help out. There is no one number fits all correct TJMax for a series of processors. Intel refers to this as a TJ Target with an unknown amount of error. All software and even Intel can only guess. Post some more details and I'll take my best guess too and explain this confusing mess further. Some Dual Core CPUs at default MHz and a low core voltage do not get very hot, even when running Prime95 when you have a good heatsink attached to them.

[kees1869]

I was using an older version of Prime95 when I originally tested that so I'll go try it again with version 25.9.
What operating system are you running and what CPU and what MHz and etc. Give me some details so I can see if there is anything on my side that might be causing this problem. What temperature do you max out at when testing?

Edit: I just ran a Cool Down Test using Vista 32 bit and Prime95 25.9 and there weren't any problems.

I'm running Vista Ultimate SP2 (32-bit) on a Q6600 @3200MHz, temperature maxes out at about 65C (ambient 25C).

>EDIT<
Disabled all startup programs and now it runs fine, I'll try to find the culprit tomorrow and let you know.
>EDIT<

[angelreaper]

After sensor test

[unclewebb]

Someone finally noticed this CPU-Z Core i7 bug. Have a look at RealTemp. It's still reporting your MHz and multiplier correctly. They haven't changed.

RealTemp during the Cool Down Test uses the Intel Clock Modulation feature of the processor. Intel included this feature within their CPUs so if users are having trouble with too much heat, they can use software to thermal throttle their processor. That's what RealTemp does during this test to create the fixed power level steps. Technically speaking, the CPU is still running externally at the same speed but internally some of the clock generator's clock pulses are being ignored. CPU-Z gets confused and reports this as a lower multiplier which isn't accurate at all.

You can reproduce this behavior anytime. Just open up the Settings window in RealTemp and click on the Clock Modulation feature. Modulate your CPU down to 12.5% and watch CPU-Z go crazy.

The official term for this feature is On Demand Clock Modulation. This feature was first introduced during the Pentium 4 era to control heat. Here's an article from February 2005 that explains what's going on.

http://ixbtlabs.com/articles2/intel-thermal-features/

[drnip]

I sure new about it as it has reported these low MHz to me several times. It reported 12,00 soemthin MHz for eva200 once! All I got to say is thank god for unclewebb and his utilities. Best uncle in the world.

[nsegative]

So according to pyongmu temp reading of inside the cpu should TJmax for the q6600 still be 100c or lower for accurate temps? It seems load difference in temp was 10c and more!

[unclewebb]

My best guess for Q6600 is:

Q6600 Stepping B3, TJMax=90C
Q6600 Stepping G0, TJMax=100C

Not even Intel publicly agrees with that opinion. They say that there is a TJ Target that they shoot for but actual TJMax might be a little different than that, usually higher.

The best test is to run your CPU underclocked at 1600 MHz and 1.10 volts and compare your reported core temperature to your room temperature near your heatsink with your case open. Your reported core temperature should be 5C or so above your room temperature. Check out the RealTemp docs at TechPowerUp for more info.

[benniebeeker]

The best test is to run your CPU underclocked at 1600 MHz and 1.10 volts and compare your reported core temperature to your room temperature near your heatsink with your case open. Your reported core temperature should be 5C or so above your room temperature. Check out the RealTemp docs at TechPowerUp for more info.

does this count for water as well as air cooling???

[nsegative]

Cool thanks will do

[unclewebb]

does this count for water as well as air cooling???

The documentation should contain that information and give you a range of values based on type of cooler or air or water. rge even made a nice chart too.

http://www.techpowerup.com/realtemp/docs.php

When you can't find that link, do an About... in RealTemp and it will be very easy to find the above page.