Real Temp - New temp program for Intel Core processors

[GeorgeVasil]

I think that the Core 0 Sensor get's Gay..<37.5% Load

The Sensor Test



Under Linpack Load Both Sensors shows the same temp..



so whats going on??

[rge]

Everest newest version and realtemp both report the same multi and mhz. cpuz is seemingly taking some type of average. cpuz is displaying .5 multis but also using effective .25 multis and .125 multis and even lower steps. Current pic of cpuz is 22 1/8 x 182. coretemp has bug.

EDIT: intel published a paper on computing mhz from turbo mode, next post.

[rge]

White paper explaining turbo...I pasted most important parts. i dont know if cpuz is doing this correctly, does not look like it. It says all active cores in cpu will operate at same frequency.
http://download.intel.com/design/pro...=tech_tb+paper
http://www.intel.com/technology/prod...demo.htm?iid=t

[bowman]

Turbo Boost is active even when RealTemp reports that my CPU is at 90C. I guess it's still running 'below operating thresholds' then.

[MJR]

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

Thanks for your help rge and msgclb getting this fixed up. Intel moved where the multiplier is located in Core i7 so I had to read the manual to find out where it went. I hate reading them.

I've noticed that CPU-Z reports 0.5 multis sometimes with Core i7 but as far as I know, they don't exist. rge thinks that it might be doing an average of all 4 cores or something like that.
Post some comparison screen shots so I can have a look.



The other new feature with this version is the ability to run an EXE or BAT file when an alarm goes off. This has been a hidden feature in RealTemp for a while but now you can access it easier in the Settings window.

I've included a 1 line bat file called RTShutDown.bat

It runs the shutdown.exe command and will shut down your computer in 10 seconds if an alarm goes off. When testing this feature I kept a command window open and typed in:

shutdown.exe /a

That will abort the shutdown procedure when you are testing. You can also edit the .bat file and change the number of seconds before shutdown occurs from 10 to whatever you want.

Depending on your operating system, this command might cause a hard shutdown so if you're working on a long Word doc you might lose some data if you haven't saved your work recently. Use at your own risk.

The last feature for the day was a new INI entry.

MinimizeOnClose=1

Some people like to have RealTemp minimize to the system tray when they accidentally click on the X close gadget. I think you should be able to figure out what the above does when added to your INI file.

So this has Vista x64 auto shutdown feature now?

[unclewebb]

So this has Vista x64 auto shutdown feature now?

You will have to try it and find out. It works for Vista x86.

Open the Settings window, click on the Alarm EXE button and point it to the file:

RTShutDown.bat

which was included in the download. Whether it works or not might depend on your permission levels, etc. It's easy enough to set a low ball alarm setting and then run Prime and see if it works or not. Where's my x64 tester burebista hiding when I need him?

Thanks rge for the Turbo info. Time to do some more reading.

The problem with reading the multiplier of a CPU is that the act of reading the multiplier, can wake the CPU up, which changes the multiplier. With Core 2 Duo I had to use a slight delay just before reading the multiplier so that if the CPU was at a lower multi because of C1E, RealTemp would be able to read it correctly. I might have to play around with this delay a little so it is more consistent while maintaining accuracy.

Edit: Vista testers might need to make an adjustment to their Power Options to get C1E to operate properly. RealTemp should show the reduced multiplier when idle, even with Core i7.



GeorgeVasil: The RealTemp test makes it pretty easy to spot the sticking point of your core0 sensor. My 8400 C0 needs about a quarter of a volt more juice to be stable at 450 x 9.0 but my sensors work a little better. Want to trade?

bowman: I think as long as RealTemp is reporting that your Thermal Status is OK, it will keep the Turbo on. Gotta love that. Kind of like having a sick friend who thinks the only place for the accelerator pedal is flat on the floor!

[bowman]

Heck, as long as it's working fine the temperature is just a number to me. In the mean time I can't play Tetris without melting the polar ice caps but that's okay..

[unclewebb]

Core i7 is just what everyone needs.
8 threads so we can all play Solitaire and surf the net without having to worry about any lag.

[GeorgeVasil]

GeorgeVasil: The RealTemp test makes it pretty easy to spot the sticking point of your core0 sensor. My 8400 C0 needs about a quarter of a volt more juice to be stable at 450 x 9.0 but my sensors work a little better. Want to trade?

Thanks for the answer M8.

Strange things here...at 4500mhz both sensors work excellent.

(because the temps are >40)

[randomizer]

Thanks for the answer M8.

Strange things here...at 4500mhz both sensors work excellent.

(because the temps are >40)

Notice that Core 0 hasn't yet reached 60C average distance to Tj Max. Until it gets that cool, you won't see it sticking Blast some fans at it and drop the temps a little and you should see it get stuck while core 1 continues to change.

[unclewebb]

GeorgeVasil: Nice fix. Run 'er at 4500 MHz and you'll never have to worry about a sticking sensor. You won't have to worry about too much of anything at that speed. I knew with your low volts at 4050 MHz that your chip was going to be a good one. Good work.

[GeorgeVasil]

GeorgeVasil: Nice fix. Run 'er at 4500 MHz and you'll never have to worry about a sticking sensor. You won't have to worry about too much of anything at that speed. I knew with your low volts at 4050 MHz that your chip was going to be a good one. Good work.

Yeah it really good chip. It needs 1.248 for 32M (Realtime),1.28 (Vantage,PCMark WPrime) and >1.336 Prime Blend (havent fully tested yet).

I think that the E0 cpus have mostly crappy sensors..(3 friends of mine have problems with their sensors) i am the 4rth one..

Thanks in advance!

[randomizer]

You should add a checkbox to Real Temp: "Fix sticking sensors". What does it do? Up the vcore and clock speed until you stop sticking! Who would want to run at stock anyway? *whistles*

[burebista]

Who would want to run at stock anyway?*whistles*

Pick me, pick me. In fact I'm not at stock speeds, I'm underclocked/undervolted.



PS. I just opened window and now are 2°C outside.

[randomizer]

You musn't do much with your PC if you run it like that all the time I can't lower my vcore except with the abit uGuru utility which overflows with fail.

[burebista]

For browsing/movies/music/office is overwhelming this way. EPU take care of my FSB and Vcore, and it does a wonderful job.
For games I put it at 3GHz and everything is fine.

[randomizer]

The uGuru chip is not supported by 99% of software. Speedfan only recently added support for something that is at least 4-5 years old, and all it does is read fan speeds and temperatures, it can't even control fan speeds. Only the abit software fully supports it, but the software stinks; it takes forever to start and isn't very stable. Is it so hard for abit to add some vcore options in the BIOS below VID? Well all hope is lost now that they've pulled out of the market.

At least I always have Real Temp for checking temps that sort of matter

[astaris]

@unclewebb
It is possible to have in realtemp a measure of the percentage of the time a core is in C3 or C6 state for i7?
In the pdf linked by rge, Intel explains how to take these measures by using the counters MSR_CORE_C3_RESIDENCY and MSR_CORE_C6_RESIDENCY (core level MSRs that run at the package level frequency, the same one of the TSC).

[xaver]

my two machines:

first Q9650 E0 @4.1ghz


second X3360 (Q9550 C1) @3.9ghz seems like it has one stuck sensor ... dont know what to do with it ?

[unclewebb]

@unclewebb
It is possible to have in realtemp a measure of the percentage of the time a core is in C3 or C6 state for i7?

I read that too and it does seem very interesting. The Intel Turbo document, brought to us by rge, has some excellent information in there. I'm just working at the moment with rge getting the RealTemp MHz working 100% with Turbo and C1E. rge says there is no other software at the moment doing this 100% correctly so it would be great to be the first one. Once that is done, I will check out those other features. Since Core i7 users don't have VID to look at anymore, they need something new to fill in the blank space.

The best and only thing to do with a stuck sensor is ignore it in the range that it gets stuck in. There's nothing else you can do. Intel doesn't approve of us trying to get core temperatures out of these sensors so if they all stick at lower temperatures, that's the user's problem and not Intel's. They're not designed for accurate core temperatures from idle to TJMax.

[unclewebb]

I had to create a separate little utility to understand what's going on with the multiplier in Core i7. MSR 0x198 that some software is reading for Core i7 is not accurate when using Turbo mode and C1E.

I followed Intel's guidelines and came up with this program to test things out.
Here's how my Q6600 looks at idle with C1E enabled and Vista Minimum processor state options at 66%



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

This tool also leads me to believe that most software may not always be correctly reading the multiplier for Core 2 based CPUs. I found a bug with my Q6600 that when the Minimum Power Option in Vista is set to 100%, two cores get locked at 9.0 like they should be but the other two cores float between 6.0 and 9.0. This new tool picks this up and will show you your average multiplier. I'm not sure if this bug is motherboard specific or maybe it's a bug in XP and Vista.

When C1E is enabled and the Power Option is dropped to 66% or less, both Dual Cores within my Quad drop to 6.0 so the ratio is 0.667 like it should be.

Give it a try and see what it shows, especially Core i7 owners.

I plan to add what I've learned into RealTemp later today for more accurate MHz for Core i7 owners.

Edit: The tool doesn't include any support for the 0.5 multis in some 45nm Core 2 processors. If you need that then let me know. It's mostly an i7 learning / testing tool.

[burebista]

E8400, underclocked, undervolted, C1E/EIST/C-State Tech enabled in BIOS.

[unclewebb]

burebista: It seems to be the Quads where things can get screwy. You can end up with different multis on each Dual Core within a Quad. If you see a number in the top ratio box that seems to be half way between then that is what's going on. This tester is mostly for Core i7 but I learned something new about Core 2 as well.

Have you tried out the new RTShutDown.bat file in Vista x64? I assume that it works but I haven't tried it yet. One user here was curious so I told him to give it a try but I never heard back.

[burebista]

Where's my x64 tester burebista hiding when I need him?

LOL, sorry man I've missed this.
Works like a charm, but I can't cancel because that cmd window disappears quickly. I've tried twice, but I can't catch the window open because it's closing too quick for me.

[unclewebb]

You're supposed to open the command window first.

You have to cancel the shutdown immediately after an alarm goes off or else your computer will go nuclear.

Thanks for testing.