Real Temp - New temp program for Intel Core processors

[unclewebb]

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



After many years, I finally bought an over priced QX9650 so I could get RealTemp fixed up. The C0% Load meter should finally work correctly on QX CPUs and the multiplier should be reported correctly if your CPU has internal timers that are functional. Some of the ES processors might be broken.

Post a screen shot if there are any problems.

Thanks for the data you provided me long ago JohnZS. Sorry it took so long to get around to fixing this.

Edit: i7 Turbo GT has been updated with Gulftown and QX support. It's ready for some beta testing over here:
http://www.xtremesystems.org/forums/...94#post4401694

[vodka3dg]

ouch... that must have cost you a leg and half an arm

thanks for the effort u're putting on this project uncle!! much appreciated.

testing 3.59

[JohnZS]

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

Thanks for the data you provided me long ago JohnZS. Sorry it took so long to get around to fixing this.

Thanks Uncle

I have bad news for you, Realtemp 3.59 reports incorrect multiplier and % load values on my C0 stepping QX9650

Any ideas?

Thanks and once again thanks for your continued support and efforts.
John

[unclewebb]

I always knew your QX9650 was unique. Can you post a screen shot at full load. It seems to work correctly on my C1. At least now I have some hardware to test on so maybe I will make some progress this time. I was working with another user who has a QX9650 - C0 and he was happy so now you've got me curious.

Edit: Here's a couple of tests I did. I ran 3 threads of Prime95 Small FFTs so that gives a very consistent 75% load + an extra percent or two for any background activity. The QX9650 has a default multiplier of 9.0 so I tested with a multiplier both higher and lower than the default and I used the half multi in both of these tests. On this QX9650, the C0% based load meter looks correct and the multiplier is also being reported correctly. If you get a chance can you try running this same test on your QX or any other test you can think of that shows a problem? Thanks John.

ouch... that must have cost you a leg and half an arm

shhhh! Don't tell the wife.

[JohnZS]

Hi Unclewebb
Thank you once again for your continued efforts. The multipliers are now reading properly after a reboot.
Not sure if this is of any use, but have a look at this.

I will try and do some proper loading with PRIME95 when I have the time to, but does it appear loadtester is not loading anymore?
Multipliers are working though
Before they were all over the place, jumping from 7 to 9 when in theory it should change from 6 (idle) to 10 (loading). After a reboot all seems well, perhaps a random glitch somewhere...
John

[unclewebb]

The Load Tester program is only designed to load 1 core. You would have to run 4 instances of this program to load each core. Prime95 is much easier.

A 90% load on one core of a quad is equivalent to a 22.5% overall load. That plus 3% load for other background processes like CPU-Z and the rest of Windows for a total load of 25.5% is reasonable.

The C0% number is not accurate on some Core 2 CPUs when lightly loaded if the motherboard supports deeper sleep states. On a Core 2 if the calculated C0% is way higher than the actual load at idle then you would be better off using the TM Load option in the Settings window. That will use a Task Manager like load meter instead. The mobile Core 2 can report a very high C0% at idle. On those it is more of a measure of how much the cores are able to go to sleep and save power so a high C0% at idle is a good thing.

The multiplier that RealTemp reports using the internal timers is an average multiplier. If you are using SpeedStep and your idle multiplier is 6.0, when your CPU is partially loaded, the tasks will be rapidly bouncing around from core to core and the multiplier can also be bouncing around all over the place. If the RealTemp multiplier is 100% accurate at full load, then it is also accurate at idle and more accurate than CPU-Z when a CPU is partially loaded.

When testing check out turning on C States. CPU-Z completely ignores what's happening inside your CPU. The multiplier at idle can be bouncing around all over the place. Using the high performance timers within the CPU can provide a very accurate look at the multiplier without needing to put any significant load on the CPU.

The Control Panel -> Power Options -> Minimum processor state controls the CPU multiplier at idle. Check that out when testing.

[caaront]

On my Vista 64 bit it reads 2 degrees hotter than all my other performance programs, other than that it is perfect.

I always run it while doing my OCCT tests, because it displays highest temp reached, so I can walk away and when I look again I might see it's currently at 77C but did hit 78C at some point.

[unclewebb]

On my Vista 64 bit it reads 2 degrees hotter than all my other performance programs, other than that it is perfect.

Have you ever thought that maybe RealTemp is right and all those other programs are not?

The core temperature can change instantaneously. Most temperature monitoring programs sample the same sensors but they only sample that sensor once per second. A degree or two difference is nothing to be too concerned about. Run a program like Prime95 Small FFTs which equally loads the cores. It results in very consistent core temperatures so all programs should report the same. Post a screen shot of this test if you are still having problems.

[Purgatory]

Hey ladies&gents...

First of all thank you very much for this usefull tool i use everyday unclewebb, i recently found out about this thread and saw that you are the creater of Real Temp...

Second of all, i'm seriously new with my G15 V2. and i'd like to know can i use any tool, plugin,...etc to use my G15 display to show RealTemp...

Thank you again and keep up the good work...

[Purgatory]

Anyone ? o_O

[unclewebb]

RealTemp includes a plugin that works with RivaTuner. I used to use this so I could see some data on my G15 so I assume that it would work on your V2 if you want to run RivaTuner. You don't need to have RealTemp running. The plugin is independent. All you need to do is install RivaTuner and then open up the Settings window in RealTemp and tell it where you have the RivaTuner.exe file located. This will install the RTCore.dll plugin into the correct directory of RivaTuner. Then when you open the RivaTuner graphs you should be able to configure it so it uses the RTCore.dll plugin.

I've received very little feedback from users about this plugin so if there are any bugs with it reporting information for your CPU then let me know and I'll see what I can do to fix things up.

The plugin was recently updated and can be downloaded separately here.
http://www.fileden.com/files/2008/3/...507/RTCore.zip

Unzip it and copy the files into the RealTemp directory and then use RealTemp to install it to RivaTuner or you can copy the files directly to RivaTuner if you know where they go. Post a screen shot if it works. It's been a while.

[burebista]

RT plugin is perfect but unfortunately RivaTuner is useless now for his HD5870.
Maybe starting from July we'll see that much awaited RivaTuner 2.25.

[Purgatory]

Hmm... I guess i can't use RivaTuner with my 5870 as "burebista" said...

Maybe you can make a G15 V2 application using LCDStudio or something to make RealTemp G15 friendly, if you have some time of course...

But i'll try RivaTuner+plugin regardless of my 5870...

Thanks for the responses and helps...

[Wishmaker]

I decided to change my multiplier from 20 to 19 and I am running some stability tests now. I noticed that RT is showing things it did not show with my x20 multiplier. If, beforehand, I used to have a delta of up to 2 degrees between cores, not more, now I am faced with deltas up to 8 degrees . Is this normal? I mean for the 20 multi I was using 190 bclk and for the 19 multi I am using 200 Bclk.

[unclewebb]

Wishmaker: I wish I had some good answers for you but I've never heard of this issue before. It would be interesting to see some Prime95 Small FFTs testing. That's the best test I've found for equally loading each thread. When CPUs are not equally loaded, you can have some random and rapid temperature changes as tasks move from thread to thread and core to core.

LinX is great for creating heat but the load is not consistent for temperature testing and comparisons from one core to the other. Post some pics of P95 Small FFTs of both of your setups so I can see the difference that you are seeing. Try to run each combo for at least 10 or 15 minutes or until the core temps are stable.

[Falkentyne]

I Just tried the 3.40 version and there were two problems:

1) it read my QX9650 at 364x12 for 4300+ mhz), instead of the correct 364x11 (4009 mhz).

2) it crashed when I pressed "reset" for the temp readout, every time.

Previous one I used was 3.07 beta, which had been working fine.

[Nightblade]

Feature request: Option to change the update interval.

[klataa]

Hi, i dont know if anyone wrote that but:

Feature request: make Realtemp to show Ati temps

[unclewebb]

Feature request: Option to change the update interval.

Why? More frequent or less frequent.

klataa: Have you tried RealTemp 3.59?

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

Just add GPU=2 to the RealTemp.ini configuration file and you might be in for a surprise. It even reports CrossFire correctly last time I checked.

Falkentyne: I sent you a PM with the latest version and fixes you mentioned. I finally have a QX to test with so things are much improved.

[SimpleTECH]

Hi, i dont know if anyone wrote that but:

Feature request: make Realtemp to show Ati temps

http://www.xtremesystems.org/forums/...postcount=4044

[Nightblade]

Why? More frequent or less frequent.

Less frequent, to reduce the number of hardware probes.

[klataa]

Works perfect! thx But i can see only temp of my primary gpu

edit

after couple restarts Realtemp shows all temps nice program!

[unclewebb]

With CrossFire, there is a power saving setting that disables the second GPU when you are sitting at the desktop. RealTemp can't ask the GPU for its temperature when this happens. The ATI Catalyst Control Center will do the same thing when the second GPU goes to sleep. As soon as you start using the second GPU then it wakes up and starts delivering temperature information to RealTemp. There is a registry mod you can do to keep the second GPU constantly awake and reporting temperature information but you'll have to do a Google search because I can't remember it off hand. Waking up the second GPU when you are sitting at the desktop is kind of pointless. As soon as you start using a 3D application that uses the second GPU, RealTemp will be recording its temperature.

Edit: Here's some info about the registry mod needed if you want to waste power and keep your second GPU awake all the time.

http://forums.amd.com/game/messagevi...hreadid=128135

Less frequent, to reduce the number of hardware probes.

Reading the core temperature of an Intel CPU is extremely efficient. The information is cached so you are not constantly reading a slow sensor like reading an ATI GPU causes. Reducing the polling interval of the CPU core temperature will make zero noticeable difference. Getting the GPU temperature from the ATI driver is a slow, flaky procedure so that's why I included the ability to reduce the polling interval of that. Nvidia doesn't have issues like that.

If you really, really need this feature then I might create an INI file option to reduce the polling interval. Try running RealTemp minimized to the System Tray for an hour or two and then have a look in the Task Manager and you will see that it is a very efficient program that doesn't put a lot of stress on your CPU. Getting the CPU temperature once a second or once every 5 or 10 seconds isn't going to be a difference that you'll ever notice.

[Nightblade]

What about probing the GPU sensor?

[unclewebb]

What about probing the GPU sensor?

There is an adjustment for that in RealTemp if you have an ATI GPU. I've never seen any problems with probing Nvidia GPUs? What type of GPU do you have?