Real Temp - New temp program for Intel Core processors

[unclewebb]

I think the one you were testing the other day was a TJMax = 105.

Since I have zero documentation from Intel, I don't think I'll bother trying to identify these correctly. I'll use Intel Core i9 for all of 'em until about 5 or 6 months from now when I finally get to get in line like everyone else and have a look at the official documentation.

I'm just happy that I've been able to make it this far with the help of the XS community. The next beta is just about done. The new Settings window will cover more screen real estate than you can ever imagine.

[Monstru]

Hehehe, looking forward to that, nig monitors MUST have a purpose after all

[jcool]

I think the one you were testing the other day was a TJMax = 105.

Since I have zero documentation from Intel, I don't think I'll bother trying to identify these correctly. I'll use Intel Core i9 for all of 'em until about 5 or 6 months from now when I finally get to get in line like everyone else and have a look at the official documentation.

I'm just happy that I've been able to make it this far with the help of the XS community. The next beta is just about done. The new Settings window will cover more screen real estate than you can ever imagine.

All of mine are Westmere-EP cores like I said
Westmere-EP will be the DP Xeons, Gulftown will be the i9 (XE only?)

As for the model names, I don't think they are decided upon yet

[randomizer]

I'm not really sure how to interpret the data in the i7 Turbo log, as I'm not sure what thread it's gathering data from. So I'll let you look at it and maybe you can see what is going on, or where something doesn't seem right with my testing.

http://www.mediafire.com/?zntbwzfjtiq - No Affinity
http://www.mediafire.com/?zvgec24mtjt - Set affinity

I also noted the times it took to do the superpi runs (I did the load testing with 32M). After 3 loops of 32M:

1m 39.150s - No Affinity
1m 46.164s - Set Affinity

However, I also did a quick 1M run without logging the multi with i7 turbo, and the results aren't different beyond margin of variation (I ran each twice):

0m 14.110s 1M - Affinity set
0m 14.109s 1M - No affinity

If you decide to enable i7 turbo to log data from all threads (and let me know what part of the log is the info I should be looking at ) I can make a pretty graph in Excel showing the multi over time for each thread.

[unclewebb]

randomizer: I think that's what rge was finding as well that letting the CPU manage what core to run a task on gave the best performance. Locking tasks to cores like the good old days seems to be less efficient.

Your log file data shows me that I need to update the logging feature so it can log each thread individually. That's the only way I'd be able to see exactly what's going on during a benchmark. Then I think it would be very obvious when one thread starts running with the highest multiplier during a Super PI bench. I'll put that on the things to do list as well.

[randomizer]

Let me know when you're done and I'll get some more data

[BlindFreddie]

@unclewebb, post #3817: Could the reason that a single task runs faster if not locked to 1 core is that it gets to use multiple lots of L1 and L2 (?) caches (the non-shared levels)?
Just an idea.

[unclewebb]

http://www.sendspace.com/file/u1iihp

There are two versions of RealTemp in this download. If you have a new 6 core Gulftown CPU then try running RealTempGT.exe

It's officially called beta 3 but everything is hooked up now and I've done some testing so hopefully it is ready for prime time. If anyone has 10 minutes to spare, post a CPU Cool Down Test of these new 32nm sensors. I don't think they are quite as good as the excellent Core i7-900 series sensors but I'll have to wait for some retail processor testing about 6 months from now before passing final judgment.

If you see any bugs let me know.

When run on a Gainestown Dual Quad system, RealTemp GT should show the first 3 cores of each CPU. I'm using the numbering scheme 0 1 2 4 5 6 partially for this reason and partially because internally in the new 6 core CPUs, they are arranged in two groups of 3.

This program works fine on a Dual Core CPU but will only show 3 of your 4 cores if you have a single Quad. It's not designed for these so use the regular version of RealTemp.exe instead.

Fungus: I found a fix for the icons disappearing when you kill explorer.exe
This new code is in RealTempGT.exe and hopefully by tomorrow I can add that code into the regular version of RealTemp
Thanks again for bringing that problem to my attention.

[jcool]

As requested, here's a cooldown test on the Gulftown. RT GT Beta 3, ~30C water, load Vcore ~1,35V DMM.
Seems the sensors are working just fine.

[loonym]

Alternative download: http://lakesidepc.com/RealTempGTBeta3.zip

[unclewebb]

Everything looks good jcool. Thanks for the data.

These new sensors don't look as good as the Core i7-900 series but at least they're not all getting stuck like the 45nm Core 2 Quads and Duos do. Oh, I forgot. The old sensors don't get stuck, they simply become saturated at lower temperatures and might "bottom out".

loonym: I checked out the lakesidepc.com site. It looks like a work in progress. At least you got some decent software on that site.

[unclewebb]

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

This version includes a new right mouse button menu and a fix for the disappearing system tray icons when explorer is killed or crashes.

Core 2 idle MHz accuracy improved, especially for Core 2 mobile chips that can access the deeper sleep states like C3.

RealTemp GT updated and first official release.

A little more info added to the GPU button for Nvidia owners. If anyone has an SLI setup and would like to see individual info for each of their GPUs in this box and is willing to help a little with testing, let me know.

The RivaTuner plugin has been updated to support the new socket 1156 CPUs as well as Gulftown 6 core and Gainestown Dual Quad computers. I don't have access to any of these systems so if you have one, post some screen shots so I can see if the RivaTuner plugin is working as intended.

A Dual Quad Gainestown system might result in 37 individual RivaTuner graphs so be prepared to turn your monitor sideways before taking a screen shot.

[jcool]

Thanks unclewebb, will try on my Dual Gainestown later

[donmarkoni]

If anyone has an SLI setup and would like to see individual info for each of their GPUs in this box and is willing to help a little with testing, let me know.

I would like to help, just tell me what I need to do.

[unclewebb]

Thanks donmarkoni. I'll put your name at the top my SLI tester's list.

Hopefully by early next week I can send something your way that you can test and compare to other GPU monitoring software. At the moment, when using SLI, RealTemp will sample each GPU temperature sensor, compare them, and report the highest temperature. My plan is to allow a user to check the temps of each GPU individually in a new window accessed by clicking on the GPU button. I would still stick with just a single GPU temp number in the system tray like it is now.

jcool: I can't remember if I mentioned this before but the RealTemp GT version should report the first 3 cores of each Quad on a Gainestown system. I'm thinking about adding the last two cores to the system tray as an option as well as displaying all 8 cores in the pop up menu over the first icon in the system tray. This way you could keep an eye on all 8 cores without the GUI having to get any more obese.

The RivaTuner plugin won't correctly identify some of your mystery CPUs but it should be able to read them all correctly.

[donmarkoni]

Thanks donmarkoni. I'll put your name at the top my SLI tester's list.

Thank You. Looking forward to it! I'll be watching closely this thread (as I always do )

[KURTZ]

what about my sensors?

[unclewebb]

Well Kurtz, they look like typical socket 1156 sensors to me. Not quite as consistent as the sensors Intel used on their socket 1366 CPUs but they're not all getting stuck like on the old 45nm Core 2 series. The amount of slope error on these new CPUs is also pretty minor.

My theory is that the actual TJMax on your last core is probably closer to 102 or 103. If you don't like seeing the temperature on that core always sagged down then go into the Settings window and bump up TJMax from 99 to 102 or 103. After you do that, the 4 cores will report very similar temperatures when equally loaded with Prime 95 Small FFTs which seems logical to me.

If you drop your Minimum processor state in the Control Panel -> Power Options to a low number does your multiplier settle down to 9.0?
Edit: I saw your other post showing 9.0 so it looks good.

How about when C3/C6 is enabled, what does RealTemp show for your multi when running a single thread of SuperPI?

The Intel Spec Finder site doesn't have the turbo documentation listed for this CPU but I think the maximum multi is 22/22/25/26 when 4, 3, 2 or 1 core is active.

[Nightblade]

Love your work Unclewebb.

Feature request: Restore mode & window position on startup. (I have dual-monitors, not sure if that is a complication.)

[unclewebb]

You're welcome.

Can you give me a few more details about your setup? What OS, what version of RealTemp are you using, monitor sizes and anything else you can think of. What type of mode are you referring to?

I know there is a bug / issue in Windows 7 that when RealTemp starts up by using the Task Scheduler, it can ignore the previous screen location of RealTemp and will open it up in the top left corner at (0,0) instead. I haven't found a way to work around this issue quite yet but if you give me a few more details I should be able to come up with something.

I was just working on this recently but I forgot about the dual monitor users so the new code might be preventing RealTemp from starting up on a second monitor now. I was trying to prevent RealTemp from opening up so far off screen that it couldn't be accessed. I might have gone a little too far.

I'll definitely be looking into this issue in the very near future.

[Nightblade]

Win7, RealTemp 3.36, primary monitor 1680x1050, secondary 1024x768. I put RealTemp into mini-mode and drag it to the 2nd monitor after starting it via a desktop icon (every time lol). Oh and the 2nd monitor is to the left of the primary

[unclewebb]

Thanks for a few more details. Give me a day or three to have a look at this. I'll have to plug in a second monitor and do some testing to see if I can make some improvements.

Nightblade: Check your PM box. There should be a version of RealTemp in there that takes care of this problem.

[Nightblade]

Ooh, thanks

Forgot to mention that when I take RealTemp out of mini-mode on the 2nd monitor it "jumps" to the right, nearly off the screen. Top-right corner of the window appears on the primary monitor.

[unclewebb]

Forgot to mention that when I take RealTemp out of mini-mode on the 2nd monitor it "jumps" to the right, nearly off the screen. Top-right corner of the window appears on the primary monitor.

Does it do that with version 3.39.1 that I sent you or did it do that before? Time to start up my second monitor so I can have a look.

Edit: Now I see the problem. One more fix.

I think 3.39.2 should work correctly. No more jumping mini mode on the second monitor.
http://www.sendspace.com/file/osbkft

[stasio]

Download link error.