SpeedFan Calibration Help

[jerreece]

So I sat down and tried to follow the CPU Temp Guide calibration deal that CompuTronix wrote up. I'm not sure if I'm even doing things right. LOL Maybe you folks can help me determine what I'm supposed to be doing.

Ambient = 23c
Chipset = 680i LT
CPU = Q6600
CPU Cooler= Xigmatek HDT-S1283
Frequency = 2.4 Ghz
Load = Prime95 - Small FFT's - 10 minutes
Motherboard = XFX 680i LT SLI
Stepping = G0
Vcore Load = 1.264/1.272
Vcore Idle = 1.168

Now, based on the (x + y) / 2 = z formula, my Tcase Idle should be 32c.
(16+2)/2=9
Ambient + Z = 32c

Now, at IDLE (CPU = 0%, 1600Mhz, 1.168V) I show:
"CPU = 33c"
Core 0 = 34 (ever since installing current heatsink, Core 0 has been higher than the others)
Core 1 = 31
Core 2 = 31
Core 3 = 33

Load Temps (CPU = 100% 2400Mhz, 1.264/1.272V):
CPU = 51
Core 0 = 52
Core 1 = 50
Core 2 = 48
Core 3 = 48

Obviously I don't have the proper spread in temps between my Tcase (CPU) and Tjunction (Core 0,1,2,3). I want to make sure I'm making the proper changes in SpeedFan to get myself set up properly. Now, normally I'd be running a 24/7 OverClock @ 3.2Ghz. But the calibration guide says to run everything on Auto settings. It would appear, that my Tcase is always 1 to 3 degrees HIGHER than my Tjunction. The guide suggests my Tjunction (at load) should be ~3c higher than my Tcase (CPU).

The guide also said, "Tjunction Load = Tcase Load + 5c."

Anyhow, I'm confused as to what I should be doing at this point. Could one of you more advanced forum members help me get things adjusted properly?

My Idle CPU reading was 33c. The formula suggests Tcase idle = 32c. So does that mean I need to offset my Tcase reading in SpeedFan by -1?

If so, that would alter my Load Tcase (CPU) reading from 51c to 50c yes? That still puts it in line with my Tjunction readings. Not sure what I should be doing.

 

[CompuTronix]

Ambient = 23c
CPU = Q6600
CPU Cooler= Xigmatek HDT-S1283
Stepping = G0

Now, based on the (x + y) / 2 = z formula, my Tcase Idle should be 32c.
(16+2)/2=9
Ambient + Z = 32c

jerreece,

You did it right except for a simple mistake. Idle power = 16 Watts, so as per the Table, X = 4.
(4 + 2) / 2 = 3
Ambient 23c + Z = 26c Tcase Idle

Also, be certain to follow the Test Setup to the letter. Your Vcore Idle is slightly too high, which translates into heat, and slightly inaccurate calibrations:

Now, at IDLE (CPU = 0%, 1600Mhz, 1.168V)

Vcore Idle in CPU-Z should be 1.10 volts. From the Guide:


Section 9: Calibrations

Part 1: Calibration - Tcase Idle

** Note 1: If BIOS does not respond properly to Auto Vcore, Frequency and Multiplier settings, then use an appropriate combination of manual settings to provide the following in CPU-Z:

CPU Frequency = 1.6 Ghz (65 nanometer processors)
CPU Frequency = 2.0 Ghz (45 nanometer processors)
Vcore = 1.10

The guide suggests my Tjunction (at load) should be ~3c higher than my Tcase (CPU).

Read it again. Tjunction Idle should be ~3c higher than Tcase Idle. From the Guide:


Section 9: Calibrations

Part 2: Calibration - Tjunction Load

(C) ... Tjunction Idle should be ~ 3c higher than Tcase Idle.

Note 2: Tjunction sensors were designed to be linear at high temperatures for Throttle and Shutdown protection, so Tjunction Idle could indicate too low or too high.

■ If Tjunction Idle is not ~ 3 higher than Tcase Idle, then disregard Tjunction Idle, and instead use Tcase Idle for accurate Idle temperature.

The guide also said, "Tjunction Load = Tcase Load + 5c."

Correct. Just re-calibrate SpeedFan with the appropriate Offsets, and your temperatures will be accurate. Refer to the following Example from the Guide:


Section 9: Calibrations

Part 2: Calibration - Tjunction Load

Example 2: Quad

Tcase Load = 45c
Tjunction Load = 50c

-Tcase/Tjunction-
--45--/--50--50--50--50--


Comp

 

[jerreece]

Okay thanks for the input there. I will try to adjust the Vcore settings on my motherboard here in the next couple days if I can. It being Christmas and all, the wife expects some family participation. lol

I'll see what I can get figured out.

Appreciate your quick response & feedback.

 

[CompuTronix]

You're very welcome, and best regards to you and your family.

Happy Holidays,

Comp

 

[jerreece]

Okay, I had to manually set my CPU Frequency to 1600Mhz, and set my Vcore to 1.104 Volts (closest I could get). Vdroop is bad on this board. Think I was actually set to 1.1325 which dropped to 1.104V.

Anyhow, here are my temps.

Ambient Temp: 22c
CPU: 30C
Core 0: 32*
Core 1: 27
Core 2: 27
Core 3: 30

*I couldn't get my system to 0% constantly, it kept fluctating between 0.4 - 0.7% CPU usage. Core 0 is the primary core so it's obviously slightly higher than the rest.

Ambient 22c + Z = 25c Tcase Idle (Z = 3)

So if I understand this right, Tcase SHOULD Idle at 25c. Which means my "CPU" temp reading is falsely high by 5c. Does this mean I should adjust my Tcase/CPU in SpeedFan by -5c?

EDIT: I went into SpeedFan under the "Advanced" tab. I adjusted the Temp 2 Offset for the CPU by -5.

Temp 1 didn't change anything. Temp 3 adjusted the "Aux" temp. Temp 2 showed a change on the "CPU" temp reading. So I think this was the right one. I guess my next step (assuming I've done things right so far) is to run the Stock LOAD test @ 1.25V?

EDIT TWO: Okay, I manually set my system to STOCK 2.4Ghz, and got it as close as I could get it to 1.25V. CPU-Z at idle shows 1.264V, and it drops to 1.24V at 100% load. However, my Ambient temps rose from 22c to 23c by the time I began the test (room is not climate controlled). Hopefully this will get me close enough for your formulas and such to work.

Here's what I showed after 10 minutes of Prime-95.

If I understand your Guide at all, in theory all my Core temps should be showing 5c higher than my adjusted Tcase/CPU temps. In this case, "CPU" shows as 44 (this was previously adjusted based on Idle tests).

So if Tcase is 44c, then Tjunction (core temps) should all be 49c.

In case the image above isn't showing, it reads:

CPU: 44c
Core 0: 51c
Core 1: 49c
Core 2: 47c
Core 3: 47c

So if I understand, essentially only my Core 1 readings are accurate. Therefore, I should be adjusting Core 0 by -2, and Core 2/3 by +2. Is this correct?

Assuming this is correct, I made those changes to SpeedFan for my Core temp readings. Here was the results of my IDLE test with the same CPU Speed & Vcore settings (idle it bumped to 1.264 however).

CPU: 28c
Core 0: 32c
Core 1: 31c
Core 2: 33c
Core 3: 34c

Again, it would appear Core 1 is my most accurate sensor. LOL! It was the one I never had to offset. The others that I offset to reflect accurate Load temps, now do not show the 3c difference to Tcase.

Anyhow, CompuTronix if you'd be so kind as to tell me if I did any of this right I'd appreciate it.

 

[jerreece]

Wow, so I have a side question for you CompuTronix, since you're so good with the temp issues. Someone just gave me a link to this article on the Tom's site in the UK.

http://www.tomshardware.co.uk/intel-dts-specs,news-29460.html

This article says that Intel has specified in October that the TjMax for the G0 Q6600 is 90c. Both CoreTemp & RealTemp have it listed as 100c. That throws off temperature readings by 10c!

Any input on this?

 

[randomizer]

Bleh, ignore the data in that article.

 

[CompuTronix]

jerreece,

You've correctly calibrated your temperatures, so enjoy your rig!

This article says that Intel has specified in October that the TjMax for the G0 Q6600 is 90c. Both CoreTemp & RealTemp have it listed as 100c. That throws off temperature readings by 10c!

No, in fact it doesn't. Randomizer and I are very familiar with that article.

During the Intel Developers Forum (IDF) in San Francisco in August, they released the Tjunction Max values for 45nm processors, but not 65nm. Then during Intel's next IDF in Taipei in October, they released the first Tjunction Max values for 65nm processors, which are shown in the article, and make no sense. Intel then updated their 65nm Tjunction Max values in November, some of which still makes no sense. This is why the author of Real Temp, (Kevin Glynn, username unclewebb), still uses Tjunction Max 100c for your Q6600 G0.

The single best source of current Tjunction Max information available anywhere on the internet can be found on the Real Temp thread over at XtremeSystems - http://www.xtremesystems.org/forums/showthread.php?t=179044&page=117 - where I see that you've posted a question for unclewebb, which he has answered. Real Temp and SpeedFan both use Tjunction Max 100c for your Q6600 G0, but since the Guide doesn't use Tjunction Max to achieve accuracy, there's no confusion about Tjunction Max, yet the outcome is always consistent with Real Temp, which proves that the Guide's calibration procedure really works.

Comp

 

[jerreece]

Awesome, thanks for your input! Especially around the holidays. Didn't expect such quick responses from ya around Christmas.

Maybe now that my SpeedFan is correctly calibrated, I can play with my OC a bit and see if I can get a 3.4 or 3.6Ghz stable.

Thanks again. Just wish Intel wasn't so flippant with their data. They really should be open about temperature data so users and manufacturers both have clear info upon which to base the safety of their systems. Shame on Intel...

 

[CompuTronix]

I couldn't agree more!

What is your VID?

 

[jerreece]

I've got the highest G0 Q6600 VID of 1.325 actually.

Which of course means to get a stable 3.2Ghz (at least on my 680i LT SLI board) I have to set the BIOS Vcore to 1.425 (which equates to 1.38/1.39 Actual)

 

[CompuTronix]

My previous CPU was a Q6600 G0 1.3000 VID, which required 1.4750 Vcore BIOS for 1.440 CPU-Z with Prime95 Small FFT's, and was stable at 3.6. Since you have the very excellent Xigmatek HDT-S1283, if you set BIOS so 1.5 Vcore isn't exceeded in CPU-Z while running Prime95 Small FFT's, then you might be able to achieve stability at 3.6 while maintaining 65c CPU / 70c Cores at ~ 22c Ambient.

 

[jerreece]

Cool. My biggest problem is this 680i chipset. I've had two of this XFX 680i LT board. The first one wouldn't OC my Q6600 AT ALL. I couldn't get anything stable above 2.4Ghz, period. That board one day just stopped working (while at default speed settings).

XFX replaced it, and the new board easily hit the 3.2Ghz overclock. This was with a different cooler, so I never pushed for anything higher (lets face it, things run fine at 3.2Ghz anyhow).

The other limiting factor of course, is my Ambient temp is related to PC temps. The longer my computer runs, the higher my Ambient temp in my office gets since we don't have Air Conditioning. Course, the heater is on this time of year anyhow.

Now if only I can find a way to OverClock my 123cc Snow Thrower...

 

[CompuTronix]

If memory serves me correctly, I seem to recall that Grimmy has had considerable OC experience with the 680i, so if you drop him a PM, I'm sure he'd be willing to help you. He has a Sticky in the "CPU & Components » CPUs" Forum: C2D Not Running Full Speed / Wrong or Low Multiplier - http://www.tomshardware.com/forum/247006-28-running-full-speed-wrong-multiplier - so you can PM him from there.

 

[jerreece]

Great! Thanks for the tip. Perhaps tonight I'll get some time to tinker and see if I can get into Windows with a higher FSB & Vcore, and start up Prime 95 for a while.