Temp Calibration Confusion

texasnightowl

Distinguished
Feb 25, 2007
94
0
18,630
So I've read the sticky on C2D temps and for the most part I have a basic understanding I think. But I am working through the calibration section and find myself confused.

Info: e6750, stock with stock intel hsf. 1 rear 120mm fan exhaust. 2 front 92mm intakes. 8800GTS 320. Corsair 520 PSU. Antec Solo case. Currently trying Vista.

Working through Step 1, I have changed the vcore to manual 1.25 and the speed to 225 for a 1.8Ghz speed. My Antec tri fans have been set to high on their switches and fan control in bios turned off. Case removed, etc. C1E is disabled as is Speedstep (EIST?). But I am not sure what PECI is?

Moving on:

Part 2
(A)
In BIOS, my system temp was reported at 40 and my CPU temp at 31. Booted to windows and I believe it is Temp "2" in speedfan that is my tcase. At least, it was the closest. Temp 1 (System) was 39 and Temp 2 (tcase/cpu) was 28. So instead of being +2 of BIOS, it was -3.
(B)
Ambient temp is roughly 24-24.5c
(C)
Offset correction for tcase? Since I'm using stock, the guide recommends that Tcase should be ambient + 2c...in my case that would be 26-27c. However, longer system was idling after booting to windows, tcase dropped to 25 (0.5 over ambient). So, I am assuming I should add an offset for Temp 2 of roughly +2?


Part 3
(A)
Both cores were running equal so I just picked one at random.
(B - D)
Assuming the character count includes "0x" then the 5th and 6th charcters were 42.
(E-F)
Result is 66
(G)
G0 stepping - Tjunction Max = 100
Result: 100-66=34c Tjunction

Part 4 (the place where I seem to be really confused)

Quote:
If Tjunction hottest core fails to indicate 12c +/- 2c higher than Tcase, refer to the Offsets section. Configure SpeedFan for Tjunction = Tcase + 12c, hottest core."

I'm not sure what my offset is supposed to be here. For the tjunction just calculated, 34, it is obviously less than 12c different from tcase (27). The delta is only 7, not 12.

But given the instruction to configure SpeedFan for Tjunction = Tcase + 12, my Tjunction would be 39, given my tcase 27, not the 34 as calculated. Then, given the incorrect tjunction reported by speedfan (21), I would use an offset of 18?

Is this right?

For what it is worth, CoreTemp Beta reported my cores at 36c. So given the current cores as reported by speedfan at 21, I would just use an offset of 15, not 18?

Then after I make the offset 18 for each core I proceed to Part 5, right?

Basically, my question comes down to:

Is my offset of +2 correct for tcase?
Is my offset of +18! correct for tjunction?
 

CompuTronix

Intel Master
Moderator
texasnightowl, good job...you're almost there. My test data using the stock cooler actually supports enough deviation to go with Tcase = Ambient +1c, so I've edited the Guide to reflect "all coolers". It was a very close tie when I wrote the Calibrations Section, so it could've gone either way. In your instance, when Tcase finally settled at 25c, this is enough thermal gradient for 24c Ambient. I've also edited Calibrations, Part 1 to refine the steps, and added a Note regarding PECI, in addition to it already being mentioned in the Troubleshooting Section, Item (H). You're not the first person to ask about this, so the Note is a welcome addition to the Guide. Check it out.

Also, reviewing my Tjunction test data collected from a variety of C2D systems, I've changed the reference and relaxed the deviation, so now Tjunction Idle = Ambient + 12c, +/- 3c. This was also a very close judgement call when I wrote this Calibrations Section. If you configure a +15c Offset for SpeedFan so it agrees with Core Temp, your Tcase Idle to Tjunction Idle Delta should be 11c. Additionally, I think if you re-run Crystal CPUID and repeat RDMSR several consecutive times while observing Core Temp and SpeedFan, you may find that when it was initially sampled, it might have momentarily toggled to a lower value, which gave you the result of 34 rather than 36.

Have another run through the Calibrations Section, and let me know if you feel there are any other steps which may need refining.

Hope this helps,

Comp :sol:
 

texasnightowl

Distinguished
Feb 25, 2007
94
0
18,630


OK...I'm right in there I think BUT my tcase to tjunction delta is falling under the +12 +/-3. Here's my current data:
Ambient temp this morning was 23.5 to 24c...roughly 1/2 degree cooler than yesterday afternoon.

Edit: Thanks for the note on PECI...however, I didn't see it in my BIOS (Gigabyte P35-DS3R).

Test 1 and 2 - Idle at 1.8Ghz and tcase offset correction
Temps include a tcase +2 offset to bring temp to 1 over ambient and NO tjunction offset correction yet so core temps are obviously wrong.

Temp 1 (System): 39c
Temp 2 (tcase/cpu): 25
Core 0: 18
Core 1: 18

Part 3 - Tjunction DTS
Cores again running very in sync. Speedfan recorded both temps at idle at 18.
(D) 43
(E) 67
(H) 100-67=33

So, at this point we have tcase = 25 and tjunction = 33 so the delta is +8, just outside the (9 - 12 - 15) range.

Part 4
I decided to go ahead and set my tjunction offsets to 15 (33-18) and proceed to Part 5 to see where we are at.
(Note: 33 also matches coretempbeta value for core temps)

Part 5
vcore set to manual at the "normal" value of 1.325; speed back to normal 2.66

Temp 1: 40
Temp 2: 31
Cores: 39

So the tcase to tjunction delta at idle remains +8 as above.

Ran Orthos Small FFT's at Priority 9 for just over 10 minutes
Temp 1 maxed at 41
Tcase/Temp 2 maxed at 57
Cores maxed at 64

tcase to tjunction has fallen to +7

I let the system cool down to
41/32/39

tcase to tjunction at +7


Results:
tcase idle 31; load 57; delta 26
tjunction idle 39; load 64; delta 25

So am I not doing something correctly? Or is my tcase to tjunction just less than others...outside the normal results?

I think I'm about to say this is it for the moment and get a better hsf and see what I get. Thinking of the Arctic 7 (pro: wouldn't have to take my system out of the case) or the Thermalright Ultima 90 + fan.
 

CompuTronix

Intel Master
Moderator
texasnightowl, I'm very confident that you've arrived at the correct outcome for your rig, however, it appears that BIOS thermal calculations are not linear for Tcase by ~ 5c, which unfortunately is a common problem. You can check for a later release and flash it, which may correct Tcase linearity. Since Tcase to Tjunction Delta increases with Load, Tcase Load is too high, and should indicate closer to 50c, relative to Tjunction 64c. In terms of perspective, G0 Steppings sometimes indicate Tcase to Tjunction Delta on the low side, so other than perhaps dropping your Tcase Offset by 1c, your Idle is close enough. With respect to overall accuracy, Tjunction Load and Tcase Idle are the most critical temperature, so you're OK, but I agree with you about upgrading from the stock cooler.

Hope this helps,

Comp :sol:
 

texasnightowl

Distinguished
Feb 25, 2007
94
0
18,630
Thanks! I'm going to go ahead and order a better HSF and then I'll step through everything again. Thanks!

(Note: Yup, BIOS is already flashed to latest...did so before loading Vista.)
 

texasnightowl

Distinguished
Feb 25, 2007
94
0
18,630
I think it is good. I like the note to let the system idle 10 minutes after boot into Windows. Adding the note about PECI to that section is good too.

A note about software processes might be helpful. Since so many of the other settings at that point are in the BIOS, moving Software Processes to the top of the list with a note (ie. remove start up programs from programs/start menu or disable them from the registry run key or to just kill them at start up) might be appropriate.

Then in Part 2, item (B), maybe clarify that the reason you are looking at the CPU temp in BIOS is to figure out which Temp x setting in speedfan is the cpu/tcase? Having read the whole thing that is obvious but as a reminder in that section.

On the whole it is great! I was obviously able to make it nearly all the way through but the low idle/tjunction delta I had was confusing me! Thanks for your work on it!
 

robbomate

Distinguished
Mar 22, 2007
4
0
18,510
I also am experiencing Tcase to TJunction delta of about 10c across all loads and overclocks.

For example with my E4300, P965, High Performance Air Cooling

@ 300Mhz FSB, 3.25 Vcore:
100% load: Tcase: 49c TJunction: 60c
Idle: Tcase: 30c TJunction: 40c

@ 333Mhz FSB, 3.325 Vcore:
90% load: Tcase: 58c TJunction: 67c
Idle: Tcase: 35c TJunction: 45c

I am worried about pushing my CPU to 100% load @ 333Mhz FSB because the TCase will exceed the 60c recommended as a maximum in the C2D Temps Guide. However, my Tcase to TJunction delta does not increase at higher loads. Does this mean my M/B is calculating the wrong temps or should I limit myself to a 300 Mhz overclock? Perhaps my fan is not attached properly, although my TJunction temps seem to indicate adequate cooling.
 

texasnightowl

Distinguished
Feb 25, 2007
94
0
18,630


So, that is where I was before. Tonight I removed the stock hsf and put in a Thermalright Ultima 90 with a Panaflo 92 fan. The fan pushes air thru the hs to the exhaust.

I did not re-do Steps 1-4. I kept with the offsets previously established. tcase=+1 and tjunction=+15

Idle with the new Ultima 90
tcase: 23
tjunction: 31

So these both reflect a drop of 8 from stock.

I'm going to have to adjust the tcase because ambient is 24. So tcase will have to become +2 instead of +1.

Load (Orthos Small FFT, Priority 9, over 10 minutes)
tcase: 45 (was 57 stock)
tjunction: 49 (was 64 stock)

So I have drops of 12 and 15 degrees respectively. These temps I can more easily live with. The tjunction temps in speedfan do continue to match coretempbeta's reading. I definitely don't have anywhere NEAR a +12 tcase to tjunction delta though.

I suppose the next steps would be either 1) put a 120 fan on instead of the 92 or 2) lap the hs and/or cpu (which I'm not ready to go for yet).