C2D L2 stepping temps readings

[dr_kuli]

Hi,

I'm a bit confused with it. Asus P5B-E reports CPU temp HIGHER than Core 0/1 temps (according to SpeedFan 4.32). CoreTemp v.0.95 reports Tjunction as 100C and core 0/1 temps at full load (Orthos @P 9 Small FFTs, TAT doesn't run on my rig) as 81/81 C after 30 minutes! SpeedFan 4.32 reports core 0/1 as 66/66 C and CPU temp as 78 C in the same moment. Concave IHS??? Any ideas/comments?

Specs:
C2D E6300 L2 stepping, Q702A120 SL9TA, Asus P5B-E BIOS 1202, FSB 450/CPU 3150 Mhz, BIOS Vcore 1.3875V, idle Vcore in WXP 1.352V, load Vcore in WXP 1.288V, delta temp idle/load higher than 30 C, Scythe Ninja Plus rev. B + AS5, well vented case (2xScythe Kama Bay intake, 2x Arctic Cooling Fan 3 TC exhaust) + Seasonic SS-401HT PSU w/120mm exhaust fan.

[/img]

 

[CompuTronix]

Read the Core 2 Duo Temperature Guide.

 

[dr_kuli]

I did, few times. Great work. I've read SpeedFan & CoreTemp documentation & forums also. The only thing which applies here is

(E) Any hardware and / or software may misreport Tcase and / or Tjunction temps.

(F) 965 chipsets may misreport Tcase and Tjunction temperatures with +/-15c offsets.

and

(I) C2D’s manufactured with concave Integrated Heat Spreaders may report high Deltas and temps.

(J) An improperly seated CPU cooler is the leading cause of abnormally high temperatures.

Still have no idea, except concave IHS...

 

[dr_kuli]

OK, so I took off the mobo, disassembled CPU & cooler, cleaned it with alcohol, checked that my IHS is NOT concave (against thick glass), applied AS5 as per manufacturer's guide, mounted CPU & cooler, double-checked everything, and so what I've got:

Tcase = 39C Idle & 71C Load
Tjunction = 46C Idle & 73C Load
Ambient = 22C
Chipset = i965P
C2D = E6300 L2 stepping
CPU Cooler = Scythe Ninja Plus rev. B
Frequency = 3150 MHz
Load = Orthos P9 Small FFTs
Motherboard = Asus P5B-E, BIOS 1202
Vcore = BIOS 1.3750V, WinXP 1.34V idle & 1.28V load

SpeedFan was adjusted (15C offset) to show exactly the same Tjunction as CoreTemp 0.95. After 15 minutes of load I've measured the temp on the edge of copper plate (under the heatpipe, plate has direct contact with IHS) at the bottom of Ninja. It was done with digital thermometer, which is officially approved as +/- 0.5C error max. The temp measured was 32.5C. I've touched the heatpipes near the plate with my fingertips - heatpipes were cold! Could anybody explain me this simple physics (thermodynamics exactly)??? I personally don't believe in such miracles as nearly 40C delta inside solid copper plate at the bottom of Ninja cooler!!! Please...

 

[CompuTronix]

dr_kuli, sorry it's taken so long to get back to you. Just a few considerations before we proceed.

(1) How did you measure Ambient?

(2) Did you test with all fans at 100% as per the Guide?

Testing

Vcore = Manual
C1E / EIST = Disabled
CPU Fan = Manual, 100%
Computer Case Fans = Manual 100%
Primary Test = TAT @ 100% 10 Minutes
Alternate Test = Orthos @ P9 Small FFT’s 10 Minutes

(3) Where positive or negative Tjunction offsets applied to SpeedFan?

(4) Tcase is presently offset by +15c and should be corrected, which would provide an appropriate Idle temp of 24c.

Hope this helps,

Comp 8)

 

[dr_kuli]

Thanks a lot for your answer.

@1) With another digital thermometer and one more analog. Both were at the top of the closed PC case.

@2) Q-Fan is disabled. Kama Bays intake are specified 30 CFM each, both run full speed 800 RPM. Ninja Fan is specified 49 CFM, runs full speed 1250 RPM. Seasonic PSU fan is specified 61 CFM at full speed, but runs way below full, as I can hear - it has its own TC sensor, no RPM monitoring. AC 3 TC fans exhaust are specified 28 CFM each, at full speed, both run about 1700 RPM at CPU load, they've TC sensors, their max RPM is 1900.

@3) SpeedFan was adjusted (INTEL CORE chip, 15C offset positive) to show exactly the same Tjunction as CoreTemp 0.95. But I think it's overkill, without any offset it shows the same as CoreTemp 0.94 beta (TjunctionMAX 85 instead of 100 as in CoreTemp 0.95), which is more believable. Now I've quitted this offset.

@4 OK, I've corrected it, seems more normal...

So now it's:

Tcase = 24C Idle & 56C Load
Tjunction = 31C Idle & 58C Load
Ambient = 22C
Chipset = i965P
C2D = E6300 L2 stepping
CPU Cooler = Scythe Ninja Plus rev. B
Frequency = 3150 MHz
Load = Orthos P9 Small FFTs
Motherboard = Asus P5B-E, BIOS 1202
Vcore = BIOS 1.3750V, WinXP 1.34V idle & 1.28V load

Delta idle/load Tcase is 32C, Tjunction 27C, seems to be OK. But it's still too small difference between Tcase & Tjunction... I think this L2 stepping has to do something with it... Any thoughts?

One more time thanks a lot.

 

[CompuTronix]

dr_kuli, I appreciate the detailed information you have provided. I agree with your action of removing the +15c Tjunction offsets and restoring to zero. Just one additional question; what is the intake temperature outside the case near the lower front area?

Thanks,

Comp 8)

 

[dr_kuli]

The case is on the floor, has 50 mm wheel set, 8x5.25" drive bays on the front, lower 6 of them are occupied with Scythe Kama Bays, intake front temp at the center of lower Kama Bay at the distance of 10 cm is 22.6C, at the top of the case is 22.9C.

 

[CompuTronix]

dr_kuli, you do indeed have one of the most interesting temperature issues I've seen. I believe that your Tcase Idle and Tjunction Load (Orthos) temps are appropriate, as per your Vcore, CPU cooler and computer case configuration, however, it's obvious that both sets of temps show nonlinear scaling, as seen with some 680 boards.

With respect to BIOS, Super I/O chips, and chipset drivers, newer isn't always better. Tcase Load remains 10c too high, and Tjunction Idle remains 5c too low, so I suspect that all is not well within this particular Asus binary bit bucket. I am continually amazed by the lack of debugging the hardware / firmware / software OEM's fail to invest in their products, which we, the end users, must by default, accept, if we want a PC.

Unless you are inclined or otherwise inspired to devote untold hours of research, and trial by fire, (no pun intended), to test various combinations of flash-uninstall-install code-from-hell into your creation, my advice to you, sir, is this; you have one relevant and critical temperature which does make sense - Tjunction Load. Configure SpeedFan to "Show in Tray" the hottest core, and declare a digital victory.

Hope this helps,

Comp 8)

 

[dr_kuli]

you have one relevant and critiacal temperature which does make sense - Tjunction Load. Configure SpeedFan to "Show in Tray" the hottest core, and declare a digital victory.

I did :lol: I will definitely try to go higher with FSB and (most probably) Vcore, then I'll post how it scales. Thank you for your help, may COLD be with you :lol: