E6400, ultra-120 and strange temperatures.

[supox]

Greetings silencers,

I've somewhat of a problem with my E6400 operating temps. Currently, the setup is such:

Ambient temp of 23C
P180 SPRC edition
P5B Deluxe/Wi-Fi (0614 BIOS if memory server me correct)
E6400 running at 8x333
Ultra-120 & AS5 (absolute minimum applied only)

Fans etc.:

2x Noctua NF-S12-1200, one in the back and the other attached to the Ultra-120
1x Scythe S-Flex SFF21D (the 800RPM one) in the top exhaust
Also removed the plastic grills which are in front of the dust filters, and took away the metal and plastic covers from the 5.25 bays.

And the current sorry state of affairs:

Idle:


Load (note the time stamps in both TAT shots):



Noteworthy things:

When the CPU is <50C, speedfan is set to 50%, or ~800RPM for the Noctua fans. The Scythe is -not- controlled, it's always running at 12V. When the temperature exceeds said limit, the fans will ramp up to 12V.

If I open the case door, CPU idle temps drop by 3-4C. Difference under load was about 1C.

My thoughts as of now:

I've installed the Ultra-120 twice to make sure thermal paste doesn't mess with heat transfer. And before I put things together, I also verified that the base of the HS was not uneven, since I had read about such cases.

The only thing left that I have not checked is whether or not the heatspreader of the E6400 is, in fact, warped and the reason behind these rather high temps.

Could really use some advice here...

Thanks in advance.

edit:

Here's a picture of the innards. And, as usual, I forgot to mention a component back there - the video card is a passive 7600GT (57C idle, highest I've seen was 65C after some hours of gaming).

[J. Sparrow]

If I open the case door, CPU idle temps drop by 3-4C. Difference under load was about 1C.

Hmmm, this seems the weirdest bit of your account: with good airflow inside the case, you should see lower temperatures with the door closed.

Assuming there's good contact between the IHS and the heatsink, I would try a quick intake duct for the U120 and swapping fans, maybe it's true what they say about the Noctua (poor static pressure); try putting the S-Flex on the CPU.

[supox]

I'll try that fan configuration.

By the way, I meant the front door, and not the side panel. Just in case there's any confusion.

edit:

Okay... maybe 1C lower at idle & load. I'm all out of cardboard so I can't try building a duct right now.

Perhaps I really should remove the HS. Any tips on how to properly test if a surface is truly flat and level are welcome - though I'll be searching for the info anyhow, since it has been propably discussed previously on more than one occasion.

edit2:

Oh yes, I did update the BIOS (1004) just to be sure. No difference, as expected.

[vizzle]

I believe the best way to test if the IHS on your CPU is flat is to take a razor blade and put it on top of the IHS and flash a source of light behind it....and see if there is any light that makes it through.

If some light does make it through, then your IHS might be a bit concave, but if not, then I believe you can consider it to be flat.

If anyone has any other recommended ways to test this, I'd be curious to know them as well.

[Felger Carbon]

I'm all out of cardboard so I can't try building a duct right now.

Remove the 4 cellophane-wrapped sticks of Ritz crackers from their box. You now have the highest possible quality cardboard available for ducting. Wait! Eat the crackers after making the duct!

Only a rude person would designate the result a Ritzy duct.

[supox]

Perhaps it won't be necessary to take any more steps. I had foolishly left the vcore on 'auto' - which in this case meant 1.375V. By lowering it to 1.325V, temps dropped to about 40C idle, and 55C under TAT load. Less when using Orthos - after 20 mins it's still at ~51C. All fans are still running at ~800RPM, too.

Figures though. It's been a while since I've bothered with overclocking. Last subject was the famed celeron 300A.

But having said all that, a CPU duct seems to be something to work on. Can't buy any Ritz crackers until tomorrow tho, shops are closed.

Thanks everyone.

[supox]

*bump*

After some more fooling around, I've managed to drop idle to 35-36 and load to 44-46.

I lapped both the Ultra-120 and the IHS of my CPU first, but that didn't show much of change (=within margin of error). So what I did was apply AS5 using the latest instructions - a thin strip across the CPU in my case. However, I also considered the possibility that the heatsink mounting would not provide adequate pressure, so I placed 3mm thick rubber washers (scavenged from a SKL3000B's HD tray) between the backplate and the retention clip. I tried to wiggle the HS around a bit, and it seemed to fit more tightly this time.

I've also gone from a 333 FSB to 375 during this process.

[thejamppa]

It seems most tool free attachments these days do not give enough adequate pressure. Its sad thing that even some screw attachments do not give enough pressure...

[bajker]

Interesting thread, it's like reading my own story. I have C2D 4300, Asus Commando, Ultra-120 and also P180, three Nexus fans (one in lower chamber, one on HS and one on back exhaust).

Before installation I checked flatness of CPU's heat spreader with digital caliper's handle against the window and thin stripe of light was clearly visible in the middle between the handle and CPU, so I lapped it with 600 sand paper. I also checked Ultra's flatness and there was light stripe again, so I lapped HS too. I checked connection between HS and CPU and it seems very good.

So, I thought that I did everything possible to ensure low temps and quiet operation. But I was wrong, temperatures are not what I expected, I read Core 2 Duo Temperature Guide and explain my problems in this thread, but did not find any good explanation.

After reading about your experience, the only explanation left is that HS is not sitting well on CPU and pressure is not adequate. I read about proper AS application on AS's site, but did it on "old fashion way", thin layer across heat spreader. Now, I will try again with thin stripe method and also increase pressure with some washers. Let you know how it goes.
Thanks.

[Felger Carbon]

When the CPU is <50C, speedfan is set to 50%, or ~800RPM for the Noctua fans.

50% of 1200RPM is 800?? Have you tried just running the Noctuas at 100%, not controlled by Speedfan, just to see what happens?

Also, I worry that you used too little thermal grease. But that's secondary to the fan cooling.

[Plissken]

I got some strange high temperature readings with my E6300 and did a lot of searching on the net for clues. It's a perfectly stable system with barely warm heatsinks. Something ain't right.

I've come to the conclusion that you cannot trust TAT or Speedfan's "CPU" temp with the E4300s or the newer E6300/6400s with the Allendale core. For Speedfan this is especially true when it comes P5B or Commando motherboards. Someone on xtremesystems proved that an E4300 CPU does not throttle above 85C as reported by TAT... not even close.

TAT is reading my E6300 as a Pentium M, which I believe has a Tmax of 100C (not 85C). Therefore it's reading of Tjunction is 15C above Core Temp and Speed Fan, and it makes sense. Right now I'm not sure what to believe, but I know I can't trust TAT, and Speedfan's CPU temp is whacky (it's above the core temps).

[skidrow]

My Rig: (8-Mar-2007)

Case: P180
PSU: Seasonic S12 600
Motherboard: Asus P5B-E
CPU: C2D 6300 @ 3220 Mhz (Lapped)
Memory: Team Xtreem PC2-6400 2GB (micron d9)
Heatsink: Thermalright Ultra-120 (Lapped)
Heatsink Fan: Stock Antec Tri-Cool
Intake Fan: Noctua NF-S12-1200
Exhaust Fans: 2 x Noctua NF-S12-1200
Internal Fan: Coolermaster SUF-S12
NorthBridge Cooler: Zalman NBF47
Fan Controller: Zalman ZM-MFC1
Graphics Card: EVGA 7600 GT
Graphics Card Cooler: Zalman VF900-Cu
Hard Drives: 2 x Seagate ST3320620AS (320 GB)

Overclock:

CPU Frequency: 460
Memory Voltage: 1.80 (Lowest "Manual" Setting)
CPU VCore: 1.375 (Very Low "Manual" Setting)
FSB: 1.200 (Lowest "Manual" Setting)

Temps:
CPU 51 (Orthos)
NB 31 (Orthos)

[supox]

50% of 1200RPM is 800?? Have you tried just running the Noctuas at 100%, not controlled by Speedfan, just to see what happens?

~1250RPM at full speed. PWM control might be to blame for the strange reading(s).