This is my first post on this forum, first build, first real overclock, so it's been a learning experience, but I've been very careful about crossing all the t's and dotting all the i's. It's still been a bit hairy, with some sound driver problems with the mobo and two marginal ram sticks that seemed like a hard drive problem but weren't.
Here's the hardware:
Intel E2160 Pentium Dual Core (M0 revision) (lapped to 2000 grit, AS5 TIM, lapped stock cooler)
XFX 680i LT mobo
4x1 GB OCZ Platinum Rev. 2 (PC6400)
400GB WD 7200rpm 16mb cache SATA3 HDD
LiteOn SATA 20x DVD DL burner drive
CoolerMaster eXtremepower 600W PSU
Thermaltake Armor w/250mm side panel fan.
ASUS 8800GT 512mb VGA
32 bit XP pro
Anyway, I've got everything pretty much sorted out now and I'm looking to crank this baby up, but my wife has made me swear that I won't buy another computer for 4 years, so it needs to last a while. To that end, I've been trying to reliably determine just what my CPU's significant temperatures are, so I can keep it nice and safe, but not leave too much clock speed on the table, so to speak.
I've read the Core 2 Duo Temp guide, and every Intel datasheet, addendum and design guide I could find on these processors, as well as numerous other sources, overclockers.com, hardware secrets, madshrimps, everything I could find, usually more than once.
I understand about the two sensors in the core and the sensor on the mobo and how they're different and all that stuff.
My question is about the calibration method in the guide, and how well it applies to the newer M0, G0 and so forth processor revisions.
The guide says to run the processor underclocked, at idle, and for 10 minutes, and by the end of those 10 minutes your Tcase should be within 1 deg C of ambient.
Well, mine isn't.
Now I realize that that is the purpose of the calibration, to correct for errors in the measurement, but this is all based on the assumption that a processor idling in such a condition is actually at Tcase = Ta+1.
How do you know?
After around 45 minutes of doing nothing but blow air on my processor at the exact frequencies, voltages, and settings specified in the guide, the temperatures were pretty locked in.
Ta was 22 deg C
Tcase was 29 deg C from SpeedFan
Tjunctions were 43 deg C, From CoreTemp 0.96, and SpeedFan w/ 15 deg offset, verified with Crystal CPUID
That means the Tcase to Tjunction delta was 14 deg C and the Tcase to Ta delta was 7 deg C.
So the first one was ballpark, but the second one was about 6 degrees high.
According to the guide I should just dial the Tcase offset into SpeedFan and be on my merry way, but since I had the case open, I decided to measure the temperature of the CPU heatsink directly.
Actual measured temperature of the heat sink and the air blowing off of it: 26.7 deg C, 4.7 deg over ambient.
Now the heatsink wouldn't be hotter than the CPU case, so Tcase must be somewhat greater than 1+ambient, just like SpeedFan said. Also since the Tjunction temp was parked at a verified 43 deg C, the resulting Tcase = Ta+1 to Tjunction delta would have been really big, 20 deg C.
But if the Tjunction temperatures are really idling at 43 deg C, that kinda sucks, cause according to your chart, under load this thing will almost climb out of the safe range at stock speeds, leaving very little room for overclocking.
I am of the opinion that whatever Intel did to these newer chips with the 73.3 degree Tcase max, they seem to believe that these chips will live happily at a significantly higher range of temperatures than they used to.
Evidence of this is in the coolers that come with two versions of the E2xxx processors.
The older 61.4 deg C spec E2xxx processors included a heatsink/fan with a degC/W rating of 0.33,
while the new 73.3 deg C spec processors include a heatsink/fan rated to 0.51 deg C/W, even though both processors have the same TDP of 65W. It seems to me that a 0.33deg C/W cooler would have a better shot at getting an idling processor down to one degree over ambient than a 0.51 deg C/W cooler, particularly since cooler efficiency isn't strictly linear, but dependent on the delta T between the heat source and Ta.
Regardless, has anybody else had this sort of thing happen? Are the high idle temps normal for this processor? I'm getting a better cooler, Scythe Mine Rev. B, but I'm still going to wonder if everything's okay if I can't get this nailed down. I suppose with the better cooler, I should be able to get Tcase down to Ta+1, so this might be moot, but I'd like to hear from people anyway.
I also did an even lower underclock, 1.2 GHz, 1.2V. Same results: Tcase 29, Tjunctions 42,43
To get an idea of my cooler capacity and the processor's power output, I ran a Prime 95 100% load test at stock speed, 1.8GHz and 1.184V and got Tjunctions of 57 deg C. This corresponds pretty well with Intel's claimed spec of 0.51 deg C/W for the cooler. The processor should have been using around 58.3W, and with ambient at 24deg C, the cooler was operating at about 0.57 deg C/W. So my processor and cooler matched the specs pretty well.
Sorry about the length of the post, but I wanted to be as comprehensive as possible in my description of my question, since I've noticed that with technology in general, the devil is in the details.
Also, I'd like to compliment Mr.CompuTronix on an excellent and informative guide, in spite of my small quibble.
And finally, please refrain from lecturing me on how pointless 4x1GB of RAM is on a 32 bit system that I'd like to overclock. I'm planning on moving to 64-bit Vista eventually and it was $35/pair, so I figured what the heck, it's still an extra 0.75GB available.
This is just my opinion so take it for what it's worth... I think people are way too worried about temperatures. If you read the sticky on temps then you know what your processor's thermal limits are and what is safe for 24/7 operation. Tjuntion is the senser within the cores themselves and in my opinion the only temp that really matters. Tcase is the senser at the heat spreader and can be off quite a bit. Keep voltages lower than max value allowed by intel (for warrenty reasons), and watch your core temps and your set. Your tjunction of 57c in prime 95 gives you at least 13c headroom, provided you ran the program for aleast 2 hours. 70c is safe and wont damage your processor. That being said, if you bench to 70c tjunction, your cpu will run way cooler during normal operation(even video encoding), and should have no problem at all lasting 4 years and beyond. Windows 32 bit can allocate
3gb+ ram depending on you setup and configuration, so it isn't a waste of money.
I know the stock speeds are ok temp-wise. However I, like a lot of people on this forum, picked this processor based on the Tom's Hardware article "$89 Pentium Dual Core that Runs at 3.2 GHz", so I was hoping to do just that, overclock it to around 3.2 GHz.
I can run the thing at 3.0 GHz (333 fsb, 1.3V) totally stable, but the Tjunction climbs to about 74 deg C when running Prime 95 which is getting into the warm zone, so going to 3.2 GHz on the stock cooler isn't going to happen. That is why I'm getting the new cooler.
I guess the real crux of my question about the temp guide was the statement that running the processor underclocked at idle will always give you a Tcase within 1 degree of ambient.
It seems like the guide is suggesting that once you know the Tcase by doing the previous procedure, you should correct the Tjunction temperatures to give an appropriate Tcase to Tjunction delta for your processor.
This doesn't seem unreasonable, since the Tjunction calibration is set at the factory, and could be in error just like anything else. And since you can't easily verify the temperature of the cores directly, establishing a base point off of ambient using this procedure would seem like the way to go. The thing is, if I had done that with my processor, the Tcase idle temp would have dropped 6 deg to 23 deg C and the Tjunction temps would have dropped 10 deg (+/-3) to 33deg C.
While that would be great, I'm not entirely convinced it's correct, and a ten degree difference in reported temperature is significant, particularly if the error grows with increasing load and overclock.
Perhaps the guide should mention a minumum CPU cooling capacity required to give a Tcase sufficiently close to ambient. The C1E halt state power consumption of the E2160 is spec'd at 8W. With a 0.33 degC/W stock cooler, this gives a rise above ambient of 2.6 degrees. With a 0.57 degC/W cooler (like mine) the theoretical rise above ambient is 4.6 degrees. Additionally, at such low temperatures the CPU cooler would be even less effective than usual due to the small temperature difference between the heat source and the surrounding air, so the error could be worse than that, as my temperature measurements seem to suggest.