Intel Stock > Ultra 120+Yate Loon...

FeRsHiZzLe · Post by **FeRsHiZzLe** » Fri Jun 29, 2007 10:28 pm

Great, about $65 to see my idle temps increase by about 10 Degrees Celcius, whats the deal?

raziell · Post by **raziell** » Fri Jun 29, 2007 11:52 pm

what is your setup?
did you check the connection of the hs?

FeRsHiZzLe · Post by **FeRsHiZzLe** » Sat Jun 30, 2007 12:31 am

E4300 on a Gigabyte S3. Temps were not as bad as now, but were still not good. I took off and did the water/glass flatness test. The water was all rough looking on the edges so I tried to lap/flatten it slightly with some wet 320 grit. The edges showed copper first which made me think I was doing it wrong, so i evened out scratches as best I could and re-installed. Now with a 200 slower mhz, my temps have risen to about 54 on idle, with stock they were like 50, and before "lapping" unsuccessfully, they were like 52. So I guess I exaggerated by saying 10 degrees haha.

continuum · Post by **continuum** » Sat Jun 30, 2007 1:51 pm

How are your load temps? Are you using enough paste?

FeRsHiZzLe · Post by **FeRsHiZzLe** » Sat Jun 30, 2007 6:48 pm

load temps peak at 71/70. I'll try reseating with more paste.

bonestonne · Post by **bonestonne** » Sat Jun 30, 2007 7:01 pm

paste or no paste, temps should be that different...also, you don't want too much of that stuff, it'll start to absorb the heat rather than transfer it to the heatsink.

are the CPU and heatsink making good contact with each other? if one surface is convex or concave more than the other, you'll have a gap where heat is not transmitted properly, and that's just really bad, it could account for your temps as well...check how level your CPU IHS is with a razor [carefully i might add]. then do the same with the base of the heatsink...theres only a major problem if both are flat...if either surface is quite far from being flat then you have your reason for the heat issue.

FeRsHiZzLe · Post by **FeRsHiZzLe** » Sun Jul 01, 2007 5:16 pm

Well now I'm not sure what to think. Until now, I've used TAT to see my temps, which are 49/50 idle now, after more sanding/re-seating. But in speedfan, my temps show 34-35 degrees on idle. Which one is right? Also TAT identifies my processor as a Pentium M...which it is not. Ideas?

FeRsHiZzLe · Post by **FeRsHiZzLe** » Sun Jul 01, 2007 7:27 pm

Also both the processor and heatsink seem very flat now, any ideas why my temps displayed in TAT are so high?

tehfire · Post by **tehfire** » Sun Jul 01, 2007 10:32 pm

FeRsHiZzLe wrote:Also both the processor and heatsink seem very flat now, any ideas why my temps displayed in TAT are so high?

IIRC the Conroe has multiple thermometers. It's possible that TAT is reading off different ones. It'd be great if you could open TAT and speedfan at the same time, but I'm not sure if they both can access the temps at the same time. If that doesn't work, have you used CPUBurn-in? http://users.bigpond.net.au/cpuburn/dow ... urn-in.zip virus free - I promise. Load two iterations of this (just open up the program twice) and run it while checking speedfan. That should give you a good idea of your temps.

cmthomson · Post by **cmthomson** » Tue Jul 03, 2007 5:38 pm

TAT uses the digital thermal sensor (DTS) which is inside the CPU chip itself. The DTS reports a negative number: the number of degrees Celsius below throttling. This is used by system vendors to turn up the fan when the headroom starts to get low.

Programs such as TAT (and Core Temp, and Everest, and others) use an assumed value for the throttling temperature, since Intel does not publish one (which makes sense since it would be lot-specific). TAT assumes 85C. I've seen values ranging from 82 to 97 with different programs.

If you run the latest SpeedFan (4.32), and look at the values for Core0 and Core1, you will see numbers that agree with Everest and TAT. The temperature it reports as "CPU temperature" could be almost anything, such as a sensor mounted on the motherboard, or the old-style thermal diode that the C2D still has to be compatible with P4 boards.

BTW, TAT reports all Core and Core2 CPUs to be Pentium M. I guess they never got around to updating the program. TAT was never meant to be public; it was released under license to system vendors to help them design their cooling systems, but someone posted a bootleg copy and now everybody has it, but Intel is not committed to supporting it.

FeRsHiZzLe · Post by **FeRsHiZzLe** » Thu Jul 05, 2007 10:54 pm

Thats what I thought should happen, but the core 0 and core 1 temps are 15-16 degrees cooler than the TAT ones. Under ~40% load on each core, SpeedFan reports 36 degrees, TAT 51. My ambient is about 80 degrees Fahrenheit, which leads me to believe the TAT ones are correct, but I just don't know.

Spanki · Post by **Spanki** » Fri Jul 06, 2007 1:54 pm

cmthomson got it mostly right...

For the E4300 cpu (along with other "Allendale" cpus and the L2 stepping E6300 and E6400 cpus):

- CoreTemp 0.95, TAT and maybe some others use Tjunction = 100C

- CoreTemp 0.94, SpeedFan 4.32, PC Wizard 2007 and most other programs use 85C

- cmthomson mentioned that some might use 82-97 (I'm not familiar with those, but don't doubt it)

...so, IF you have one of the cpus listed above, at least TAT and CoreTemp 0.95 will report temperatures 15C higher than those others that use 85C for the Tjunction value. If you are using a E6600, E6700 or early (B2 stepping) E6300 or E6400, then all those programs report the same (they use 85C).

Spanki · Post by **Spanki** » Fri Jul 06, 2007 2:27 pm

Here you go.. read this.

FeRsHiZzLe · Post by **FeRsHiZzLe** » Sat Jul 07, 2007 12:24 am

Well thank you for all the info. Now, if my E4300 is a Conroe, Stepping=2, Revision=L2, (according to CPU-Z) do you guys think I should go by the 85 or 100 TJunction temps?

Spanki · Post by **Spanki** » Sat Jul 07, 2007 1:19 am

It's up to you - whatever you're more comfortable with. Personally, I think 85C is the right number for desktop cpus (quads may be different though)... as mentioned in that link, TAT was originally created for testing mobile rigs.

On my system (L2 stepping E6400), using 85C also make the numbers make more sense... I can believe that my 'core' temp is 6-7C hotter than the 'cpu' (Tcase) temp being measured ~1mm from there (center point between the 2 cores), but I find it difficult to believe a 21-22C delta in that small a distance.

Using 85C also makes it easier to relate with other users, without having to explain that the temps you're quoting are high because they're based on a 100C TjMax (Tjunction).

In either case, the only real meaningful number is the DTS value (how far you are from throttling), and that's the same in both cases.

FeRsHiZzLe · Post by **FeRsHiZzLe** » Sat Jul 07, 2007 11:57 am

Well, thank you very much. Learned a lot from this thread, and am a lot more comfortable with my temps now.

Spanki · Post by **Spanki** » Sat Jul 07, 2007 12:41 pm

No problem - glad it made sense.

Spanki · Post by **Spanki** » Sat Jul 07, 2007 12:43 pm

Edit: deleted... ignore

cmthomson · Post by **cmthomson** » Sat Jul 07, 2007 7:26 pm

Spanki wrote:cmthomson got it mostly right...

The programs have been updated since the last time I looked, but the basic information remains accurate: the DTS is a negative number relative to the unpublished throttling temperature, which the various programs make various estimates of...