justaguywithacomp

Distinguished
Aug 9, 2008
5
0
18,510
I just finished my wc system and after checking the temps I am a little unsure (q9450/forumla2/dtek block/2x120 external mount radiator). The first image is the various temps I get at stock, I flashed my bios and the generic cpu temp dropped by exactly 10 degrees so that obviously is incorrect (saying < ambient at idle where before it made sense) everything else remained the same.
temps.jpg

This is under full load for about 10 minutes (get the same temps after an hour or so, havent tested longer yet), I ran the sensor test prior to starting and left it open. It looks fairly obvious that core 2/3 are stuck and the temp difference between 0/1 remains almost constantly at 7 degrees.

Ocing it up to 3200 from 2.6 gives me the following screen shot, again under load for about 10 minutes (close to the same again after an hour)
temps2.jpg


Core 2/3 now register what I think are realistic temps for my setup and the 7 degree difference remains between 0/1.

I've remounted the block 3 times and get nearly identical results (<.5 diff) for both above situations.

My question is what is more likely, core0/1 are correct and for some odd reason core 0 is always 7 degrees warmer or that core 2/3 are accurate but are just not registering until reaching a certain temp?
 

justaguywithacomp

Distinguished
Aug 9, 2008
5
0
18,510
didnt even notice the voltage in the screen shot, im at 1.25 while over clocked currently so that must have been a program glitch.

Ambient temps are around 22 degrees.
 

justaguywithacomp

Distinguished
Aug 9, 2008
5
0
18,510
So I did another remount and this time I got 3 degrees lower under load, so about 47 40 35 35 unclocked. Double checked for bubbles as well by slowly rocking the case all around and nothing extra came out.

What is the location of the cores on this chip? Perhaps the layout and access is causing me to tighten one side down more, if I know the location of core 0 I can try and give it a little more emphasis.

Also I have another mounting bracket, I'm using the standard one with the back plate that comes with the block but am wondering if switching to the other will help any.

What temps should I be expecting on this chip water cooled on stock?
 
I got 6 degrees between the highest and lowest under modest load - SLACR Q6600 @ 3Ghz (333)

Your core 0 seems a bit high ...

Those stupid Intel push/click HDF's are a total pain in the A$$.

Remounted mine a few times too.

Base is well poilished tho.

Wouldn't like to remove the IHS to see if direct contact reduced the difference between the ocres though I think it might be part of the problem for a few.

Most of my dual core cpus rarely vary more that a couple of degrees from each other.

 

justaguywithacomp

Distinguished
Aug 9, 2008
5
0
18,510
I'll try another remount and give the core 0 a little extra and see if I get any difference.

Is there any opinion on cores 2/3? Are those temperatures likely correct while under load/oc'd (38c) ? If they are it does look like something is uneven. Hopefully not the chip.
 

Grimmy

Splendid
Feb 20, 2006
4,431
0
22,780
In my opinion, you would need to calibrate what the tjmax is.

I'd say just go with the CPU temp, which Speed Fan shows. That would be the over all temp on the IHS, which is the Tcase sensor. Intel does publish the thermal spec of all their CPU's which would be the Tcase max, in which you shouldn't exceed.

When you read what RealTemp says when you test the sensors, the core sensors are part of a process for thermal throttling and thermal shut down. So for Idle temps, it won't be totally reliable at all, even if you calibrate them (to figure out the tjmax) since those sensor tend to stick at lower temps.

Edit:

On a side note, I got kinda sick of looking at more then 1 temp in my task bar, since I knew they would be off some, even though I'm using a Q6600 (65nm sensors don't really stick much). I find the Tcase temp pretty reliable and simple, since all I need to remember to keep it below 71C.