Wow. That made a difference!

Snooper

Senior member
Oct 10, 1999
465
1
76
I'm getting ready to transfer the guts of my system (P5P800, 3.6ghz 560 P4 - freebie from work, etc, etc, etc) to my new P180 case (it should be here on Tuesday). One of the reasons I ordered it in the first place is that I have been having a hard time keeping that 3.6ghz cpu cool. It has been running right at 71C-72C during 100% CPU load! The last straw was when I ran ThrottleWatch.

And I DID get to watch throttling! During extended 100% session (running a session of Einstein@home and SETI@home pegs it every time), it would bounce between throttling 12% to 50% and even occasionally to 66%! That DID explain why the temperature just went up to 71-72C and stayed there, even when my room was warmer...

Anyway, I went ahead and ordered some ArticSilver 5 when I ordered the case (from a different vendor) because I wanted to redo it with a better thermal paste. The stuff I used before was a "good" to "very good" paste in it's day (about 5 years ago. This stuff lasts forever it seems... And no, I don't remember the name nor do I know where the tube is at this time to look it up!), but I wanted to try the current champ.

And I am VERY glad I did! I just finished pulling the heatsink (a Zalman 7700) off and cleaned up the sink and the CPU before reinstalling it. It's been running at 100% load for the last 20 minutes and the temperature has settled down at 69C with absolutely ZERO throttling going on. Nice. This stuff is supposed to get better (to the point of serveral degrees C after many heat cycles they say) with time, so I will keep my eye on it and see what it does.

Hopefully, that P180 case will provide better airflow for the CPU in general so the CPU temps will go down even more. Of course, if that doesn't do it, I bet the temps WILL come down when I hit it with the water cooling...
 

Snooper

Senior member
Oct 10, 1999
465
1
76
I agree completely. The 90nm process caused all kinds of problems for Intel. It was not as good a process as it should have been (and definitely not compared to the previous 130nm process). The leakage current was always too high and they could not lower the voltage enough to compensate for the increase leakage with the thinner gate oxide. That's why the P4 hit the brick wall the way it did at 3.8 ghz. And that was pushing it. It's a pretty big problem when you have an architecure that is designed to run FAST, but your process won't allow it due to thermal problems...

From what I have heard, 65nm is a WHOLE different ball game. I can't wait to see what comes out when this process is fully ramped and refined. Thankfully, I can at least see the light at the end of a very slow tunnel!
 

Snooper

Senior member
Oct 10, 1999
465
1
76
The Pentium M was definitely at better match to the 90nm process. It was designed from the ground up to minimize power consumption and they worked in some pretty neat tricks to do that while keeping the instruction / clock up above the P4. I can't wait to see how the 65nm processors come out. Think multicore Pentium M running at P4 speeds. Or better.
 

Fullmetal Chocobo

Moderator<br>Distributed Computing
Moderator
May 13, 2003
13,704
7
81
Thanks for the info!! (Not neffing--just glad to see someone actually put out information instead of asking damn questions... :) )
Tas.
 

Duvie

Elite Member
Feb 5, 2001
16,215
0
71
so you run 1 instance of each of those DC apps??? 2 instances together???

Try running 1 instance and then run something that stresses the vid card and builds itsheat up....Then let us know if it holds...prolonged loading will eventually fill the case with warmer air...the newer may help to alleviate this as well....


I run DC X2 with nowhere near the highend level of that HSF and I see 52c with dual FH clients running for the last week...
 

Snooper

Senior member
Oct 10, 1999
465
1
76
Duvie,

Yes, a total of two instances (HT CPU). They absolutely peg the CPU utlization at 100% and leave it there until I stop them. This is the same "testing" I was doing early this week when I was getting all the insane throttling going on. I'm sure if I fired up a good 3D game and ran it for a while, my temps would creap up a bit. Not sure how much though as I SHOULD have pretty good case ventilation in my In-Win Q500. It's currently setup with a 120mm fan up front/bottom blowing in, the power supply blows out and I mounted a 120mm fan blow hole at the top of the case pretty much centered over the gap between the front of the powersupply and the back of the optical drives. There is always a nice, steady stream of warm air coming out of that blow hole. The MB temp is stable at 30C (it's a bit warm in my room right now...)

One thing to keep in mind is that this CPU was a freebie from work. Basically, it was sorted, packaged, and tested and while IT made the grade as far as reliablity is concerned, too many of the other die out of that lot (a group of 25 wafers that are always processed together) didn't, so all the die in the batch were not sold due to possible long term reliability concerns. I don't have a clue WHAT failed on the other die. But I DO know this thing always runs hot.

When I first power it up in the morning, the CPU will be at 50C right when the computer completes booting. After it idles for a bit, it will cool back down to 46C (actually, with the redo on the thermal paste, that has settled down to around 40C it appears). It's a hot momma no matter how you look at it.

If that P180 case doesn't improve the CPU temps much, I think I will go play in the water cooled world for a bit. I've always wanted to and now I have a good reason!

And I'm still not complaining! This was a $600 CPU when they gave it to me (about 2 months before this speed bin was available for sale...)! ;)
 

Snooper

Senior member
Oct 10, 1999
465
1
76
I've got things loaded up this morning (turned the system off last night) and my temps are locked in at 67C (vs. 69C last night). It IS a bit cooler in my room this morning, but the MB temps is only 1C lower than last night (30C vs. 31C.). The temps are nice, but the important thing is that it has not throttled one single bit since I applied AS5. That's what counts!

Now if that P180 will just GET here already!