Need Assistance Cooling AMD 64 x2 5000+

[kenwood850]

I am at wits end. I have been trying to get an AMD Athlon(tm) 64 X2 5000+ CPU to run cooler on an ASUS M2NBP-VM-CSM motherboard using an Artic Cooling Freezer 64 Pro AC heat sink and fan with Artic Silver 5 thermal compound. The OS is the latest Slackware.

Under no load (99+% idle as reported by "top") "sensors" reports the following:

k8temp-pci-00c3
Adapter: PCI adapter
Core0 Temp: +43.0°C
Core0 Temp: +40.0°C
Core1 Temp: +49.0°C
Core1 Temp: +51.0°C

acpitz-virtual-0
Adapter: Virtual device
temp1: +40.0°C (crit = +75.0°C)


As soon as I put it a load of about 45% on both cores as reported by "top", it soon reaches 64°C where a script I am running shuts it down so the CPU won't be damaged.

The script check the temperatures every 10 seconds and the output looks like this. It would appear I applied the load around 21:35:22.


01-04-10, 21:34:02, +45.0°C, +50.0°C, +44.0°C, +54.0°C, +40.0°C
01-04-10, 21:34:12, +44.0°C, +50.0°C, +44.0°C, +52.0°C, +40.0°C
01-04-10, 21:34:22, +45.0°C, +50.0°C, +44.0°C, +52.0°C, +40.0°C
01-04-10, 21:34:32, +44.0°C, +50.0°C, +44.0°C, +51.0°C, +40.0°C
01-04-10, 21:34:42, +44.0°C, +50.0°C, +45.0°C, +52.0°C, +40.0°C
01-04-10, 21:34:52, +45.0°C, +50.0°C, +42.0°C, +54.0°C, +40.0°C
01-04-10, 21:35:02, +45.0°C, +49.0°C, +42.0°C, +54.0°C, +40.0°C
01-04-10, 21:35:12, +44.0°C, +49.0°C, +44.0°C, +52.0°C, +40.0°C
01-04-10, 21:35:22, +45.0°C, +49.0°C, +42.0°C, +53.0°C, +40.0°C
01-04-10, 21:35:33, +50.0°C, +56.0°C, +48.0°C, +60.0°C, +40.0°C
01-04-10, 21:35:43, +52.0°C, +58.0°C, +50.0°C, +63.0°C, +40.0°C
01-04-10, 21:35:54, +52.0°C, +56.0°C, +50.0°C, +60.0°C, +40.0°C
01-04-10, 21:36:04, +53.0°C, +58.0°C, +52.0°C, +62.0°C, +40.0°C
01-04-10, 21:36:15, +54.0°C, +60.0°C, +50.0°C, +63.0°C, +40.0°C
01-04-10, 21:36:25, +54.0°C, +59.0°C, +52.0°C, +62.0°C, +40.0°C
01-04-10, 21:36:37, +54.0°C, +60.0°C, +49.0°C, +63.0°C, +40.0°C
01-04-10, 21:37:24, +55.0°C, +60.0°C, +53.0°C, +64.0°C, +40.0°C
Shutdown

I have tried various amounts of AS5 from the rice size grain and half they specify to perhaps twice as much, cleaning the heat sink and CPU thoroughly with 91% isopropyl alchohol between trials, and the results are always about the same. I am not overclocking and I have taken all the defaults in the BIOS. The AS5 when I separate the CPU from the heat sink looks like it was making good contact in that the AS5 has a stippled appearance on both pieces. Coverage has been about 70% to 100% depending on how much AS5 I used.

The one thing that is puzzling is that when I go into the BIOS during a boot, the BIOS reports the motherboard temperature as 32°C and the CPU temperature as 32 or 33°C. In theory I would think the temperature "sensors" reports at idle would be the same as what the BIOS reports, but it isn't. Sensors is reporting about 10 to 15 degress more under what I suppose are the same conditions.

Based on all this I have several questions.

1. Why the discrepancy between what the BIOS reports as the no load temperature and what sensors reports?

2. Could sensors be in error?

3. If "sensors" is accurate, what the heck can I do differently to bring the temperature down?

Thanks.

 

[Conumdrum]

Try Realtemp. Never trust the bios or anything else. I'm Intel for now, but one of these is best for AMD. Please refer to the AMD CPU forum for the exact one.

RealTemp
CoreTemp
HW monitor

Benching software and such is very varied. I use these for each purpose:
These are pretty standard and used by many.
Monitoring the PC temps overall: HW monitor aka hardware monitor
CPUZ for CPU info
GPUZ for GPU info
CPU only: Real Temp
GPU only: ATI Tool, I have a Nivida GTX280, so it works on Nvidia

Loading/benching tools:
CPU loaders: Prime95 and OCCT
GPU Loaders: ATI Tool and the best one is Furmark, nothing pushes the GPU harder right now.
Benching for overall graphics/gaming performance is 3DMark06

 

[kenwood850]

Try Realtemp. Never trust the bios or anything else. I'm Intel for now, but one of these is best for AMD. Please refer to the AMD CPU forum for the exact one.

RealTemp
CoreTemp
HW monitor

Benching software and such is very varied. I use these for each purpose:
These are pretty standard and used by many.
Monitoring the PC temps overall: HW monitor aka hardware monitor
CPUZ for CPU info
GPUZ for GPU info
CPU only: Real Temp
GPU only: ATI Tool, I have a Nivida GTX280, so it works on Nvidia

Loading/benching tools:
CPU loaders: Prime95 and OCCT
GPU Loaders: ATI Tool and the best one is Furmark, nothing pushes the GPU harder right now.
Benching for overall graphics/gaming performance is 3DMark06[/quotemsg

Unless I am missing something, these all look to be Windows type programs. Slackware is a Linux flavor

 

[Conumdrum]

Ohh, LOL, your kinda out of luck. Make a dual boot then check your temps in windows.

I see you did mention slackware. I have no idea what your meaning except you said linux.

 

[kenwood850]

The fun never stops. I took off the Slackware on the boot partition and installed Windows XP for testing. I ran CoreTemp and temps seemed about the same as lm_sensors was reporting. I also did a little more research on the web and found this particular CPU might have had problems with reporting temperatures. As a result I loaded Prime 95 and ran the CPU stress test after setting up the Core Temp to shut it down if it reached 69. Well, after about a minute it reached 69 and shutdown.

I then tried to restart it and the power came on and the fans spun for about 3 seconds and then it shut down with no beep codes and nothing on the monitor. I waited 5 minutes and tried again with the same results. Okay I thought, I was a little to aggressive in my testing and I fried the CPU. Lesson learned. I then put the old CPU back in and the results were the same. When you start it, it run for about 3 seconds and then shuts down with no indication as to why. I tried clearing the CMOS by removing the battery and putting the jumper in the shorting position for 10 seconds, but that did not seem to help.

I am probably no longer in the right forum for this, but any idea what might have happened?

 

[kenwood850]

This morning I took out the PS and tested it to make it sure it was working fine and it was. I then tried it again on the motherboard with all the drives disconnected and lo and behold it worked. I then put the 5000+ back in and it worked as well. I then tried the 5000+ with all the drives connected and it sill worked. I have no idea what happened.

In any case, I still had the problem with the high temperatures. I then took the CPU out and checked the flatness with a razor blade and it was definitely concave. A check of the HS surface showed it was relatively flat but with deep machining grooves. I spent several hours with 800 grit paper on glass lapping both surfaces until they were as flat as I could make them and relatively smooth. When I put it back together there was a substantial improvement in the core 0 temperature but the core 1 temperature although improved still left a lot to be desired. I took it apart one more time and lapped it again, but could not improve on the flatness. The finish might have been improved a little but that was all. When I put it back together the temperatures were pretty much the same.

What I have now is

Core 0 at 0% load 42°C
Core 0 at 100% load 57°C

Core 1 at 0% load 48°C
Core 1 at 100% load 72 °C

At this point, unless someone else has some suggestions (and I welcome them), I think I will button it back up, put Slackware back as the OS, and run it as the server it was designed to be with SETI running in the background, and not worry about the temperatures. I have done all I can think of to do and have put more time into it then it probably deserves. If it blows up, I will at least have the good excuse to do the the MB and CPU upgrade I have been wanting to do but couldn't justify.