It was sometime since I returned from vocation and I hadn’t use the gen64 system for a demanding task. Recently I had to do some simulations with ns2 and the performance was really poor! Meaning that some simple scenarios that demanded about 30 seconds (on the same system) required about 6-7minutes to complete. This utterly irritating behavior was most unexpected so actions should be taken to revert it.
One of the concerns was the fact that the system was recently updated new kernel was installed and some new dynamic libraries were in effect. The first issue attended was to update the whole simulating platform in order to be in accordance with the new system environment. Notice that I compile the ns2 statically in order to prevent such disasters. The results were the same. The computer was acting like a pentium 3 with 64Mbytes of RAM (do not try to simulate with ns2 in such systems its a bit paranoid)
Secondly I tried to alter the kernel preemptive-ness and latency parameters (
CONFIG_HZ_100=y
CONFIG_HZ=100
upto
CONFIG_HZ=300 ) but nothing changed.
After some serious thought and even though no signs of actual swapping taking place I decided to mess with the swap parameters of the system, in order to minimize such an effect if present. so
sysctl -w vm.swappiness=10 (as root to temporarily change the swap file/partition behavior - if someone needs to change such characteristics permanently he should alter the file /etc/sysctl.conf)
Notice that due to simulation requirrements I have 12Gbytes of swap file and using the swap intensively may render the system totally unstable. Of course that wasn’t the case.
Lastly I decided to check the BIOS and/or o/c the system a bit. And there I discovered the true nature of my problem. It wasn’t software related! It was a hardware problem. The amd64 was reaching temperatures of 60-63 degrees Celsius (approx 140 degrees in the Fahrenheit scale)!!! That totally lead to one conclusion. The CPU cooler (Stock AMD) was dead.
I opened the case and the cooler was working. So it was high time to operate the machine in order to attend the issue once and for all.
So check a photo from the interior of the machine. After applying some thermal paste or even changing the stock cooler the result was magnificent. An Amd64 machine as it ought to work.
We usually believe that problems exist only in software, not in hardware. When a failure happens to a piece of hardware the results are instantly noticeable (i.e. no booting at all). Yet as shown here some minor electromechanical problem can always damage a system. Without giving away any significant symptoms. In my case surfing experience was the usual one.
I am grateful that the system wasn’t affected by this overheating problem and I am still stressing the system to check if there is perm-damage to the CPU.
Leave a Comment