CPU Heat w/ Distributed.Net Client? 19
yamla asks: "I run a dual-Celeron machine (
ABIT BP6) and
primarily use Linux. I am concerned about heat and so I have added a second fan and used thermal paste. When I run the distributed.net client, the heat is CPU1:45C(113F), CPU2:49C(120F), SYS:55C(131F). Without this client running, my system is about ten degrees cooler at CPU1:34C(93F), CPU2:35C(95F), SYS:44C(111F). My question is simple. Should I stop running the distributed.net client or is this difference in heat nothing to be concerned about?" Wow! I've heard of hot software, but this might be too literal for comfort.
"Silica virus" (Score:1)
Probably just fine (Score:2)
If you have decent cooling (ie, good heatsink), it'll stay within spec. Can't remember the numbers off the top of my head, but those don't look too bad. If you can run rc5des for a long time without crashes, they forget about it. If you're not overclocking, you'll be fine. I run dual celerons in an extremly hot case (two big scsi drives in there, too), but things are just fine, since I have somewhere around 10 total fans (yes, it's loud). You don't need to go to that extreme.
If it doesn't break, don't fix it.
... and you're complaining? (Score:4)
On a comparative level, I used to have both a p133 and a p150, both which burnt your finger if you dared to touch them after a cpu-intensive job, but which never had any issues.
BP6 & D.net, and Linux (Score:2)
Also, I would like more information on what people are using for that hardware monitoring. lm_sensors seems to be the only option I've found, with several different GUI interfaces. Are there any other options?
A bit on the hot side... (Score:3)
About 55 C is a bit on the hot side. Your probably best off getting a better heat sink for both of your CPU's and maybe adding yet another fan. Personally I perfer keeping my CPU's down in the 30 to 40 C range to reduce termal stress. All my systems have good enough cooling that they run at no more then 40 C (room temp 25 C) except for my junk Cyrix (got it for free as it's unstable due to being badly overclocked once) that runs at about 45 C
Having your system 10 C cooler is not suprising at all. When you CPU is doing nothing Linux puts the CPU into a low-power mode. The second anything happens, such as a interupt, the CPU is instantly switched back on. In this low-power mode very little heat is produced, you can guess the results of that. :)
Personally I wouldn't worry for now and keep on using Distributed.net Intel processors have pretty good heat tolerence. However while some people say that since the maximum temperature is 80 C you shouldn't worry at all I disagree. If you're running your CPU at 80 C you are going to cause quite a lot of problems! That's like always running your car at the red line. It won't wreck your engine immediatly but over a period of time will cause a lot of problems.
Re:BP6 & D.net, and Linux (Score:1)
Same situation (Score:1)
As an intesting OT note, various OSes appear to put differing amounts of load on a CPU. In windows, with the d.net client running, my temp is around 40-42. In BeOS, with nothing running, the CPUs hover around 45-47 degrees. My best guess is that the Be idle loop includes some sort of instructions that make the CPU work more than even the d.net client does in Windows. I have no numbers for Linux, but that may be interesting to check out...
-Smitty
SYS 55C? (Score:2)
The one thing that looks odd to me if your temperature for the motherboard. I think that in my case the CPUs were always higher than the SYS temperature. Does your case have adequate ventilation and cooling? (ie, more fans than just the one in the power supply, which is half covered by ribbon cables)
Good luck,
-OT
Re:SYS 55C? (Score:2)
I am also thinking along the same lines. Try to get better case airflow.
On my main system I use a 90mm fan infront of my hard disks. It also blows out. I really need a larger case as I've maxed out HD slots. Initialy I used cable ties to mount the fan to the HD holding bracktes. I later switched to a piece of cardboard cut to fill in the opening around the fan. I use screws to hold the cardboard to the front of the fan. The fan is now held in place by the cardboard only. When cutting the cardboard I sized it a little tight, and just jammed it in place. It provides a better air seal this way. I placed a couple of images on my web server. Fan and cardboard mounted [visi.com], and fan and cardboard from back side [visi.com] showing harddisks behind.
Re:SYS 55C? (Score:2)
Hot Electrons (Score:1)
In the normal operation of transistors there is a great deal of rapid acceleration of electrons going on. Especially, as feature sizes go down, switching speeds go up and heat goes up.
Most electrons accelerate, as they should, from one end of the transistor (drain in, MOS) to the other (source, in MOS). But there are 'rogue' electrons that shoot off in other directions at high speed and hit the transistor itself (oxide layer below the gate, in MOS) and stay there. Eventually, as enough electrons do this, the transistor is damaged and will malfunction (in MOS, the oxide layer will begin conducting, which is bad).
Hotter tempartures accelerate this process.
This means that all modern ICs will eventually fail (feature sizes are sufficiently small, switching is fast enough and heat is high enough). From what I know, a pentium II is designed to last, on average, 10 years.
The moral of the story?
No matter what you do, your chip is going to fail eventually but you might be able to prolong its life by keeping it cooler.
Cool (Score:2)
My guess is that the increase in temperature caused by the dist.net client is the same increase as any other 100% CPU load software would cause. Modern x86 OS'es (which means Linux, NT etc. but NOT win9x) run a special idle instruction (the name of which I forgot) which powers down parts of the CPU when its idle. This is unlike apm/acpi 'suspend mode' etc. in that it is very fast, and wakeup only takes a few cycles. When the CPU load is 100%, the OS (obviously) has no time to run this instruction, and the CPU is fully powered all of the time.
The point is that the dist.net causes no more thermal stress than any other CPU-heavy app (quake anyone?
(/rant)
dufke
-
Re:Cool (Score:1)
First of all, isn't the 'sys' temperature you quote the temp in the case? If so, how can it be higher than the CPU temp... hardly possible as long as you are cooling with the case air and no peltiers. MB sensors are hardly the most accurate thermometers...
What if the case sensor is in the hottest place on the motherboard? And the hottest place is *not* the CPU's. Quite possible actually. For instance harddrives, especially fast high-RPM models, can generate a lot of heat. If your harddrive is close to the case sensor your are going to get a lot of heat there.
The CPU isn't the only thing getting hot on your motherboard too. There are all sorts of other chips and even a power regulator for the CPU. All of those can get very hot.
There is one more possibility too... Depending on how the CPU fans are positioned they could be blowing hot air straight onto the MB heat sensor and taking in cool air from above the motherboard.
BTW The idle instruction is HLT.
Temperature Sensors, Fans, & Noise (Score:1)
I was thinking of unplugging the front fan to reduce noise a bit. But I don't know how that will affect the temperature of this 650 mhz k7.
Tips/Tricks anyone?
Re:Temperature Sensors, Fans, & Noise (Score:1)
i do this with my duel celeron system I have 3 of the 6 fans on switches and only run them when im playing games burning cds ect. This allows me to leave it on overnight. The temperature when im surfing the web is only 1 or 2 degrees warmer with them off.
Talking about the bp6 iv verified that the motherboard temp sencer runs 10C over the auctual temperature and that the cpu sencers also run a bit high
Re:BP6 & D.net, and Linux (Score:1)
Re:BP6 & D.net, and Linux (Score:2)
Re:Temperature Sensors, Fans, & Noise (Score:1)
Re:SYS 55C? (Score:1)
My understanding (from reading bp6.com [bp6.com]) is that the SYS temperature on this motherboard tends to be higher than the CPU temps because they placed the temperature monitor right around the voltage regulators. Because these things generate heat, the SYS temperature doesn't reflect the temperature of the motherboard in general.
For what it is worth, I have experienced exactly zero lockups so far on my system that were not obviously driver related.
Note that I am running two different speed celerons in my system. Using Linux 2.2.14, you need to disable the TSC skew correction in the kernel. If by chance you are in this situation, email me [mailto] and I'll send you the patch I use.