I have a few rack mount servers in an old chassis.

I had been running BOINC on one of the 1U machines (P4 HT) under FC1 and all seemed to be fine. The box probably runs a little hot but BOINC never reported any problems. I recently upgraded that machine to FC2 and now BOINC goes into "throttling mode."

example:

Message from syslogd@comstarb at Fri Dec 3 18:00:13 2004 ...
comstarb kernel: CPU1: Running in modulated clock mode
2x48_100157606 - PH 3 TS 210728 - 11/02/2063 04:00 - H:M:S=0524:48:24 AVG= 2.59 DLT=11.98
2x48_100157606 - PH 3 TS 210729 - 11/02/2063 04:30 - H:M:S=0524:48:25 AVG= 2.59 DLT= 1.00

Message from syslogd@comstarb at Fri Dec 3 18:00:18 2004 ...
comstarb kernel: CPU0: Temperature above threshold

Message from syslogd@comstarb at Fri Dec 3 18:00:18 2004 ...
comstarb kernel: CPU1: Temperature above threshold

Message from syslogd@comstarb at Fri Dec 3 18:00:18 2004 ...
comstarb kernel: CPU1: Running in modulated clock mode

Message from syslogd@comstarb at Fri Dec 3 18:00:18 2004 ...
comstarb kernel: CPU0: Running in modulated clock mode
2x48_100157606 - PH 3 TS 210730 - 11/02/2063 05:00 - H:M:S=0524:48:27 AVG= 2.59 DLT= 1.99
2x48_100157606 - PH 3 TS 210731 - 11/02/2063 05:30 - H:M:S=0524:48:28 AVG= 2.59 DLT= 1.00
2x48_100157606 - PH 3 TS 210732 - 11/02/2063 06:00 - H:M:S=0524:48:29 AVG= 2.59 DLT= 1.00
2x48_100157606 - PH 3 TS 210733 - 11/02/2063 06:30 - H:M:S=0524:48:31 AVG= 2.59 DLT= 2.00

Message from syslogd@comstarb at Fri Dec 3 18:00:23 2004 ...
comstarb kernel: CPU1: Temperature above threshold

Message from syslogd@comstarb at Fri Dec 3 18:00:23 2004 ...
comstarb kernel: CPU0: Temperature above threshold

Message from syslogd@comstarb at Fri Dec 3 18:00:23 2004 ...
comstarb kernel: CPU0: Running in modulated clock mode

Message from syslogd@comstarb at Fri Dec 3 18:00:23 2004 ...
comstarb kernel: CPU1: Running in modulated clock mode
2004-12-03 18:00:28 [---] Received signal 2
2004-12-03 18:00:29 [---] Exit requested by user

I had come across an update for the mobo BIOS, that mentioned incorrect temperatures were reported. That has not cleared up the issue. When I run this box outside the chassis, it is OK. Apparently the issue is it is too hot for FC2 in the chassis. But since this was not the case in FC1 it still has me wondering.

Has anyone had a similar problem with FC2, incorrectly responding to high temperature with CPU under full load, or is it accurate and FC1 was not aware of the problem before?

Placing a fan in front of the case can help but the electric bills are starting to get out of my budget!

---

I guess your old kernel just wasn't aware of the throttling.
The kernel only reports the throttling, it's done by the cpu itself.

1U is to small for a P4 at full load, IMO.

If your rack is big enough you should try separating the seti box from the
others and take the lid off the 19" module to get more fresh air in.
This works wonders.

Regards Hans

P.S:

Don't forget to dust your box from time to time :o)

---

> I guess your old kernel just wasn't aware of the throttling.
> The kernel only reports the throttling, it's done by the cpu itself.
>
> 1U is to small for a P4 at full load, IMO.
>
> If your rack is big enough you should try separating the seti box from the
> others and take the lid off the 19" module to get more fresh air in.
> This works wonders.
>
> Regards Hans
>
> P.S:
>
> Don't forget to dust your box from time to time :o)
>
>
Thanks for the reply, Hans.

I realize I am trying to run things on the temperature edge here. I try to keep things as clean as posible.

I had similar situation with another machine (I have three that are hardware identical as can be) that has more airspace above the hole in the case above the CPU. A patch in FC2 seemed to clear up the same issue. Since the ambient temperature is cooler now in LA than in the summer I did not expect to see this problem occur now.

For the time being, I switched the HD's between the throttling box and one that runs Win 2K Adv.Server. There are no messages from the new Win box regarding throttling; the CPU utility reports about 60 degrees C. The Switched FC2 is running smoothly but I have temporarily cleared space above it. I'll see what happens when it the other stuff is in place and it has a little more airspace than it had before. Otherwise I will have to reorganise the rack.

What utilities are avialable to report temps (CPU & ambient) in FC2, if anyone knows?

---

> Thanks for the reply, Hans.
>
> I realize I am trying to run things on the temperature edge here.

I'm having good success ATM with _lowering_ the CPU VCore.
The machine I'm posting from (Prescott P4 3.2) runs at a VCore of 1.2V instead of the nominal 1.38V.
No problems so far, CPU temp dropped by 10°C. Use at own risk!

Only a few boards actually support setting a lower VCore.
(Gigabyte here. I recommend them)

> I try to keep things as clean as posible.
My comment was more or less tongue-in-cheek :o)

Regards Hans

---

> What utilities are avialable to report temps (CPU & ambient) in FC2, if
> anyone knows?

Don't know about FC2, but I use RedHat 9 and use lm_sensors to monitor the cpu and motherboard temps and the fan speeds. Check to see if lm_sensors comes with FC2. If not, you can find it at http://secure.netroedge.com/~lm78/

Charlie

---