Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 24 Nov 2011 10:00:52 +0200
From:      Oles Hnatkevych <>
Subject:   8.2 broken CPU speed detection on IBM server
Message-ID:  <>

Next in thread | Raw E-Mail | Index | Archive | Help

We've got very strange problem with (at least) two identical IBM Servers

System Information
         Manufacturer: IBM
         Product Name: IBM System x3650 -[79792AG]-

FreeBSD 8.2-RELEASE is installed.

Depending on the phase of the moon when kernel boots with a change close 
to 70% the CPU speed is wrongly detected, affecting timing operation for 
the system.

This looks this way:

FreeBSD 8.2-RELEASE #0: Thu Sep 22 13:43:28 EEST 2011 i386
Timecounter "i8254" frequency 1193182 Hz quality 0
CPU: Intel(R) Xeon(R) CPU            5110  @ 1.60GHz (10748.77-MHz 
686-class CPU)
   Origin = "GenuineIntel"  Id = 0x6fb  Family = 6  Model = f  Stepping = 11
   AMD Features=0x20100000<NX,LM>
   AMD Features2=0x1<LAHF>
   TSC: P-state invariant

10.748 Ghz!!!

As a result we see 296 interrupts per seconds on CPU (six times less)

# vmstat -i
interrupt                          total       rate
irq15: ata1                           35          0
irq17: aac0                       525059         22
irq23: uhci0 uhci2+                37899          1
cpu0: timer                      7063047        296
irq256: bce0                     5275133        221
irq257: bce1                    43824849       1842
cpu3: timer                      7063046        296
cpu2: timer                      7063043        296
cpu1: timer                      7063045        296
Total                           77915156       3275

After that ping, top and other applications depending on timer slow down 
six times (i.e. pings once in 6 seconds).

And to round this up:

# time sleep 1
real    0m6.737s
user    0m0.000s
sys     0m0.000s

Playing with BIOS settings did not help at all. Even pressing 'restore 
defaults'. Updating BIOS did not help either. On the live system kernel 
has been recompiled, but that does not matter since on another test box 
the generic kernel is affected by the same problem.

What to do???? Try 7-CURRENT? Install Linux? ;)

Oles Hnatkevych
Skype: don_oles

Want to link to this message? Use this URL: <>