/proc/cpuinfo reporting low BogoMIPS

[paulfxh]

I run Conky in all of the OSes (Raspbian, Gentoo, Arch) I use on my RPi to display a range of parameters including CPU frequency.
This latter I read from /proc/cpuinfo.

Since, the firmware rpi-update of 9/01/2013 and all subsequent updates, the number reported by BoGoMIPS (from /proc/cpuinfo) has reduced to exactly two-thirds of the true value.
In searching through this forum, I can find only a few oblique references (here and here) to this phenomenon but no suggestion of a solution.

In both links referenced above, the CPU frequency shown by BogoMIPS is again exactly two thirds of the true value.

In my case, the under-reporting is easily overcome by multiplying the BogoMIPS value by 1.5 in my Conky script.

However, I'm puzzled that this has not caused more concern in the RPi community.
Have I missed something?

[RaTTuS]

if you have underclocked your RPi then the bogomigs will be wrong ,
see /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq

[paulfxh]

Sorry, I forgot to mention that, in all OSes, I have OVERclocked the processor to 850 MHz.
BogoMIPS always indicated 845 MHz before 9/01/2013 but 564 MHz thereafter.

[efflandt]

How do you know that it was being properly calculated before? I have been using Linux for many years and I do not recall seeing bogomips directly match the same number as the frequency. I seem to remember my old 33DX386 (33 MHz) being about 26 bogomips. I don't have any arm computers to compare, but a quick look at regular PC's including the old Celeron below and newer PC's seem to show bogomips about double the frequency:

Code: Select all

efflandt@realhost:~> cat /proc/cpuinfo
processor	: 0
vendor_id	: GenuineIntel
cpu family	: 6
model		: 6
model name	: Celeron (Mendocino)
stepping	: 0
cpu MHz		: 333.060
cache size	: 128 KB
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 2
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca cmov pat pse36 mmx fxsr
bogomips	: 665.19

[paulfxh]

How do you know that it was being properly calculated before?

Thanks for your reply.
How do I know?
Well, I've been overclocking the cpu on my RPi, on several OSes, for the last six months and keep a close eye on the BogoMIPS displayed in /proc/cpuinfo (it's up in front of me all the time in Conky)
For all that time up till about two weeks ago, the BogoMIPS value displayed was invariably about 99.5% of the cpu_freq value I had input to /boot/config.txt
So as far as a mere human being can be certain of anything, I'm pretty sure what was being calculated before was not far off the mark.

quick look at regular PC's including the old Celeron below and newer PC's seem to show bogomips about double the frequency:

This is probably because the computers you've looked at have dual core processors which means twice the number of calculations per minute as a single core (which is what the RPi has) operating at the same speed.

[ascent]

I'm actually experiencing the same with an unoverclocked Pi, but the strange thing is that with two different Pi's, the values differ:

Pi #1

Code: Select all

[root@raspi /]# grep -i bogomips /proc/cpuinfo
BogoMIPS        : 464.48

Pi #2

Code: Select all

[root@raspi /]# grep -i bogomips /proc/cpuinfo
BogoMIPS        : 697.95

Both Pi's run Pidora 18 but the slower one has a lower firmware (fromJun 6 2013 18:33:28) where the faster one has a newer one (Jun 29 2013 15:05:17). After running rpi-update on the "slower" Pi, they now both have the same bogomips value. Perhaps this is useful for anyone.

Wouter