Re: Power measurements

["Phil Endecott" <phil_fvnqz_endecott@chezphil.org>]

Rafa³ Bilski wrote:
>>> % dmesg | grep eps 
>> 
>> Thanks, yes, I found this after I posted my last message.  Apparently it
>> should use 860 mV at 1.2 GHz and 844 mV at 400 MHz.
>> 
> Most C7-Eden CPU's have same voltage on 
> high and low frequency. Looks like Your's has different voltages. If 
> You don't see 844mV at 400MHz then probably VID pins aren't connected 
> to VRM on motherboard.
> Mode	MHz		Voltage (mV)	Power (W)
> P0	1200		860		7
> P1	400		844		3

May I ask where these numbers come from?

>> My aim is to make this machine as
>> power-efficient as possible, which I've achieved mainly by working on
>> the PSU (the first one I had was < 40% efficient, now I'm up to about
>> 70%), and by using a solid-state disk.  It's getting increasingly
>> difficult to find things to improve!
> Newer PSU's should have 80%-85% efficiency. But 70% isn't that bad.

Don't believe that until you've measured it.
The numbers that they advertise are always in the best conditions, 
which is normally maximum load.  If you operate it in different 
conditions the performance will be worse.  For example, the power 
supply that is giving me 70% efficiency advertises itself as having 
>80% efficiciency.

> You should probably poke some registers in northbridge. I don't 
> know much about newer VIA chipsets, but older (CLE266) had many bits 
> which could improve power savings - dynamic clock stop to different 
> chipset components, ACPI C3 can put memory to sleep and so on. 
> Unfortunatly these bits are disabled by default and are never touched 
> by BIOS.

Is there any documentation available for any of this?
Hmmm, I really should learn something about ACPI.

> Btw. Can You benchmark Your's C7 for me? I would like to compare it to 
> my P4-M which is eating much more power. I'm not sure how much exacly 
> because I had to change FSB frequency, but something about 25W.
> Maybe "nbench"?

Here is nbench for my Jetway J7F2 with VIA C7 @ 1.2G with the 
'ondemand' governor:

BYTEmark* Native Mode Benchmark ver. 2 (10/95)
Index-split by Andrew D. Balsa (11/97)
Linux/Unix* port by Uwe F. Mayer (12/96,11/97)

TEST                : Iterations/sec.  : Old Index   : New Index
                     :                  : Pentium 90* : AMD K6/233*
--------------------:------------------:-------------:------------
NUMERIC SORT        :          354.96  :       9.10  :       2.99
STRING SORT         :          39.736  :      17.76  :       2.75
BITFIELD            :      1.1172e+08  :      19.16  :       4.00
FP EMULATION        :          46.664  :      22.39  :       5.17
FOURIER             :          2729.8  :       3.10  :       1.74
ASSIGNMENT          :          8.4321  :      32.09  :       8.32
IDEA                :          1042.5  :      15.94  :       4.73
HUFFMAN             :           525.7  :      14.58  :       4.66
NEURAL NET          :          4.7302  :       7.60  :       3.20
LU DECOMPOSITION    :           298.8  :      15.48  :      11.18
==========================ORIGINAL BYTEMARK RESULTS==========================
INTEGER INDEX       : 17.572
FLOATING-POINT INDEX: 7.148
Baseline (MSDOS*)   : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==============================LINUX DATA BELOW===============================
CPU                 : CentaurHauls VIA Esther processor 1200MHz 1200MHz
L2 Cache            : 128 KB
OS                  : Linux 2.6.21-1-686
C compiler          : gcc version 4.1.3 20070601 (prerelease) (Debian 4.1.2-12)
libc                : libc-2.5.so
MEMORY INDEX        : 4.507
INTEGER INDEX       : 4.295
FLOATING-POINT INDEX: 3.964
Baseline (LINUX)    : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.

The results for 'performance' are

INTEGER INDEX       : 17.530
FLOATING-POINT INDEX: 7.157

i.e. essentially the same.  For 'powersave' (400 MHz) the results are:

INTEGER INDEX       : 5.828
FLOATING-POINT INDEX: 2.379

which is almost exactly 1/3 of the speed.

For comparison a 1 GHz C3 gives:

INTEGER INDEX       : 13.785
FLOATING-POINT INDEX: 5.219

Comparing the C3 and the C7, the C7 gives 14.61 integer-index per GHz 
and the C3 gives 13.79 integer-index per GHz, so the C7 is 6% faster 
per GHz.  For floating point it's 14% faster per GHz.  (I think my C3 
probably has slower RAM, but I'm not sure about that.)


How does this compare with your Intel chip?


Regards,

Phil.