[mythtv-users] CPU power (was Possible small HD frontend)
Michael T. Dean
mtdean at thirdcontact.com
Sat Mar 22 17:41:54 UTC 2008
On 03/22/2008 01:08 PM, Brian Foddy wrote:
> On my X2, it made a bit difference (reference my just recent post). I've
> also tried it on a dual 3GHz Xeon P4, and tho the cpu slowed and the machine
> was stable, the kill-a-watt showed no difference at all, but the machine
> was definitely slower. I hope this weekend to try it on a dual Opteron
> machine I have to see it it makes a difference.
Remember, guys, there's a lot more than just a CPU that's taking power,
so a "wall" measurement (as through the Kill-a-Watt) will reflect total
/system/ power usage.
It's been pretty well documented that though the Core 2 architecture is
more power efficient than the AMD K10 architecture, when comparing, i.e.
an Opteron vs a Xeon system, you'll likely see significant at-the-wall
power savings when using an Opteron. Since the Xeon systems use
FB-DIMM's and FB-DIMM's consume an /enormous/ amount of power at idle
compared to DDR2 (used by the Opterons). And though the FB-DIMM's power
usage "at load" (i.e. in heavy transaction processing) drops, it never
drops to the same power usage as the DDR2's until it reaches full load.
Therefore, when running a Xeon system, your CPU may be pulling less
power, but the difference is negligible compared to the extra power
pulled by the FB-DIMM's. And, to make a Xeon system with FB-DIMM's
"power efficient", you need to fully load /all/ the FB-DIMM's (this can
actually mean that having more memory than needed in such a system costs
a /lot/ of money in the long run).
See
http://arstechnica.com/news.ars/post/20070724-amd-vs-intel-power-efficiency-in-the-server-room.html
(and the links within).
I'm pretty sure I've seen other comparisons where mobo's/chipsets were
tested revealing some pretty astounding differences, though I don't feel
like digging them up (and don't remember the details well enough to give
the results). :)
Mike
More information about the mythtv-users
mailing list