OK, this is a topic that pops up now and again in various guises, but
I've been working on upgrading my ageing server and reinstating the two
that have been taken off line due to hardware failure (one of which was
identical the other even older!). As a result I did some testing of
power usage between the old and potential new machines and got quite a
big shock, so I was wondering whether anyone else either had some
comparative figures or knew of a site that had some - or maybe could
point out a noddy mistake I've made that indicates I'm not as badly off
as I think I am!!
So for my systems I've put together the following:
My existing box is based on an old Celeron 533 with 512M RAM and twin
250G Seagate HDs setup with a RAID mirror. When I put my power monitor
on it it reported a peak of around 60W and settled back to around 47W as
a typical operating figure. A quick estimate of cost gave me a ball park
running cost of around the £40 per year mark give or take.
So far so good. I then proceeded to test my new server and decided I was
thankful that I hadn't paid for it!!
The new box is based on a nice Tyan motherboard (with serial console
access to the BIOS) with twin Athlon MP 2000+ processors and 1G RAM. For
testing there was only a single 40G Seagate HD in. Now when I put the
power monitor on that it peaked at around 180W and settled back down to
around the 172W mark give or take. At which point I picked myself up off
the floor and did my calculations. This came it at around the £140 per
year mark for electricity usage - eeek!!
Now I was expecting a jump, but that was a phenomenal jump. I decided
that, since the case had 4 80cm fans, plus the PSU fan and twin CPU fans
I try disconnecting a few to see what impact it had. Not a lot sadly. If
I removed all 4 of the 80cm case fans it only dropped back by around 8W.
I checked the power usage of the CPUs and these appear to measure up at
around the 52W to 58W mark, which still leaves a significant chunk,
although the Celeron looks to be using a mere 28W by comparison.
That doesn't add up well with my plans to reinstate the two other
servers, as it would head towards an yearly electricity bill of around
£450 for just those!
A quick test on some newer hardware (my 'top of the range' desktop) gave
me a figure of 95W, which sounds better as that only doubles the annual
running costs. Not that I really mean 'top of the range', just the best
I have here - it's an Athlon 2500+ with 1G RAM and a 160G HD.
So no it is time to start calculating the pros and cons of replacing my
'new' servers with lower power ones (or patching up the old ones with
parts from my bit boxes!). That causes problems with the fact that the
HD caddies are IDE ones, so won't support SATA drives, and to replace
them means scrapping my rack mount cases and buying new ones :( I'll
also, out of interest, be pulling these new servers to pieces for more
power tests (i.e. how much improvement does using a different PSU make -
not that this is a practical 'upgrade' since they aren't the standard
form factor being 2U rack cases).
Anyhoo, just wondered whether anyone else had done any similar
measurements. My best case scenario is that I've made a noddy error in
my calculations, so I'm more than happy for somebody to call me a Muppet
:)
--
Paul Tansom | Aptanet Ltd. |
http://www.aptanet.com/ | 023 9238 0001
======================================================================
Registered in England | Company No: 4905028 | Registered Office:
Crawford House, Hambledon Road, Denmead, Waterlooville, Hants, PO7 6NU
