Hi,
I've been looking at the power requirements for HP blade servers. Using
their power calculator, I need 80 Amps, roughly 1.6 Amps per blade (2 x
dual core opterons in each blade). Anyone have any opionions on whether
this is a 'real world' figure? are HP being conservative? If so, how
conservative?
I'm asking because if I require more than 64 Amps in the data centre,
we'll have to provision for more power and cooling, which will be
expensive.

PGP SIGNED MESSAGE
Hash: SHA1

In article <1140629682.000026.189920@z14g2000cwz.googlegroups. com>,
monopole <paul.dixon@gmail.comwrote:
>I've been looking at the power requirements for HP blade servers. Using
>their power calculator, I need 80 Amps, roughly 1.6 Amps per blade (2 x
>dual core opterons in each blade). Anyone have any opionions on whether
>this is a 'real world' figure? are HP being conservative? If so, how
>conservative?

For two processors and one disk, 1.6A could be in the ballpark (at 120V,
anyway). I've been checking current draw on some homebrew servers lately,
and these are the numbers I've run across:

* Sempron 2800, five SATA disks, RAID controller, power supply w/ active PFC
startup: 2A (all disks spin up at once AFAICT)
idle: 1A
CPU at full load: 1.15A
* dual 242, 16 SATA disks & one PATA disk, RAID controller,
redundant power supply with three hot-swappable units
startup: 6.6A (even with staggered start on the RAID array!)
idle: 3.6A
(CPU at full load unknown)

These were built with ordinary desktop and server boards, 3.5" hard drives,
etc. The drives in each RAID array are 400GB Seagate SATA drives (four in
the first server, 16 in the second). I'd expect that a blade server might
draw less power, especially if it has only one or two disks (2.5" disks at
that) hanging off it.

_/_
/ v \ Scott Alfter (remove the obvious to send mail)
(IIGS( http://alfter.us/ Top-posting!
\_^_/ rm -rf /bin/laden >What's the most annoying thing on Usenet?

PGP SIGNATURE
Version: GnuPG v1.4.1 (GNU/Linux)

wCJFJ5IPKrc6YqJlqQyjGFA=
=8QAQ
PGP SIGNATURE

Keith wrote:

You don't have an HP sales droid who can figure this out for you? .chips
is really the wrong place to ask such things.

unfortunately the sales droid just repeats the figures from the
spreadsheet.

any suggestiion on a more relevant newsgroup to post in?

Cheers, Paul.

"monopole" <paul.dixon@gmail.comwrote in message
news:1140695803.167601.238660@
Keith wrote:
>>
>You don't have an HP sales droid who can figure this out for you?
>.chips
>is really the wrong place to ask such things.
>>
unfortunately the sales droid just repeats the figures from the
spreadsheet.

any suggestiion on a more relevant newsgroup to post in?

Cheers, Paul.

So HP gives you their answer, developed by their engineers, and you don't
like it. So you prowl newsgroups looking for someone to tell you it is
all right. Are you going to expect HP to do something if you install the
system and 64 amps isn't enough?

You could always buy it and upgrade the power later if it draws too much.
:-) And sweat a lot instead of AC.

I don't get what you are hoping to accomplish.

del cecchi

monopole <paul.dixon@gmail.comwrote in part:
I've been looking at the power requirements for HP blade
servers. Using their power calculator, I need 80 Amps, roughly
1.6 Amps per blade (2 x dual core opterons in each blade). Anyone
have any opionions on whether this is a 'real world' figure?

We usually talk Watts: That's only 192W for two dual cores,
power supply. mobo & disk. Frankly very light for CPUs
with TDP of 95-110W each. HP must be using some very high
efficiency PSUs to even think this might work. They must be
staging startup. And counting on some sort of load factor.

are HP being conservative? If so, how conservative?

Why would they be? Will it win them sales?

I'm asking because if I require more than 64 Amps in the
data centre, we'll have to provision for more power and
cooling, which will be expensive.

Do you think you're the only one in this position? HP marketroids
are never asked? I'd put in 100 Amps MINIMUM. They want 50 blades,
they gotta pay. Going cheap on power & cooling is just a recipe for
disaster: the fuse/main breaker blowing under high load! The worst
possible thing, at the worst possible time. Avoidable stupidity.
-- Robert

Thu, 23 Feb 2006 16:26:57 GMT, Robert Redelmeier
<redelm@ev1.net.invalidwrote:

>monopole <paul.dixon@gmail.comwrote in part:
>I've been looking at the power requirements for HP blade
>servers. Using their power calculator, I need 80 Amps, roughly
>1.6 Amps per blade (2 x dual core opterons in each blade). Anyone
>have any opionions on whether this is a 'real world' figure?
>
>We usually talk Watts: That's only 192W for two dual cores,
>power supply. mobo & disk. Frankly very light for CPUs
>with TDP of 95-110W each.

The HP BL35p blade server at least uses the 68W versions of the
dual-core chips.

HP must be using some very high
>efficiency PSUs to even think this might work. They must be
>staging startup. And counting on some sort of load factor.

My understanding is that it's pretty much a worst-case scenario power
draw. They do have some fairly extensive recommendations for setting
up power supplies for a variety of configurations. Some (most?) of
these power requirements are for 220V, so that's going to really
change how many watts we're talking about.

Either way, these blades definitely do pack a LT of processing power
into a small package with low power requirements. But they also
aren't going to be over-estimating the power requirements. The 80A
figure won't be average power consumption, but it'll probably be
pretty close to peak.

Tony Hill
hilla <underscore20 <atyahoo <dotca

I'm running 2x dual core opterons (270 HE) in my servers. Yes, the HE
provision would normally require less power than the blade. But then I
have a hard drive to feed so the power consumption should be near the
same.

I measured 1.5A while the system is working. ( 1.2 when idle).

Just as a comparison, my older servers that run dual Xeon 2.8 (and a
drive) require 1.9 A for the same work.

Tom S.

Well, it seems the HP power calculator was used incorrectly (too many
power supplies selected - better for an engineer to do this than a
sales droid!) so the current draw has dropped massively. It's now
prediciting a steady state running of around 0.5Amps per blade.

That's a heck of a lot less cooling. Thanks for your input guys.

Paul