Hi,

Gee, today I have decided to meassure power consumption of E280R with
dual X7009A (900Mhz, USIII cu) cpus.

Sun Specs of USIII speak about ~57Watt @1050Mhz, so I was kind of
really really, REALLY surprised to see that this beast eats 350-390
Watt when IDLE and when loaded.

Jesus, is a X7009A that power hungry or is that E280R chassis which
consumes so much?

Don't USIII throttle down when they have nothing to do?

Here are some other meassurments:

E420R with two X1195A: 230-250W
V120 / 650Mhz: 40-65W
Dual 1Ghz P3 Coppermine Intel SC5100 server: 75-90W
Dual 1.4Ghz P3 Tualatin Dell 2550: 100-130W
Dual 2.8Ghz Xeon Dell Precision 470: 295W

We are currently building a new Datacenter partly using older hardware
due to budget limitations and wanted to use E280R's for Postgres/LDAP
and SecurID backend but with our local energy prices this sounds like
a very, very expencive game in a long term.

How does V240 compare to E280 in terms of power hunger?

Kirill
-------------------------------------------------
When replying via E-Mail, please remove duplicate
"@" from the address.
-------------------------------------------------

Kirill Ponazdyr wrote:

> Hi,
>
> Gee, today I have decided to meassure power consumption of E280R with
> dual X7009A (900Mhz, USIII cu) cpus.
>
> Sun Specs of USIII speak about ~57Watt @1050Mhz, so I was kind of
> really really, REALLY surprised to see that this beast eats 350-390
> Watt when IDLE and when loaded.
>
> Jesus, is a X7009A that power hungry or is that E280R chassis which
> consumes so much?
>
> Don't USIII throttle down when they have nothing to do?
>
> Here are some other meassurments:
>
> E420R with two X1195A: 230-250W
> V120 / 650Mhz: 40-65W
> Dual 1Ghz P3 Coppermine Intel SC5100 server: 75-90W
> Dual 1.4Ghz P3 Tualatin Dell 2550: 100-130W
> Dual 2.8Ghz Xeon Dell Precision 470: 295W
>
> We are currently building a new Datacenter partly using older hardware
> due to budget limitations and wanted to use E280R's for Postgres/LDAP
> and SecurID backend but with our local energy prices this sounds like
> a very, very expencive game in a long term.
>
> How does V240 compare to E280 in terms of power hunger?
>

Don't know about "actual" but the plated value is around 400W which IME is
off the low end of modern systems.

P.

Kirill Ponazdyr <ng1@@codeangels.com> wrote:
> We are currently building a new Datacenter partly using older
> hardware due to budget limitations and wanted to use E280R's for
> Postgres/LDAP and SecurID backend but with our local energy prices
> this sounds like a very, very expencive game in a long term.

Looks like your data show that there is quite a bit more to the power
consumption of a system than just the power-consumption of its CPUs

As for the comparisons to other systems, you really need to add-in
number and type of I/O cards in each system, and whether or not there
were cables connected to those cards (eg a GbE NIC draws more power
when it is connected to a switch than when it is disconnected, an HBA
may be the same), how many and what type (RPM's etc) of discs, how
many and perhaps what size of DIMMs, etc etc etc.

> How does V240 compare to E280 in terms of power hunger?

Surprisingly enough, I had occasion to measure power consumption of a
V240 with 1.0 GHz CPUs in it:

Power Consumption:

Sometime after the initial measurements were made, the author
purchased (for this and other nefarious poproises) a P3 International
Model P4400 "Kill A Watt" power meter. This device purports to
measure a number of power parameters, in this instance, the author was
concerned with Watts consumed in various states:

State Watts
-------------------------------------------------------------------
Power-off, plugged-in, mgmt card alive 13
Power-on initial Memory Test 225
Sitting at "ok" prompt 222
Booted and idle 222
Running a loopback netperf TCP_STREAM test 232-234
Running a loopback single-byte netperf TCP_RR test 225
Running a netperf TCP_STREAM over 1 GbE port 227
Running 2 inbound TCP_MAERTS across 2 GbE ports 228-231
Running 2 netperf TCP_STREAM across 2 GbE ports 229-231
Running 4 netperf TCP_STREAM across 4 4 GbE ports 231-232

The system had 8 GB of RAM (8x 1GB DIMMS), two 1.0 GHz (1002 MHz)
UltraSPARC IIIi CPUs and a single 10K RPM SCSI disc.

[I didn't write it down at the time, but I suspect that all four GbE
ports (core, not-add-on) were connected in all the states above, but
it has been a while - another reason to be painfully verbose when
writing-up this sort of stuff...]

I also had occasion to do the same with a V20z (1.6 GHz):

The following are various and sundry values for the power consumption
of the V20z in various states/configurations. The power consumption
was measured with a P3 International, Model P4400 "Kill-a-Watt" power
meter.

System State/Config Power (W)
--------------------------------------------------------------------
Plugged-in, switch-off, Service Processor 25
booted, mains off. Single CPU in place. No
add-on NICs, PS/2 kbd and mouse, monitor.

Mains on with one CPU in place, booted and 154-155
root logged-in. 1GB RAM reported by dmesg.

Running loopback netperf TCP_STREAM test 163-165

Mains on with two CPUs in place, booted 202-203
and root command line logged-in. 2GB RAM
reported by dmesg. All hanging off of CPU 0

Running 2-CPU, loopback netperf TCP_STREAM 215-216

Running 2-CPU, loopback netperf TCP_RR 219-220

(The DIMMS were 512MB)

rick jones
--
oxymoron n, commuter in a gas-guzzling luxury SUV with an American flag
these opinions are mine, all mine; HP might not want them anyway...
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...