When I first fired it up, it was around 300W. Over time, a few
minutes after Solaris is running, I've noticed it started dropping.
Now it's around 245W. I'm including the LCD monitor as well. I've
got a 1500VA UPS, and now I'm looking at 18m of runtime in the event
of a power outage.
What I wonder is, how might this compare to the energy usage of a
recent desktop PC, like maybe a Core 2 Duo desktop, NVidia graphics
card? That other desktop I have running is a dual Pentium IV 3.0Ghz w/
Nvideo PCI-E graphics. It uses around 118W-138W. The LCD monitor is
using around 23W.
* hackr_d@yahoo.com:
> When I first fired it up, it was around 300W. Over time, a few
> minutes after Solaris is running, I've noticed it started dropping.
Understandable, as the fans don't run full speed any more which is what
they do during the power on diagnostics.
> Now it's around 245W.
245W doing what? Idle? What services are running, what programs etc.
> What I wonder is, how might this compare to the energy usage of a
> recent desktop PC, like maybe a Core 2 Duo desktop, NVidia graphics
> card?
Again, a Core 2 Duo PC doing what? What type of Core 2 Duo, what type of
Nvidia gfx card? The variation between the specific components regarding
power consumption is quite huge. An average C2D-based desktop PC with
midrange gfx card easily needs less than 100W in idle and less than 245W
under full load.
> That other desktop I have running is a dual Pentium IV 3.0Ghz
Definitely not as there is no dual Pentium 4 (P4 can't run in
multiprocessor systems). You either have a P4 with Hyperthreading (shows
two logical CPUs to the OS but still is single core/single processor), a
Pentium D which is dual core but not dual CPU, or a XEON (which can be
all of them, Hyperthreading, dual core and running in dual processor
config).
> w/
> Nvideo PCI-E graphics.
Which model of Nvidia gfx card?
> It uses around 118W-138W.
Again, doing what? Without being more specific about how and under which
circumstances you gathered your Watt numbers they are just meaningless.
BTW: a lot of cheap household power meters don't work well with
Switching Power Supplies connected to them and show false results.
On Aug 21, 5:38 pm, Benjamin Gawert <bgaw...@gmx.de> wrote:
> * hack...@yahoo.com:
>
> > When I first fired it up, it was around 300W. Over time, a few
> > minutes after Solaris is running, I've noticed it started dropping.
>
> Understandable, as the fans don't run full speed any more which is what
> they do during the power on diagnostics.
>
> > Now it's around 245W.
>
> 245W doing what? Idle? What services are running, what programs etc.
It stays steady at 245W. The LCD monitor is using 23W, so it's
actually using 223W. I'm using the "meter" built into my battery
backup. Maybe it isn't reliable?
> > What I wonder is, how might this compare to the energy usage of a
> > recent desktop PC, like maybe a Core 2 Duo desktop, NVidia graphics
> > card?
>
> Again, a Core 2 Duo PC doing what? What type of Core 2 Duo, what type of
> Nvidia gfx card? The variation between the specific components regarding
> power consumption is quite huge. An average C2D-based desktop PC with
> midrange gfx card easily needs less than 100W in idle and less than 245W
> under full load.
Just trying to find a ballpark figure for these, as I was concerned
about my Blade's power consumption. My DV8000 laptop uses around 60W
idle, and can go as high as 160W watching a video fullscreen.
Charging the battery uses as much as 210W when it's close to fully
drained.
>
> > That other desktop I have running is a dual Pentium IV 3.0Ghz
>
> Definitely not as there is no dual Pentium 4 (P4 can't run in
> multiprocessor systems). You either have a P4 with Hyperthreading (shows
> two logical CPUs to the OS but still is single core/single processor), a
> Pentium D which is dual core but not dual CPU, or a XEON (which can be
> all of them, Hyperthreading, dual core and running in dual processor
> config).
It must have hyperthreading then. I only saw one CPU, and task
manager shows 2 CPUs.
>
> > w/
> > Nvideo PCI-E graphics.
>
> Which model of Nvidia gfx card?
>
GeForce 7100 GS.
> > It uses around 118W-138W.
>
> Again, doing what? Without being more specific about how and under which
> circumstances you gathered your Watt numbers they are just meaningless.
> BTW: a lot of cheap household power meters don't work well with
> Switching Power Supplies connected to them and show false results.
This is an APC 1500VA UPS, which has a power meter function built-in.
Maybe it isn't accurate, as my Blade's power consumption seems to
"drop" over time, idle. The services were sshd, no other services.
Java Desktop, I'm sure, uses up CPU and hence more power. An
interesting experiment (under Linux, pretty sure Solaris doesn't have
this) would be to try the frequency throttling to see if that helps.
And the US III+ can do it. The module is us3_cpufreq.
Mostly I am talking about idle power consumption. The Blade 2000
seems to use a good bit idle.
>
> This is an APC 1500VA UPS, which has a power meter function built-in.
> Maybe it isn't accurate, as my Blade's power consumption seems to
> "drop" over time, idle. The services were sshd, no other services.
> Java Desktop, I'm sure, uses up CPU and hence more power. An
> interesting experiment (under Linux, pretty sure Solaris doesn't have
> this) would be to try the frequency throttling to see if that helps.
> And the US III+ can do it. The module is us3_cpufreq.
>
> Mostly I am talking about idle power consumption. The Blade 2000
> seems to use a good bit idle.
>
> DMM
Don't expect the ups power meter to be accurate. Better to buy one of
the domestic plug in style power meters and use it at the output side of
the ups. The better ones will allow selection of va as well as watts on
the display.
The initial variation in power could be the ups charging the internal
batteries. Even if nominally fully charged, there's always a certain
amount of top up charge to replace self discharge, from when the unit
was off. 50 watts difference sounds too much for the fans alone, unless
you have loads of them in the box.
The other thing is that if you have full height drives, replace them
with hh when you can, as they have significantly lower consumption...
> The other thing is that if you have full height drives, replace them
> with hh when you can, as they have significantly lower consumption...
Full height drives don't fit into the SB1000. Come think of it, I
don't think anyone has made full height drives for *years*. The only
modern 3.5" HDDs I know of are half height (i.e., 1.6") and 1" high.
> When I first fired it up, it was around 300W. Over time, a few
> minutes after Solaris is running, I've noticed it started dropping.
> Now it's around 245W. I'm including the LCD monitor as well. I've
> got a 1500VA UPS, and now I'm looking at 18m of runtime in the event
> of a power outage.
>
> Output of uname -a:
>
> SunOS sunblade 5.11 snv_94 sun4u sparc SUNW,Sun-Blade-1000
>
> What I wonder is, how might this compare to the energy usage of a
> recent desktop PC, like maybe a Core 2 Duo desktop, NVidia graphics
> card? That other desktop I have running is a dual Pentium IV 3.0Ghz w/
> Nvideo PCI-E graphics. It uses around 118W-138W. The LCD monitor is
> using around 23W.
That power consumption is normal. My Blade 1000 (2x900MHz) draws about
280W on idle without display (the 130nm-CPUs are power hogs). That's why
I'm turning to Solaris on x86. The machine (AMD 4850e, 4GB, 2 disks) is
already here, power consumption is less than 80W.
On Aug 22, 3:45 pm, ufo <d...@spam.me> wrote:
> > When I first fired it up, it was around 300W. Over time, a few
> > minutes after Solaris is running, I've noticed it started dropping.
> > Now it's around 245W. I'm including the LCD monitor as well. I've
> > got a 1500VA UPS, and now I'm looking at 18m of runtime in the event
> > of a power outage.
>
> > Output of uname -a:
>
> > SunOS sunblade 5.11 snv_94 sun4u sparc SUNW,Sun-Blade-1000
>
> > What I wonder is, how might this compare to the energy usage of a
> > recent desktop PC, like maybe a Core 2 Duo desktop, NVidia graphics
> > card? That other desktop I have running is a dual Pentium IV 3.0Ghz w/
> > Nvideo PCI-E graphics. It uses around 118W-138W. The LCD monitor is
> > using around 23W.
>
> That power consumption is normal. My Blade 1000 (2x900MHz) draws about
> 280W on idle without display (the 130nm-CPUs are power hogs). That's why
> I'm turning to Solaris on x86. The machine (AMD 4850e, 4GB, 2 disks) is
> already here, power consumption is less than 80W.
Yeah, I thougt about throwing together a nice low-cost desktop
machine, and installing Open Solaris on it. I was looking at an EVGA
uATX mobo with nVidia graphics onboard. Maybe the 1.2GHz US IIICu's
are a little more efficient? Anyways, that's going to be a real
challenge for Sun: make 'em fast, scalable, but also very energy
efficient as well. People care more about Green these days than they
did in the past. Else, the masses are going to be moving the the Next
Best and Cheapest thing, which is commodity (Intel or AMD/ATI)
motherboards.
Note 1066/800 MHz front side bus. It's probably dated, but will still
give strong performance. The Sun Blade 2000 has a "System clock
frequency" of 150 MHZ. Now I know SPARCs may not have front side bus
like PC, but they must be similar. I'm not familiar with SPARC arch
at all. I know it must have something more efficient than the PC's
FSB.
On 2008-08-22, Rich Teer <rich.teer@rite-group.com> wrote:
> On Fri, 22 Aug 2008, ChrisQ wrote:
>
>> The other thing is that if you have full height drives, replace them
>> with hh when you can, as they have significantly lower consumption...
>
> Full height drives don't fit into the SB1000. Come think of it, I
> don't think anyone has made full height drives for *years*. The only
> modern 3.5" HDDs I know of are half height (i.e., 1.6") and 1" high.
Well ... given that I have never seen a 3.5" drive which would
match the height of a full-height 5.25" drive, I think that it is
reasonable to call the 1.6" drives "full height" in the world of 3.5"
drives, and 1" drives half height.
Anyway -- what he meant was to switch from 1.6" drives to 1"
drives, which does significantly reduce the power consumption. Looking
at a pair of 36 GB FC drives, one 1.6" and one 1", I find:
height 12V 5V Total Cap Model # Watts/
inches Current Current Watts GB Seagate Terabyte
================================================== ============================
1.6" 1.5A 1.5A 25.5W 36 ST136403FC 725.33
1" 1.1A 1.4A 20.2W 36 ST336704FC 574.57
1" 1.1A 1.0A 18.2W 73 ST373405FCV 225.29
1" 0.9A 1.4A 17.8W 156 ST3146807FC 116.84
So -- the first saving is to replace 1.6" drives with 1" drives.
The second (keeping the same capacity) is to replace two 36GB drives
with one 73GB drive. If you need more storage, replace the one 73GB
drive with a 146 GB drive.
Note that I also have a 12V only 146 GB drive (or is it 5V
only?), but it is in a running system (this one) so I can't pull it out
to check without shutting down the system. IIRC it does draw fewer
Watts than the dual voltage version.
So -- yes, you can save power by upgrading the disk drives. I
don't know what the 500 GB drives draw, since I don't have one yet. But
I am running three trays of 7 73GB 1" drives each (just upgraded from
36GB drives -- mostly 1" but a few 1.6"). And thanks to zfs, I was able
to replace them without having to shut down and copy data from one array
to another. I just replaced one drive at a time, and when they all were
73 GB drives, the capacity of the zfs pool jumped to represent the new
drives. (These are arrays of five drives, plus two hot spares shared
between two arrays, and two drives (one per tray) just sitting there for
experimentation at need.
The 73 GB drives came from a lucky purchase at a hamfest -- 19
working drives and one dead one for $165.00. :-)
Enjoy,
DoN.
--
Email: <dnichols@d-and-d.com> | Voice (all times): (703) 938-4564
(too) near Washington D.C. | http://www.d-and-d.com/dnichols/DoN.html
--- Black Holes are where God is dividing by zero ---
Sun Blade 2000 energy consumption 
Linear Mode
