EPIA vs Sempron

  • Thread starter Thread starter Joris Dobbelsteen
  • Start date Start date
J

Joris Dobbelsteen

I'm looking for a comparison of the power consumption of the EPIA PD10000
compared to an AMD Sempron in a power conservative way.

The box will be doing nothing most of the time, so I'm very interested in
the power usage. For the PD10000 they are easy to find. For the AMD it's a
lot harder (especially if including a motherboard and such devices).

Does anyone know where I can find this or maybe have any data?
I need information of a fairly complete system...

- Joris
 
I'm looking for a comparison of the power consumption of the EPIA PD10000
compared to an AMD Sempron in a power conservative way.

The box will be doing nothing most of the time, so I'm very interested in
the power usage. For the PD10000 they are easy to find. For the AMD it's a
lot harder (especially if including a motherboard and such devices).

Does anyone know where I can find this or maybe have any data?
I need information of a fairly complete system...
Click to expand...


Which Sempron, Skt. A or 754?
Depends on the speed of the Sempron too.

Also depends on what you set the bus speed to, and the
optional adjustment of the CPU vcore, since it need not run
at full stock vcore if/when underclocked. Likewise the EPIA
can be underclocked for further power reduction, as well as
reduction of power used by motherboard and memory.

Too many variables to come up with a figure, especially
since we don't know what level of performance you really
need. Even if the system will be idle most of the time,
that tells us nothing of the performance level you would
need at those moments it's _not_ idle but rather IS doing
whatever-it-is you need done with it.

If you just want a generic answer, skt A Sempron should be
about 25W higher. skt 754, 16W higher. Rest of system
(drives, other cards, amount of memory) can be assumed
constant. However, a fair comparison would assume the
Sempron is underclocked since it does offer substantially
higher performance, so the power differences would be even
less presuming it's either not a skt A board, or if it is,
it supports and you have working (via OS) ACPI HALT state.

Then there's video. Integrated on both or no? Again more
variables that could end up as significant as which
CPU/platform.

For ultimate lowest power possible regardless of the
performance, underclock the EPIA to 66MHz FSB, 66MHz memory
bus, and if the board allows it, try to reduce the CPU vcore
some too. Use least amount of high density memory possible
and compact flash in lieu of a hard drive. I'd expect total
wattage to be under 35W in that config, but that's just a
guesstimation. When going that low even having a floppy or
optical drive is a significant % so a clearer definition of
the role of the system would need be considered.
 
kony said:
Which Sempron, Skt. A or 754?
Depends on the speed of the Sempron too.
Click to expand...

Ill probably go for the 754 socket, because its currently available. If I go
that route, just bought the EPIA board. Still I would like to know for
probable future solutions.
Also depends on what you set the bus speed to, and the
optional adjustment of the CPU vcore, since it need not run
at full stock vcore if/when underclocked. Likewise the EPIA
can be underclocked for further power reduction, as well as
reduction of power used by motherboard and memory.

Too many variables to come up with a figure, especially
since we don't know what level of performance you really
need. Even if the system will be idle most of the time,
that tells us nothing of the performance level you would
need at those moments it's _not_ idle but rather IS doing
whatever-it-is you need done with it.
Click to expand...

It will be idle arround 99% of the time. Its doing common server tasks, such
as firewall, proxy, fire serving, print server, web server, FTP server,
database server, etc...
Sometimes I will load the device with a short but demanding task. Its
usually related to database processing and currently the EPIA V10000 takes
5-10 minutes to do so. I expect better performance from my current PD10000
because the increased memory performance and 256 MB extra RAM I currently
have (its currently 1GB). Of course these tasks will benefit from the
sempron.
Nightly tasks are not much of a consideration for the performance, they run
at night because I will not be using the system at those times and could not
care less about performance.
If you just want a generic answer, skt A Sempron should be
about 25W higher. skt 754, 16W higher. Rest of system
(drives, other cards, amount of memory) can be assumed
constant. However, a fair comparison would assume the
Sempron is underclocked since it does offer substantially
higher performance, so the power differences would be even
less presuming it's either not a skt A board, or if it is,
it supports and you have working (via OS) ACPI HALT state.
Click to expand...

Any disks and such should not be considered, because these are equal between
the systems. Thus they can be factored out.
Then there's video. Integrated on both or no? Again more
variables that could end up as significant as which
CPU/platform.
Click to expand...

Its a server, so I don't really care. Most are capable of doing SVGA at
1024x786, which is sufficient for that device.
For ultimate lowest power possible regardless of the
performance, underclock the EPIA to 66MHz FSB, 66MHz memory
bus, and if the board allows it, try to reduce the CPU vcore
some too. Use least amount of high density memory possible
and compact flash in lieu of a hard drive. I'd expect total
wattage to be under 35W in that config, but that's just a
guesstimation. When going that low even having a floppy or
optical drive is a significant % so a clearer definition of
the role of the system would need be considered.
Click to expand...

It will need a hard disk because of the 200 GB storage that will provide.

I recon the Sempron uses about 16W more when idle than the EPIA solution in
the generic case. Of course with the Sempron providing significantly better
performance...

- Joris
 
It will be idle arround 99% of the time. Its doing common server tasks, such
as firewall, proxy, fire serving, print server, web server, FTP server,
database server, etc...
Click to expand...

So you say, but it does not matter how moften it is idle as
much as how much performance it needs have, the two are
entirely different parameters even with the aforementioned
tasks, which could be serving 1000s of clients or somewhere
between 0 and 1.
Sometimes I will load the device with a short but demanding task. Its
usually related to database processing and currently the EPIA V10000 takes
5-10 minutes to do so. I expect better performance from my current PD10000
because the increased memory performance and 256 MB extra RAM I currently
have (its currently 1GB). Of course these tasks will benefit from the
sempron.
Nightly tasks are not much of a consideration for the performance, they run
at night because I will not be using the system at those times and could not
care less about performance.
Click to expand...

At night you have the most benefit from the faster system.
By having the faster system the cued jobs can finish
significantly faster and the entire system can power down.
That alone could make far more difference than any other
considerations, IF the aforementioned server tasks don't
need 24/7 uptime.

Any disks and such should not be considered, because these are equal between
the systems. Thus they can be factored out.
Click to expand...

If it is manditory that these disks always be running, yes.
If there is the potential to spin them down when a job
finishes, it would need be factured how much quicker the
faster CPU/platform finishes such jobs. The ultimate goal
for lowest power is not to use a crippled CPU to get there
but to shut down any and all unused parts of a system if not
whole thing. I can see that may not be possible on this
with the server functionality.

Its a server, so I don't really care. Most are capable of doing SVGA at
1024x786, which is sufficient for that device.
Click to expand...

You should care because the video card is a larger power
difference than the CPU in many cases. Possibly not in this
one since the EPIA is SUCH a low power consumer but then
there are still a couple dozen watts that could be wasted by
a poor video card choice, with integrated video tending to
be the lowest, it could easily make as much difference in a
situation where the CPU were idle 99% of the time.


It will need a hard disk because of the 200 GB storage that will provide.

I recon the Sempron uses about 16W more when idle than the EPIA solution in
the generic case. Of course with the Sempron providing significantly better
performance...
Click to expand...

I don't have any accurate figures but would expect some more
competitive power reduction from a moderate underclock of
the Sempron. Lowering Vcore by 0.2V and the clockspeed by
25% (or more as needed to retain stability at the lower
vcore) might put the power consumption difference in a
negligable range but using a newer platform with the Sempron
might even ofset that, for example if the skt 754 board has
GbE integrated but the Epia didn't and had to have an
external card (or other cards/features) added.
 
Back
Top