Hi All,

I am trying to build myself a little home server, mainly as a central
repository for files with both local and remote access. Whilst I've
been building PCs for over 20 years I'm a bit out of touch and rarely
went into and detail (general purpose boxes as such).

I nearly have a solution via FTP to a 4G Pen Drive in the back of my
router but whilst it's extremely quiet and low power it isn't 'nice'.
;-)

So, on Sat a mate gave me an old Asrock K8V-XE Mobo and Sempron 2600+
chip.I found a 1G DIMM and stuck in a couple of 500G SATA drives and a
trial copy of WHS. So far so good, runs pretty well and / but draws
60W.

However it was free, has 4 x SATA II (and 2 x PATA ports / 4 devices)
and eSATA (so then 3 internal + 1 eSATA) + I have an eSATA drive
enclosure, for external backups).

I think a bit of the problem (power wise) may be that I don't think
the Sempron supports AMD's 'Cool and Quiet' so it may be that a
similar spec CPU that does may end up using less power?

I've also got an Asrock ATOM 330 A330GC (MIni-ITX) board on it way. It
only has 2 x SATA II but does have GB Ethernet and time will tell how
much (less?) power it draws. ;-)

Anyway, I have some thoughts / questions surrounding the above if I
may.

1) Is it possible to fit an AMD 'Mobile' (s754) chip in this desktop
board and if so would it draw much less power (I'm not particularly
bothered about 'performance' as long as it's generally useable. I
can't justify the long term running costs for a potential average
utilisation of 1%). ;-)

2) If not a mobile chip what's the lowest spec desktop chip that would
support 'Cool n Quiet' (or whatever power saving tech's apply).

3) The A330 board was the cheapest I could find and with 2 x 1Tb
'Green' drives would probably be sufficient. However, it does have an
ATA100 IDE connector and I though I could stick a drive on there if I
needed the extra capacity [1][2].

As an alternative to going very low power I've been playing with WOL
on this WHS (and the desktops) and had reasonable success, certainly
when it comes to waking it up (locally and remotely).

Lastly, I've currently got the WHS in a cheapo mini tower case that
happens to have a reasonable layout with plenty of space for drives
and has a std PSU. I assume such a low power solution would (should)
have a matching low power '80 Plus' PSU? An online-calculator gave me
a 120W PSU but those I have found seem to be of the smaller (ITX?)
form factor?

Cheers, T i m

[1] Can you actually get 'big, green and PATA' drives (cheap)?

[2] I've also got a bi-directional PATA<>SATA converter on the way
with the thought that whilst it might add a bit more load (compared
with going without) it might cover the question above (giving me
another SATA port). But, what sort of performance degradation would I
see if it's only ATA 100 and SATA(1) please? I have a USB floppy and
DVDRW drive. Oh, and can you generally set which port carries the boot
/ system drive as I believe WHS makes most use of the system drive (so
I would like it to be as quick as is practical).

p.s. An old P4 / 1.7 averaged about 56W and a PIII / 800 about 50 (but
neither had 'Green' drives).

On Mon, 08 Feb 2010 12:29:47 +0000, T i m wrote:
> Hi All,
>
> I am trying to build myself a little home server, mainly as a central
> repository for files with both local and remote access. Whilst I've
> been building PCs for over 20 years I'm a bit out of touch and rarely
> went into and detail (general purpose boxes as such).

While not an answer to any of your questions (hence snipped) I've run a
low power mini-itx system consisting of a 600MHz processor, 120GB 2.5 inch
drive and wifi connection. Total power consumption was about 16Watts.

Although more expensive than free :-) the EPIA mb and case can sometimes
be got off eBay (I paid ~ £30 for mine) and the drive was salvage from a
laptop upgrade. The main thing is that when you're running a machine
24 * 7 * 365 the power cost is about £1 per Watt, so paying a little more
for low power does pay off.

On Mon, 08 Feb 2010 16:40:06 GMT, pete <no-one> wrote:

>On Mon, 08 Feb 2010 12:29:47 +0000, T i m wrote:
>> Hi All,
>>
>> I am trying to build myself a little home server, mainly as a central
>> repository for files with both local and remote access. Whilst I've
>> been building PCs for over 20 years I'm a bit out of touch and rarely
>> went into and detail (general purpose boxes as such).
>
>While not an answer to any of your questions (hence snipped) I've run a
>low power mini-itx system consisting of a 600MHz processor, 120GB 2.5 inch
>drive and wifi connection. Total power consumption was about 16Watts.

Sweet.

I've fired up the a330 with one 500G Samsung F2 SATA and it's
currently installing WHS via a USB DVDRW and it's averaging 40W.

I dare say that will drop with the DVD, kbd and mouse out?
>
>Although more expensive than free :-)

(yeah, hard to beat that one) ;-)

> the EPIA mb and case can sometimes
>be got off eBay (I paid ~ £30 for mine) and the drive was salvage from a
>laptop upgrade.

I've seen those std 5-1/4 bays that take 4 x 2.5" laptop drives. Very
neat and low power but probably an expensive way to get say 2TB?

> The main thing is that when you're running a machine
>24 * 7 * 365 the power cost is about £1 per Watt, so paying a little more
>for low power does pay off.

A little yes. I'll have to see what this settles down to then see how
'awkward running it WOL would be (so far it's been ok and it's very
early days). If WOL isn't any / much trouble then I might stick with
the bigger setup (the K8V-XE board, possibly with a different CPU)
and use the A330 to replace the Apple Mini I'm currently running as a
workstation.

Cheers, T i m

The message <g2uvm516crk7f80rdktljb27d9jopkhul7>
from T i m <news> contains these words:

> Hi All,

> I am trying to build myself a little home server, mainly as a central
> repository for files with both local and remote access. Whilst I've
> been building PCs for over 20 years I'm a bit out of touch and rarely
> went into and detail (general purpose boxes as such).

> I nearly have a solution via FTP to a 4G Pen Drive in the back of my
> router but whilst it's extremely quiet and low power it isn't 'nice'.
> ;-)

> So, on Sat a mate gave me an old Asrock K8V-XE Mobo and Sempron 2600+
> chip.I found a 1G DIMM and stuck in a couple of 500G SATA drives and a
> trial copy of WHS. So far so good, runs pretty well and / but draws
> 60W.

> However it was free, has 4 x SATA II (and 2 x PATA ports / 4 devices)
> and eSATA (so then 3 internal + 1 eSATA) + I have an eSATA drive
> enclosure, for external backups).

> I think a bit of the problem (power wise) may be that I don't think
> the Sempron supports AMD's 'Cool and Quiet' so it may be that a
> similar spec CPU that does may end up using less power?

> I've also got an Asrock ATOM 330 A330GC (MIni-ITX) board on it way. It
> only has 2 x SATA II but does have GB Ethernet and time will tell how
> much (less?) power it draws. ;-)

> Anyway, I have some thoughts / questions surrounding the above if I
> may.

> 1) Is it possible to fit an AMD 'Mobile' (s754) chip in this desktop
> board and if so would it draw much less power (I'm not particularly
> bothered about 'performance' as long as it's generally useable. I
> can't justify the long term running costs for a potential average
> utilisation of 1%). ;-)

> 2) If not a mobile chip what's the lowest spec desktop chip that would
> support 'Cool n Quiet' (or whatever power saving tech's apply).

> 3) The A330 board was the cheapest I could find and with 2 x 1Tb
> 'Green' drives would probably be sufficient. However, it does have an
> ATA100 IDE connector and I though I could stick a drive on there if I
> needed the extra capacity [1][2].

> As an alternative to going very low power I've been playing with WOL
> on this WHS (and the desktops) and had reasonable success, certainly
> when it comes to waking it up (locally and remotely).

> Lastly, I've currently got the WHS in a cheapo mini tower case that
> happens to have a reasonable layout with plenty of space for drives
> and has a std PSU. I assume such a low power solution would (should)
> have a matching low power '80 Plus' PSU? An online-calculator gave me
> a 120W PSU but those I have found seem to be of the smaller (ITX?)
> form factor?

> Cheers, T i m

> [1] Can you actually get 'big, green and PATA' drives (cheap)?

> [2] I've also got a bi-directional PATA<>SATA converter on the way
> with the thought that whilst it might add a bit more load (compared
> with going without) it might cover the question above (giving me
> another SATA port). But, what sort of performance degradation would I
> see if it's only ATA 100 and SATA(1) please? I have a USB floppy and
> DVDRW drive. Oh, and can you generally set which port carries the boot
> / system drive as I believe WHS makes most use of the system drive (so
> I would like it to be as quick as is practical).

> p.s. An old P4 / 1.7 averaged about 56W and a PIII / 800 about 50 (but
> neither had 'Green' drives).

Here's a posting I made sometime in February last year in this
newsgroup which I copied so I could repost it whenever this question
comes up (saves me a shed load of typing ;-). I think this will provide
you with some food for thought on this subject.

StartQuote:
================================================== ================
The message <gmjkd0$8af$1>
from Gordon Henderson <gordon+usenet> contains these words:

> In article <6v45ggFi6gfiU5>,
> John Stumbles <john.stumbles> wrote:
> >I keep thinking of sticking a large HDD or two into a smallish
> >low-powered base unit running NFS, rsync backups and apache (for
> >intranet) and putting it in the server room (aka cupboard under the stairs
> >:-)). However a quick trawl through dabs, ebuyer and novatech reveals a
> >complete lack of cheap PCs of any kind: nothing much under £200, whereas
> >a year or two back you used to be able to get a basic machine with a
> >Linux distro for about £160. Any pointers?

> If you're up for building your own...

> Cheapest low power general purpose mobo right now is probably an Intel
> Atom, however the VIA boards are lower power (but slower, not an issue
> for what you need)

> Atom mobo:

> [..]

> VIA Motherboard:

> [..]

> Although that's out of stock, but due soon, and there are other VIA
> motherboards there too.

> Add in a suitable case, 2 WD Green drives and off you go. My home
> server/firewall/backup thing has an Atom mobo, and 2 WDC Green 1TB drives
> in it and it sucks about 45 watts. If I'd used a VIA board it would be
> about 10W less, but I wanted the Atom for other reasons.

> One advantage of the VIA boards is lack of fan, the Atom boards have
> a weeny whiney little fan on them and the one in my workstation is
> driving me nuts and an investigation might just be undertaken to see
> if it can be slowed down or even removed/replaced... (it's not on the
> processor but the bridge chip!)

That's interesting, I'm running a FreeNAS box with four Samsung HD103UJ
disks mounted and this takes just 54 to 55 watts (depending on drive
temperature - I don't spin them down). This box is, as near as damnit,
totally silent despite the low level of 120Hz spindle vibration which,
in the typical 'tin box' can be magnified to a level of annoyance out of
all proportion, especially if resting on a desktop surface that has
sounding board resonance modes around this frequency.

With each drive accounting for about 8 watts each in the power budget
when spinning, this leaves the remaining 22 or so watts accounted for by
the Asrock AliveNF7G-FULLHD R3.0 and an underclocked 1250 AMD Semperon
running at 800MHz with the core voltage dropped to 0.850 volts. I have a
single stick of 1GB DDR2 memory fitted (the benefit of dual channel is
rather lost in this application, as I proved by trying it with a couple
of 1GB dimms).

Mind you, the original[1] 145 watt mini ATX psu fitted in this
Gateway2000 desktop case does contribute a good 5 to 10 watts
improvement over the typical standard 3 to 4 hundred watt ATX psu
designs generally available.

The Gateway case is especially good for this job since it is quite
sturdily made and had space to fit the extra two drives, using MoBo
stand off pillars, just behind the front panel where they could get
first sniff of the incoming air to keep them nice and cool (too cool as
it turns out) without the need to supplement the extremely quiet airflow
generated by the lone PSU fan.

For file serving / print serving, you need very little processor power,
so you can safely underclock to the minimum speed possible (but be wary
of departing from standard minimum FSB clocks though since this could
effect I/O performance - HDD and Gbit ethernet).

The normal multiplier setting on the Sempron is 11 (afaicr) which, on
the "200 MHz" FSB, clocks it at 2.2GHz. I originally tried the minimum
of 4 with a 150MHz FSB setting (600MHz) but found this was knocking the
data transfer rate down by a good 10% or more, so I reset the FSB back
to its normal setting of "200 MHz", raising the CPU clock to 800 MHz.

The penalty in power consumption was less than a watt, even after
raising the Vcore from its minimum setting of 0.800v up to 0.850v, so it
seemed a worthwhile compromise to regain full I/O performance. TBH. I
was quite pleasantly surprised to discover that such control over the
CPU multiplier setting was even possible (Hats off to AMD for
re-instating this feature!).

Mind you, all this undervolting (especially the undervolting) and
multiplier and FSB adjustments were only made possible by the cmos
configuration including them in the setup menu.

An awful lot of MoBo bioses have shite cmos configuration menus that
limit such options, especially the Vcore one where it is arrogantly
assumed that only overclockers will be tuning the system so only
increases of voltage over and above whatever the cpu default happens to
be are offered. Unfortunately, this often only becomes apparent after
you've set it up and booted into the cmos setup.

When I was choosing a processor and MoBo combination for the NAS
upgrade, I was looking for the cheapest combo that offered a minimum of
4 sata ports with built in Gb ethernet port and took 'pot luck' over the
cmos options. I guess I was lucky, but the dynamic Vcore power saving
option with this generation of CPU is probably what guaranteed the
availability of a full range of Vcore settings in the cmos setup.

The Intel CPU route might not offer the same multiplier control (I'm
not sure if Intel have also seen the light and restored this as an end
user option) so I can't say whether such might be a better choice.
However, if you can find on-line reviews or detail specs that list the
cmos options for your prospective purchase, you should be able to avoid
the less suitable examples.

The power savings of an especially low power motherboard and processor
might not be necessary unless you're only fitting one or two drives and
planning on subjecting them to the stress of spindown power saving in
which case, a 5 or 10 watt reduction on a base consumption of 15 to 20
watts would represent a sizeable reduction percentage-wise.

Once you go for an instant response 4 drive setup, the marginal power
saving between an especially low power consumption MoBo and an
undervolted, underclocked mainstream one becomes much less important.

If I placed more value on the saving of some 30 odd quid a year on the
electricity bill over the integrity of and instant access to my data, I
could employ spindown power saving to the tune of some 30 watts. Since I
don't, that extra 5 or 10 watts used by my choice of MoBo over a
specialised low power one, becomes of little significance.

The choice of PSU becomes more significant once we're considering
system power levels around the 50 watt or less mark since a cheap
commodity unit could well account for an extra 10 watts over and above
what a modestly sized high efficiency unit could achieve. I didn't shop
around for such, since it just so happened that the original 145 watt
unit in the Gateway 2000 desktop style case that I decided to use as a
replacement for the NEC Powermate II box I had been previously using
just happened to be noticably less power hungry than the typical 300/400
watt ATX psu.

If you're looking to putting a cost effective 3 or 4 disk NAS box or
fileserver together, I could recommend you look out for a Gateway 2000
desktop (not tower) case for its solidity and efficient PSU but beware
of a possible issue with this PSU and certain MoBos as detailed in
note[1]. I can also recommend that Asrock AliveNF7G-FULLHD R3.0 and the
1250 AMD Semperon I'm currently using (and Samsung HD103UJ drives as
well ;-)

The only other snag with this Gateway case is that the I/O shield area
isn't the standard letterbox cutout, the i/o holes have been directly
punched out of the rear panel to accomodate the original MoBo.

You'll either have to hack out the existing holes or create a standard
I/O shield letterbox aperture to accomodate a standard shield plate to
allow a replacement MoBo to be installed. Not a showstopper, just a
trivial impediment to upgrading the MoBo for someone prepared, if needs
be, to spend a fiver in Aldi on a pair of 'Tinsnips' ;-)

[1] Mind you, it's not quite so original now. When I upgraded from an
Asrock micro ATX P4 MoBo, which would allow the CPU vrm to be powered
off the 5 volt rail in the absence of a 4 pin 12 volt connector, to the
current AM2 socket one which has (rather unconscionably) no such option,
I had to wire a 4 pin 12v connector into the PSU on account it lacked
this refinement.

I also had to add the guts of a cheap 5.2v 1.2A smpsu wallwart to beef
up the surprisingly low powered 5VSB rail (just 100mA rated!!!) which
wasn't anywhere near enough to allow the new board to fire up.

It was no wonder I had startup issues with some of the previous MoBos
I'd tried on previous occasions. I guess I hadn't looked quite so
closely at the voltage ratings to spot that it wasn't even a mere one
(or even half!) amp rating. One wouldn't normally bother trying to
modify the innards of a PSU, but, in this case, it was deemed well worth
the trouble to hang on to the quietness and high efficiency of the
original.

I do have a 270 watt micro ATX PSU of the type intended for those SFF
cases which I could have fitted with very little modification to the
case, but, although its efficiency was noticably better than the usual
commodity ATX units, it wasn't quite as good as the original (but I keep
it in the "Server Spares" box just in case the original finally pops its
clogs and heads off for 'Silicon Heaven'(tm Red Dwarf) ;-)

================================================== ================

EndQuote:

I've since replaced two of those 7200rpm drives with a couple of
Samsung's eco green 1TB units which seem to, as best as I can judge,
save a couple of watts apiece on the standard drives. I've also upped
the cpu clock multiplier a little to improve the Gbit throughput (6
instead of 4 for a 1.2GHz clock speed, afaicr - just over half rated
speed with a slight Vcore increase to, afaicr, 0.9v) and fitted a second
1GB ddr2 dimm to enable dual channel memory mode.

The net effect of all those changes means the server now consumes some
53 watts. I estimate that if all the drives were the eco green types,
this would probably drop to just below the 50 watt mark, which, for an
always available 4TB server, is pretty economic.

Write speeds, when the source files on this nearly 6 year old win2k box
aren't fragmented, just exceeds 2GB per minute (about 35MB/s) and read
speeds can exceed this by a small margin. I think a more up to date
system box would show even better results since I suspect I'm
encroaching on the Gbit speed limit of the win2k box but I don't often
see any decent spec systems _with_ Gbit ethernet included to test this
out.

HTH & HAND

On Tue, 9 Feb 2010 03:28:43 GMT, Johnny B Good
<jcs.computersbutt> wrote:


> Here's a posting I made sometime in February last year in this
>newsgroup which I copied so I could repost it whenever this question
>comes up (saves me a shed load of typing ;-). I think this will provide
>you with some food for thought on this subject.

Thanks.
>
<snip>
>
>> One advantage of the VIA boards is lack of fan, the Atom boards have
>> a weeny whiney little fan on them and the one in my workstation is
>> driving me nuts and an investigation might just be undertaken to see
>> if it can be slowed down or even removed/replaced... (it's not on the
>> processor but the bridge chip!)

And this one goes through a rumbly stage as it spins up and is also
only 3 wire so I might replace that anyway with something bigger, 4
wire and held off the Mobo.
>
> That's interesting, I'm running a FreeNAS box with four Samsung HD103UJ
>disks mounted and this takes just 54 to 55 watts (depending on drive
>temperature - I don't spin them down).

This seems to be a common value, the 50W range I mean. As you mention
elsewhere you are in the 'diminishing returns' area when trying to get
lower.

<snip>

> With each drive accounting for about 8 watts each in the power budget
>when spinning, this leaves the remaining 22 or so watts accounted for by
>the Asrock AliveNF7G-FULLHD R3.0 and an underclocked 1250 AMD Semperon
>running at 800MHz with the core voltage dropped to 0.850 volts. I have a
>single stick of 1GB DDR2 memory fitted (the benefit of dual channel is
>rather lost in this application, as I proved by trying it with a couple
>of 1GB dimms).

I went for just 1 DIMM for similar reasons.
>
> Mind you, the original[1] 145 watt mini ATX psu fitted in this
>Gateway2000 desktop case does contribute a good 5 to 10 watts
>improvement over the typical standard 3 to 4 hundred watt ATX psu
>designs generally available.

And that's another thing. Whilst I didn't see much difference in the W
between two (cheapo) PSU's there was quite a difference in the VA and
I think we are billed on VA?
>
> The Gateway case is especially good for this job since it is quite
>sturdily made and had space to fit the extra two drives, using MoBo
>stand off pillars, just behind the front panel where they could get
>first sniff of the incoming air to keep them nice and cool (too cool as
>it turns out) without the need to supplement the extremely quiet airflow
>generated by the lone PSU fan.

I try to think of the case thermo/aerodynamics like that, sometimes
blocking off side and back vents to ensure an actual 'flow' of air
through the case and not just round and round within.
>
> For file serving / print serving, you need very little processor power,
>so you can safely underclock to the minimum speed possible (but be wary
>of departing from standard minimum FSB clocks though since this could
>effect I/O performance - HDD and Gbit ethernet).

I don't think I'd go as far as playing with the clocking but might
look for a more suitable chip (with the Quiet n Cool feature etc).
>
<snip>
>
> When I was choosing a processor and MoBo combination for the NAS
>upgrade, I was looking for the cheapest combo that offered a minimum of
>4 sata ports with built in Gb ethernet port and took 'pot luck' over the
>cmos options.

It's funny isn't it, considering how many manufacturers and boards
there are out there at the moment how quickly your choices become
limited if you have a specific requirement (like low power, 4 x SATA,
GB and cheap). My Sempron setup has the 4 x SATA and was cheap (I just
bought the RAM) but only has 100M Ethernet. Now, a couple of my PC's
have GB, as is my switch but I'm not sure how much difference GB makes
in the real world (I think I read no where near as much as 10X
better). However, I guess twice as fast could mean a backup takes half
as long so could be worth it?

<snip>

> Once you go for an instant response 4 drive setup, the marginal power
>saving between an especially low power consumption MoBo and an
>undervolted, underclocked mainstream one becomes much less important.

Good point.
>
> If I placed more value on the saving of some 30 odd quid a year on the
>electricity bill over the integrity of and instant access to my data, I
>could employ spindown power saving to the tune of some 30 watts. Since I
>don't, that extra 5 or 10 watts used by my choice of MoBo over a
>specialised low power one, becomes of little significance.

True. Outside the W <> VA question I'm looking at ~45W(Atom) V 60W
(AMD) with much more versatility in the AMD setup. However, I'll have
to measure it again with a GB card to be sure ad if my up-time_needs
don't warrant it being on 24/7 then the *difference* matters even
less. (It *was* sometimes awkward that my hardware NAS wasn't on 24/7
but because it wasn't big or flexible enough to do anything else ...
it didn't need to be).
>
> The choice of PSU becomes more significant once we're considering
>system power levels around the 50 watt or less mark since a cheap
>commodity unit could well account for an extra 10 watts over and above
>what a modestly sized high efficiency unit could achieve. I didn't shop
>around for such, since it just so happened that the original 145 watt
>unit in the Gateway 2000 desktop style case that I decided to use as a
>replacement for the NEC Powermate II box I had been previously using
>just happened to be noticably less power hungry than the typical 300/400
>watt ATX psu.

I've got a 120W PSU in the old Flex PIII system I was playing with
before and I might try the Atom board in there with the thoughts of it
being a workstation to replace the Mini (then it wouldn't matter that
it only had one 3.5" hdd location).
>

> The only other snag with this Gateway case is that the I/O shield area
>isn't the standard letterbox cutout, the i/o holes have been directly
>punched out of the rear panel to accomodate the original MoBo.

I cut down a wireless card to put in the (Gigabyte) slimline Flex case
and the Atom board has a single PCI slot. ;-)
>
> You'll either have to hack out the existing holes or create a standard
>I/O shield letterbox aperture to accomodate a standard shield plate to
>allow a replacement MoBo to be installed. Not a showstopper, just a
>trivial impediment to upgrading the MoBo for someone prepared, if needs
>be, to spend a fiver in Aldi on a pair of 'Tinsnips' ;-)

I have a pair of nibblers that are brilliant for opening up fan holes
(I normally cut the perforated fan positions out and replace hem with
finger guards). The head of the cutter goes through one hole and you
just chomp your way round, finishing off with a half round file. No
chance of distortion but every chance of sending a cutting into yer
PC's so do it elsewhere. ;-)

Mine are like this:

http://en.wikipedia.org/wiki/File:Nibblerdetail1.jpg

<snip>

> I do have a 270 watt micro ATX PSU of the type intended for those SFF
>cases which I could have fitted with very little modification to the
>case, but, although its efficiency was noticably better than the usual
>commodity ATX units, it wasn't quite as good as the original (but I keep
>it in the "Server Spares" box just in case the original finally pops its
>clogs and heads off for 'Silicon Heaven'(tm Red Dwarf) ;-)

Can you easily / cheaply get std format PSU's in low power and high
efficiency (80 Plus?) ratings? I haven't seen many?
>

>EndQuote:
>
> I've since replaced two of those 7200rpm drives with a couple of
>Samsung's eco green 1TB units which seem to, as best as I can judge,
>save a couple of watts apiece on the standard drives.

I was thinking of doing similar if I stick with the Atom. For
cheapness I have both setups in small/midi ATX (£22) cases that happen
to have loads of drive bays. Seeing all the space around the Mini-ITX
board and knowing I might be able to find a small / PSU and only
needing room for two HDD's that it wouldn't take much to make a wooden
case (probably using the metalwork from a case for the actual
mountings) but wondered about RFI? Mind you, many cases have plastic
front panels (but could be RFI treated) but what of the Perspex boxes?

> I've also upped
>the cpu clock multiplier a little to improve the Gbit throughput (6
>instead of 4 for a 1.2GHz clock speed, afaicr - just over half rated
>speed with a slight Vcore increase to, afaicr, 0.9v) and fitted a second
>1GB ddr2 dimm to enable dual channel memory mode.

I'm still interested in exploring a chip that supports Cool n Quiet
and or a Mobile CPU.
>
> The net effect of all those changes means the server now consumes some
>53 watts. I estimate that if all the drives were the eco green types,
>this would probably drop to just below the 50 watt mark, which, for an
>always available 4TB server, is pretty economic.

Yup, I think that's about the best bang / watt.
>
> Write speeds, when the source files on this nearly 6 year old win2k box
>aren't fragmented, just exceeds 2GB per minute (about 35MB/s) and read
>speeds can exceed this by a small margin. I think a more up to date
>system box would show even better results since I suspect I'm
>encroaching on the Gbit speed limit of the win2k box but I don't often
>see any decent spec systems _with_ Gbit ethernet included to test this
>out.

The think is, like your pot_luck to the CMOS settings thing I think if
your focus has been best VFM / ease / low energy for max capacity then
to specifically try to put network performance into the equation
(beyond insuring it's working reasonably etc) might be a factor too
far.

Cheers and thanks for the feedback. ;-)

T i m

On Tue, 09 Feb 2010 10:45:04 +0000
T i m <news> wrote:

> It's funny isn't it, considering how many manufacturers and boards
> there are out there at the moment how quickly your choices become
> limited if you have a specific requirement (like low power, 4 x SATA,
> GB and cheap).

I'd have thought four drives and low power were mutually exclusive
really. Or did you cover that further up the thread?

On Tue, 09 Feb 2010 10:45:04 +0000, T i m <news> wrote:

>>> One advantage of the VIA boards is lack of fan, the Atom boards have
>>> a weeny whiney little fan on them and the one in my workstation is
>>> driving me nuts and an investigation might just be undertaken to see
>>> if it can be slowed down or even removed/replaced... (it's not on the
>>> processor but the bridge chip!)
>
>And this one goes through a rumbly stage as it spins up and is also
>only 3 wire so I might replace that anyway with something bigger, 4
>wire and held off the Mobo.

I replace those little whiners with larger (normally 60mm) fans
running slow, whenever there's room. Mounting the buggers in the right
position can be... rather bodgey. Remember to leave the heatsink on
the chip.

If there's already a case fan creating a decent through draft, switch
out the whiner+heatsink for a bigger passive heatsink (Zalman do
several).

(Why not repurpose the Mini to this, though?)

Cheers - Jaimie

On Tue, 9 Feb 2010 11:08:39 +0000, Rob Morley <nospam>
wrote:

>On Tue, 09 Feb 2010 10:45:04 +0000
>T i m <news> wrote:
>
>> It's funny isn't it, considering how many manufacturers and boards
>> there are out there at the moment how quickly your choices become
>> limited if you have a specific requirement (like low power, 4 x SATA,
>> GB and cheap).
>
>I'd have thought four drives and low power were mutually exclusive
>really. Or did you cover that further up the thread?

Erm, don't think so, but then I've been going round in circles a bit
as I get to grips with it all.

Like, if you are running WHS I think it suits 3 x drives to make the
best use of drive extender. I have 3 new 500G drives but they aren't
quite as low power as some of the 'green drives (but only by a couple
of W / drive it seems).

Then there's the PSUs. With the thing only (now) drawing ~40W I'm not
sure how efficient the 'free-with-case' 350W PSU is but buying
anything more efficient *will* be disproportionately expensive
(infinitely so if you count this PSU as 'free with he case').

I have a few PSU's here so I think I'll do some tests and measure the
input power(W and VA) on them all and see what turns up.

But yes, I guess, model for model, 1 x 2TB drive will be less power
consumption than 3 x 750G but then I lose some of the point of the
WHS, drive extender and the flexibility (I had 2 x 500G in the AMD
powered server and when I got the Atom board and wanted a drive I
'removed' the second / active drive from the first server using the DE
'remove' feature that farms all the data off onto other drives. Nice).

I think at the moment it's down to the AMD - 3 x 500Gand using it
on-demand / WOL (probably / realistically good enough for our needs
here) or Atom + 2 'Green' 750-1TB drives an on 24/7 (better if I
consider an i/p CCTV add-on as well).

Cheers,

Confused of London. ;-(

On Tue, 09 Feb 2010 12:02:32 +0000, Jaimie Vandenbergh
<jaimie> wrote:

>On Tue, 09 Feb 2010 10:45:04 +0000, T i m <news> wrote:
>>I replace those little whiners with larger (normally 60mm) fans
>running slow, whenever there's room. Mounting the buggers in the right
>position can be... rather bodgey.

I've just ordered from eBay an 80mm 4 pin (PWM) fan as the ATX case
I'm currently using has one of those telescopic ducts from a fan
(basic 2 wire from stock) mounted on the side case, pointing at the
CPU / other heat sinks. The Asrock A330 Mobo has a 4 wire chassis fan
output so hopefully would work with the BIOS and fan to keep a
reasonable temperature with the minimum of power and noise.

> Remember to leave the heatsink on
>the chip.

Funny that the CPU runs bare and it's the bridge that gets hot. ;-)
>
>If there's already a case fan creating a decent through draft, switch
>out the whiner+heatsink for a bigger passive heatsink (Zalman do
>several).

I had considered that also, possibly with a 120mm PWM case fan on the
rear. However, I'm not sure about the airflow when used like that
(conflicting with the PUS fan) so would, with an otherwise sealed /
ported case, prefer a fan in front of the case covering the HDDs *and*
positive pressure through the case (and hence also over any passive
cooling and through the PSU).
>
>(Why not repurpose the Mini to this, though?)

Erm, I did 'consider' the C2D Mini that's sitting idle upstairs (was
daughters when she was doing a graphics / design course but she's
currently in Scotland and chose to take the much bigger PC she built
up there instead?) but for the sever role I wanted it all in one box,
no external PSU or drives etc. Plus, I think I understood to run WHS
I'd still have to have Bootcamp on there and then have all the
implications re imaging for backups etc etc?

A small part of the reason for having a NAS / Server was to be able to
transfer some of the stuff that currently cluttering up this (160G)
1.6G Mini onto to something with more space (and knowing how tricky by
comparison to the boxes I've just assembled it would be to upgrade the
160G hdd to something bigger and how easy would be how to clone the
existing Bootcamp / OSX / XP stuff off to save having to start again
<shudder>).

Cheers, T i m

On Tue, 09 Feb 2010 13:57:47 +0000, T i m <news> wrote:

>On Tue, 09 Feb 2010 12:02:32 +0000, Jaimie Vandenbergh
><jaimie> wrote:
>>I've just ordered from eBay an 80mm 4 pin (PWM) fan as the ATX case
>I'm currently using has one of those telescopic ducts from a fan
>(basic 2 wire from stock) mounted on the side case, pointing at the
>CPU / other heat sinks. The Asrock A330 Mobo has a 4 wire chassis fan
>output so hopefully would work with the BIOS and fan to keep a
>reasonable temperature with the minimum of power and noise.

Sounds perfect.

>>If there's already a case fan creating a decent through draft, switch
>>out the whiner+heatsink for a bigger passive heatsink (Zalman do
>>several).
>
>I had considered that also, possibly with a 120mm PWM case fan on the
>rear. However, I'm not sure about the airflow when used like that
>(conflicting with the PUS fan) so would, with an otherwise sealed /
>ported case, prefer a fan in front of the case covering the HDDs *and*
>positive pressure through the case (and hence also over any passive
>cooling and through the PSU).

Yep, I do the same - more intake than outflow makes controlling the
airflow *so* much easier. I very rarely add an extra out fan, but I
have been known to add cardboard baffles so as to use a back fan port
as in inflow without the air going straight to the PSU.

>>(Why not repurpose the Mini to this, though?)
>
>Erm, I did 'consider' the C2D Mini that's sitting idle upstairs (was
>daughters when she was doing a graphics / design course but she's
>currently in Scotland and chose to take the much bigger PC she built
>up there instead?) but for the sever role I wanted it all in one box,
>no external PSU or drives etc.

Sensible enough. I use a 4xSATA USB2 case with internal PSU for
external storage in one place, not the fastest thing in the world due
to the USB2 bottleneck but effective bulk storage and fine for network
access. It has a built in PSU - but the Mini doesn't, of course.

> Plus, I think I understood to run WHS
>I'd still have to have Bootcamp on there and then have all the
>implications re imaging for backups etc etc?

Yes, you're right about Bootcamp, but that would only add a little
inconvenience on rebuild-from-scratch, but not for imaging/restoring.
You can do the image/restore from OSX, see below.

>A small part of the reason for having a NAS / Server was to be able to
>transfer some of the stuff that currently cluttering up this (160G)
>1.6G Mini onto to something with more space (and knowing how tricky by
>comparison to the boxes I've just assembled it would be to upgrade the
>160G hdd to something bigger and how easy would be how to clone the
>existing Bootcamp / OSX / XP stuff off to save having to start again
><shudder>).

It's dead easy, I do the same every time I update my primary machine
(currently a Mini).

Starting position: OSX, BootCamp with Windows.
Tools needed: OSX, WinClone ([url down] but
appears to be down... readily available from versiontracker etc)

1. Use WinClone to take an image of your BootCamp Windows partition,
and store it somewhere.
2. Mount your new HD in a USB case and plug it in.
3. Use Disk Utility to partition the new drive as you would wish.
4. Use Disk Utility to 'restore' your live OSX partition to the new
drive.
5. Swap the new drive into the Mini.
6. Boot off the new drive.
7. Run BootCamp Assistant up to the point where it wants a reboot.
Don't reboot, cancel out instead. This adds the fake BIOS environment
to the boot process and only needs to be done once per hard drive.
8. Use WinClone to restore the Windows partition from wherever you put
it to its new home.

You can use OSX and WinClone to do out-of-band imaging/restoring of
the Windows partition later too (ie just steps 1 and 8), saves on
messing about with boot CDs. Normal WHS backups can be done on the
fly, as well.

Cheers - Jaimie

On Tue, 09 Feb 2010 14:26:42 +0000, Jaimie Vandenbergh
<jaimie> wrote:

<snip good stuff>
>
>Sensible enough. I use a 4xSATA USB2 case with internal PSU for
>external storage in one place, not the fastest thing in the world due
>to the USB2 bottleneck but effective bulk storage and fine for network
>access. It has a built in PSU - but the Mini doesn't, of course.

I did look at RAID NAS's but they seemed expensive and limited in
comparison to a soft 'server' type solution. I also went through the
Openfiler, FreeNAS, Ubuntu Server OpenSolaris set but not being a
'*nix man' made most of those solutions too heavy for me right now. I
generally got them working but it was often only to a limited degree
and mostly after much frustration reading / Googling for solutions. By
comparison I've installed and configured the Trial of WHS several
times on many platforms and added quite a few options with scarcely a
glance at any of the documentation.
>
>> Plus, I think I understood to run WHS
>>I'd still have to have Bootcamp on there and then have all the
>>implications re imaging for backups etc etc?
>
>Yes, you're right about Bootcamp, but that would only add a little
>inconvenience on rebuild-from-scratch, but not for imaging/restoring.
>You can do the image/restore from OSX, see below.

Oh, ok. I may well do that in any case then. ;-)
>

>It's dead easy, I do the same every time I update my primary machine
>(currently a Mini).
>
>Starting position: OSX, BootCamp with Windows.
>Tools needed: OSX, WinClone ([..] but
>appears to be down... readily available from versiontracker etc)

Ok.
>
>1. Use WinClone to take an image of your BootCamp Windows partition,
>and store it somewhere.
>2. Mount your new HD in a USB case and plug it in.
>3. Use Disk Utility to partition the new drive as you would wish.
>4. Use Disk Utility to 'restore' your live OSX partition to the new
>drive.
>5. Swap the new drive into the Mini.
>6. Boot off the new drive.

Hmm, I got this far when upgrading from the 60 to 160G drive when I
first got this Mini and it just didn't work. I mean it *worked* as in
I could see all the drives each time and it went through the motions
etc but when trying to boot the Mini from the new drive it just
spluttered and died. Doing any FS checks with DU ended up with screen
after screen of (from memory) permissions issues and what have you?
;-( I ended up installing everything from scratch. Ironically the
old disk in the same external caddy boots happily off the Mini?

>7. Run BootCamp Assistant up to the point where it wants a reboot.
>Don't reboot, cancel out instead. This adds the fake BIOS environment
>to the boot process and only needs to be done once per hard drive.

Ok.

>8. Use WinClone to restore the Windows partition from wherever you put
>it to its new home.
>
All sounds easy. ;-)

>You can use OSX and WinClone to do out-of-band imaging/restoring of
>the Windows partition later too (ie just steps 1 and 8), saves on
>messing about with boot CDs.

Ok.

>Normal WHS backups can be done on the
>fly, as well.

Nice (n thanks).

T i m

On Tue, 09 Feb 2010 14:46:47 +0000, T i m <news> wrote:

>>4. Use Disk Utility to 'restore' your live OSX partition to the new
>>drive.
>>5. Swap the new drive into the Mini.
>>6. Boot off the new drive.
>
>Hmm, I got this far when upgrading from the 60 to 160G drive when I
>first got this Mini and it just didn't work. I mean it *worked* as in
>I could see all the drives each time and it went through the motions
>etc but when trying to boot the Mini from the new drive it just
>spluttered and died. Doing any FS checks with DU ended up with screen
>after screen of (from memory) permissions issues and what have you?
>;-( I ended up installing everything from scratch. Ironically the
>old disk in the same external caddy boots happily off the Mini?

Odd. I've done this at least 15 times (really! Across 8 or so Macs)
and never had any issues...

a) The EFI boot code handles spotting bootable partitions (including
Windows ones once the BootCamp Assistant does its thing), so nothing
like setting "active" flags needs to be done.

b) Succesfully starting to boot means that the disk partition style
was okay (Intel Macs need GUID or APM format disks, but not PC-style
MBR ones), and also that the partition filesystem was HFS+ as it
should be.

c) So it sounds like the clone from the old disk to new disk got
screwed up somehow. Never seen that myself, but it's hardly unlikely.

Could the "restore" operation have been merging files into an existing
OSX install on the destination, rather than erasing the partition and
filling it from scratch? There's a tickbox for "erase destination"
that should be ticked at point 4 above to avoid that.

Indeed, at point 4 you can use any other Mac cloning tool, such as
Carbon Copy Cloner or Silverkeeper or whatever. They're generally
twice as fast as Disk Utility, and more foolproof.

Cheers - Jaimie

In article <g2uvm516crk7f80rdktljb27d9jopkhul7>, T i m
<news> writes

>I nearly have a solution via FTP to a 4G Pen Drive in the back of my
>router but whilst it's extremely quiet and low power it isn't 'nice'.

Low hassle factor
Noiseless
More space = just stick another drive in

>I think a bit of the problem (power wise) may be that I don't think
>the Sempron supports AMD's 'Cool and Quiet' so it may be that a
>similar spec CPU that does may end up using less power?

Very much. Look at AMD's processor finder.

>1) Is it possible to fit an AMD 'Mobile' (s754) chip in this desktop
>board

Yes, I think so. (RTFM? :p)

> and if so would it draw much less power (I'm not particularly
>bothered about 'performance' as long as it's generally useable.

Any cpu made in the last few years will do the job. The main issue IMO
is disk latency.

>2) If not a mobile chip what's the lowest spec desktop chip that would
>support 'Cool n Quiet' (or whatever power saving tech's apply).

have a look at AMD's processor finder. CnQ is not problem free - I have
to turn it off if I want to run anything that will use 3D features of
the graphics or I get lockups.

If you can knock down the CPU speed to reduce power - for instance
setting the FSB to 100MHZ instead of 133, that'll give you a dramatic
reduction in power used and may be enough to run the cpu without a fan
(headless), making it quieter. The AMD XP-M processors are reputed to be
particularly good at this, particularly the XP2500-M. The net's full of
obsessive spotty geeks who have already done all the legwork for you
down to the last milliwatt.

Power management in the latest releases of Linux seems to deal well with
AMD chips.

>As an alternative to going very low power I've been playing with WOL
>on this WHS (and the desktops) and had reasonable success, certainly
>when it comes to waking it up (locally and remotely).

I think, for a server, you'll find the latency (waiting while the server
wakes up and restores the pre-hibernation state) too annoying.

>Lastly, I've currently got the WHS in a cheapo mini tower case that
>happens to have a reasonable layout with plenty of space for drives
>and has a std PSU. I assume such a low power solution would (should)
>have a matching low power '80 Plus' PSU?

Nah. It's all a big marketing con, borne of idiots that look and talk
funny. If you can, get a brand-name psu which claims that it meets the
E-standard (80% or more), that'll do.

As soon as you realise that the PC industry is about endless
upgradeitis, all becomes clear. New processor? You need a new mobo!
etc.

What you need to remember is that more power (label on psu) is NOT what
you want here! If the estimator you used says 120W (and remember that's
running at full pelt, which you won't be), a suitable psu will be ca.
150 - 200W to place it at the peak of the efficiency curve.

> An online-calculator gave me
>a 120W PSU

Ok, ish. Seems rather high (assuming we're talking "standard" PC
hardware and you're aiming at low power.) You could borrow one of those
plug-in power meters. (There's been extensive discussion of them here
fairly recently), but I feel that your effort would be better spent on
concentrating on the lowest-power components you can put together for
not much. 60W would be about right.

Johnny'll be along soon enough, he's done a lot of experimenting with
low-power setups and will have an opinion.

For the purposes of your task, the Atom and the Sempron will be
virtually the same. The Atom will run a lot cooler (it's designed to run
with no heatsink), but there isn't much in the performance.

>[1] Can you actually get 'big, green and PATA' drives (cheap)?

Yes, Look at the WD Green series. Getting good reviews atm.

>[2] I've also got a bi-directional PATA<>SATA converter on the way

Are you SURE about that? They usually only work one way.

>with the thought that whilst it might add a bit more load

Hardly anything. One chip and a few support components.

> (compared
>with going without) it might cover the question above (giving me
>another SATA port). But, what sort of performance degradation would I
>see if it's only ATA 100 and SATA(1) please?

Absolutely none.

>Oh, and can you generally set which port carries the boot
>/ system drive as I believe WHS makes most use of the system drive (so
>I would like it to be as quick as is practical).

Odd question. Think about using a flash drive for the OS (64gb are cheap
now) and use spinning media for your data drives.

>p.s. An old P4 / 1.7 averaged about 56W and a PIII / 800 about 50 (but
>neither had 'Green' drives).

Nor processors with good power management.

On Wed, 10 Feb 2010 10:31:15 +0000, Mike Tomlinson
<mike> wrote:


>>I nearly have a solution via FTP to a 4G Pen Drive in the back of my
>>router but whilst it's extremely quiet and low power it isn't 'nice'.
>
>Low hassle factor
>Noiseless
>More space = just stick another drive in

I bought a 32G flash drive to replace the 4G one that is in there now
and I'm not sure if it was a device, compatibility or (ftp) client
issue but 500G of utility files were sent to the drive and most
folders more than one deep were empty.
>
>>I think a bit of the problem (power wise) may be that I don't think
>>the Sempron supports AMD's 'Cool and Quiet' so it may be that a
>>similar spec CPU that does may end up using less power?
>
>Very much. Look at AMD's processor finder.

<Googles > tries>, Hmm, selecting s754 then looking at the wattage
option gives me 51 as the lowest?

http://products.amd.com/en-us/DesktopCPUResult.aspx

>
>>1) Is it possible to fit an AMD 'Mobile' (s754) chip in this desktop
>>board
>
>Yes, I think so. (RTFM? :p)

The FM I've found for the Mobo just gives the range of CPU's that were
generally available at the time and no mention of anything
'different'. Sorta 'this is what we would recommend rather than 'you
might also be able to fit this but why would you want to' type of
thing. This Mobo was made before power saving was 'important'. ;-(
>
>> and if so would it draw much less power (I'm not particularly
>>bothered about 'performance' as long as it's generally useable.
>
>Any cpu made in the last few years will do the job. The main issue IMO
>is disk latency.

Ok.
>
>>2) If not a mobile chip what's the lowest spec desktop chip that would
>>support 'Cool n Quiet' (or whatever power saving tech's apply).
>
>have a look at AMD's processor finder. CnQ is not problem free - I have
>to turn it off if I want to run anything that will use 3D features of
>the graphics or I get lockups.

Well that one wouldn't affect me and and a server but other 'issues'
might of course.
>
>If you can knock down the CPU speed to reduce power - for instance
>setting the FSB to 100MHZ instead of 133, that'll give you a dramatic
>reduction in power used and may be enough to run the cpu without a fan
>(headless), making it quieter.

Ok, I can try that (thanks). The current CPU fan is pretty quiet but
passive would be quieter still, consume less power and be more
reliable.

> The AMD XP-M processors are reputed to be
>particularly good at this, particularly the XP2500-M. The net's full of
>obsessive spotty geeks who have already done all the legwork for you
>down to the last milliwatt.

I'll re-Google then ... I found some generic stuff but nothing that
seemed (to me) close enough to be sure that it wouldn't all be a waste
of time and money.
>
>Power management in the latest releases of Linux seems to deal well with
>AMD chips.

Ok.
>
>>As an alternative to going very low power I've been playing with WOL
>>on this WHS (and the desktops) and had reasonable success, certainly
>>when it comes to waking it up (locally and remotely).
>
>I think, for a server, you'll find the latency (waiting while the server
>wakes up and restores the pre-hibernation state) too annoying.

I think the rate of 'annoying' would end up being tempered with the
size of the electric bill (already larger than we would like). Even if
we ended up with a low power solution I think I would still have it
power down if it was easy to do so and if there was nothing that
needed access 24/7.
>
>>Lastly, I've currently got the WHS in a cheapo mini tower case that
>>happens to have a reasonable layout with plenty of space for drives
>>and has a std PSU. I assume such a low power solution would (should)
>>have a matching low power '80 Plus' PSU?
>
>Nah. It's all a big marketing con, borne of idiots that look and talk
>funny. If you can, get a brand-name psu which claims that it meets the
>E-standard (80% or more), that'll do.

Ok thanks. ;-)

I did some W <> V/A measurements yesterday and the std PC PSUs were
quite a way off. A cheapo (included) 400W PSU on the Atom system was
drawing 41W and 69VA. Luckily we are billed on W. By the way of
calibration I checked the toaster and it was nearly identical W and VA
(nearly 100% resistive load). Again, I have read PSU's giving up to a
10W extra load and some drawing quite a bit when off (compared to
other PSU's on the same system).
>
>As soon as you realise that the PC industry is about endless
>upgradeitis, all becomes clear. New processor? You need a new mobo!
>etc.

I have been building systems since my first IBM-PC XT clone so know
what you mean. It used to make me smile overhearing sales folk in PC
shops telling the potential customer that '... and it can be upgraded
in the future with a faster CPU if required'. Technically they were
right but I don't know how people ever did so.
>
>What you need to remember is that more power (label on psu) is NOT what
>you want here! If the estimator you used says 120W (and remember that's
>running at full pelt, which you won't be), a suitable psu will be ca.
>150 - 200W to place it at the peak of the efficiency curve.

Yup, like I said back up there ^ somewhere. And that (again) doesn't
seem so easy, well not at the mainstream prices of the ~400W PSUs.[1]
>
>> An online-calculator gave me
>>a 120W PSU
>
>Ok, ish. Seems rather high (assuming we're talking "standard" PC
>hardware and you're aiming at low power.) You could borrow one of those
>plug-in power meters. (There's been extensive discussion of them here
>fairly recently),

I've been using my power meter (the portable one, the one displaying
the total house consumption fluctuates too much to accurately spot the
PC going on/ off) all along.

> but I feel that your effort would be better spent on
>concentrating on the lowest-power components you can put together for
>not much.

Hence the Atom Mobo I mentioned in my opening post? ;-) Also 'not
much' isn't the same as trying to do the best with what you have or
what is easily and cheaply available. Like, the AMD board and chip
were free but (currently, pre clock lowering experiments) draw about
60W. The Atom board I ordered thinking it would be lower power sits at
around 40W. Whilst that's ~£20/year saving it has other pros and cons
.... like the AMD board has 4 x SATAII an external eSATA (might be
handy for a backup drive) quiet CPU fan already and seems to be pretty
quick (when working on the server directly). On the other hand the
Atom board has Gb Ethernet (but only 2 SATA and has a noisy fan etc).
I'll stick a Gb card in and see what difference it makes to the power.

> 60W would be about right.

If I was to only have the 60W server on half the day it would still be
lower cost than the 40W one on all day.
>
>Johnny'll be along soon enough, he's done a lot of experimenting with
>low-power setups and will have an opinion.

He did <g> and it was good but no specific answers to my particular
questions / kit.
>
>For the purposes of your task, the Atom and the Sempron will be
>virtually the same. The Atom will run a lot cooler (it's designed to run
>with no heatsink), but there isn't much in the performance.

Yep, I'd have to say the Atom does feel a tiny bit slower but still
good enough.
>
>>[1] Can you actually get 'big, green and PATA' drives (cheap)?
>
>Yes, Look at the WD Green series. Getting good reviews atm.

Yeah, spotted them. Don't think they do a PATA model though?

FWIW, the 3 x 500G Samsung SATA drives I already have (HD502HJ) show:

Spin-up Current (Max.) 2.0A
Seek (typical) 5.8W
Read/Write (typical) 6.5W
Idle (typical) 5.0W
Standby (typical) 1W
Sleep (typical) 1W

Their 'Green' drive: HD502HI

Spin-up Current (Max.) 2.0A
Seek (typical) 4.8W
Read/Write (typical) 5.1W
Idle (typical) 3.9W
Standby (typical) 1W
Sleep (typical) 1W
>
>>[2] I've also got a bi-directional PATA<>SATA converter on the way
>
>Are you SURE about that? They usually only work one way.

These ones were advertised as 'Bi-directional' and if it's just a
physical interface converter (serial to parallel) there's no reason
why it couldn't be (I think there may be two SATA connectors,
depending what role you have it in).
>
>>with the thought that whilst it might add a bit more load
>
>Hardly anything. One chip and a few support components.

It all adds up. ;-)
>
>> (compared
>>with going without) it might cover the question above (giving me
>>another SATA port). But, what sort of performance degradation would I
>>see if it's only ATA 100 and SATA(1) please?
>
>Absolutely none.

Ok.
>
>>Oh, and can you generally set which port carries the boot
>>/ system drive as I believe WHS makes most use of the system drive (so
>>I would like it to be as quick as is practical).
>
>Odd question.

Really? I have 3 drive powers, one ATA100 and two SATA!!-3G. I
understand WHS uses the 'system drive' as an initial dumping ground
then migrates any duplicated data over to other drives (it's form of
mirroring). Therefore, I was thinking it made sense for this
particular drive to be the fastest, given the choice etc?

>Think about using a flash drive for the OS (64gb are cheap
>now) and use spinning media for your data drives.

Hmm, interesting.
>
>>p.s. An old P4 / 1.7 averaged about 56W and a PIII / 800 about 50 (but
>>neither had 'Green' drives).
>
>Nor processors with good power management.

But FWIW, both drawing less than this AMD (4 and 10W) and only 10W
more than the Atom.

I think if cost wasn't an issue it might be a Via board with 2 x 1TB
'green' laptop drives in a real ITX case and PSU. ;-)

Cheers, T i m

[1] In connection with my PC building history I knew a Co who had a
big factory in Tawan and they made their own PSUs. Probably at least
10 years ago it was costing them £1.50 to get a PSU on the shelf in
the UK.

p.s. For the Samsung 500G laptop drive HM500LI.


Voltage +5V±5%
Spin-up Current(Max.) 1000 mA
Seek(typical) 2.6 W
Read/Write(typical) 2.4 W
Low Power Idle(typical) 0.7 W
Standby(typical) 0.25 W
Sleep(typical) 0.2 W

The message <lpc2n5p2v68gkh65c6k7ppjlnt96sf8hfc>
from T i m <news> contains these words:

> On Tue, 9 Feb 2010 03:28:43 GMT, Johnny B Good
> <jcs.computersbutt> wrote:


> > Here's a posting I made sometime in February last year in this
> >newsgroup which I copied so I could repost it whenever this question
> >comes up (saves me a shed load of typing ;-). I think this will provide
> >you with some food for thought on this subject.

> Thanks.
> >
> <snip>
> >
> >> One advantage of the VIA boards is lack of fan, the Atom boards have
> >> a weeny whiney little fan on them and the one in my workstation is
> >> driving me nuts and an investigation might just be undertaken to see
> >> if it can be slowed down or even removed/replaced... (it's not on the
> >> processor but the bridge chip!)

> And this one goes through a rumbly stage as it spins up and is also
> only 3 wire so I might replace that anyway with something bigger, 4
> wire and held off the Mobo.
> >
> > That's interesting, I'm running a FreeNAS box with four Samsung HD103UJ
> >disks mounted and this takes just 54 to 55 watts (depending on drive
> >temperature - I don't spin them down).

> This seems to be a common value, the 50W range I mean. As you mention
> elsewhere you are in the 'diminishing returns' area when trying to get
> lower.

> <snip>

> > With each drive accounting for about 8 watts each in the power budget
> >when spinning, this leaves the remaining 22 or so watts accounted for by
> >the Asrock AliveNF7G-FULLHD R3.0 and an underclocked 1250 AMD Semperon
> >running at 800MHz with the core voltage dropped to 0.850 volts. I have a
> >single stick of 1GB DDR2 memory fitted (the benefit of dual channel is
> >rather lost in this application, as I proved by trying it with a couple
> >of 1GB dimms).

> I went for just 1 DIMM for similar reasons.
> >
> > Mind you, the original[1] 145 watt mini ATX psu fitted in this
> >Gateway2000 desktop case does contribute a good 5 to 10 watts
> >improvement over the typical standard 3 to 4 hundred watt ATX psu
> >designs generally available.

> And that's another thing. Whilst I didn't see much difference in the W
> between two (cheapo) PSU's there was quite a difference in the VA and
> I think we are billed on VA?

No, it's always real watts. Domestic customers are never seperately
billed for going over a set KVA limit since regulations on domestic
appliances take care of that problem (eg the requirement for PF caps in
fluorescent lamp fittings only applies if the lamp rating exceeds 20
watts). Of course, it's a different story as far as business customers
are concerned when their requirements demand a 3 phase supply to power
heavy duty electrical equipment.

> >
> > The Gateway case is especially good for this job since it is quite
> >sturdily made and had space to fit the extra two drives, using MoBo
> >stand off pillars, just behind the front panel where they could get
> >first sniff of the incoming air to keep them nice and cool (too cool as
> >it turns out) without the need to supplement the extremely quiet airflow
> >generated by the lone PSU fan.

> I try to think of the case thermo/aerodynamics like that, sometimes
> blocking off side and back vents to ensure an actual 'flow' of air
> through the case and not just round and round within.

Absolutely! The last thing you need are extra fans and vents to disrupt
a nice "Crossflow" of air through the system box. It also helps if the
drives are first to be blessed by the incoming airflow before it gets
heat contaminated by the cpu heatsink exhaust.

I recently had to work on one of those miniature Dell Towers, a
Dimension 5150C, and noticed (just as I'd expected) the extremely high
drive temperature as a result of the front cooling fan blowing air
through the BTX style CPU heatsink onto the hard drive immediately
between it and the back panel. The drive had reached 61 degrees C, one
degree in excess of its maximum rating.

I'd already disassembled the fan cooler to vacuum it out so I decided
to refit it so as to make it exhaust through the front panel so the hard
disk would then recieve incoming cooling air from the generous venting
on the back panel instead of being cooked by the CPU's exhaust heat.
This dropped its temperature to 49 deg C, not particularly cool, but
still a good 12 degrees cooler!

I have to conclude that this situation was either an error on the
assembly line or else a very serious design flaw. You do have to wonder
at the returns rate on this model, if it were the latter situation.

> >
> > For file serving / print serving, you need very little processor power,
> >so you can safely underclock to the minimum speed possible (but be wary
> >of departing from standard minimum FSB clocks though since this could
> >effect I/O performance - HDD and Gbit ethernet).

> I don't think I'd go as far as playing with the clocking but might
> look for a more suitable chip (with the Quiet n Cool feature etc).
> >
> <snip>
> >
> > When I was choosing a processor and MoBo combination for the NAS
> >upgrade, I was looking for the cheapest combo that offered a minimum of
> >4 sata ports with built in Gb ethernet port and took 'pot luck' over the
> >cmos options.

> It's funny isn't it, considering how many manufacturers and boards
> there are out there at the moment how quickly your choices become
> limited if you have a specific requirement (like low power, 4 x SATA,
> GB and cheap). My Sempron setup has the 4 x SATA and was cheap (I just
> bought the RAM) but only has 100M Ethernet. Now, a couple of my PC's
> have GB, as is my switch but I'm not sure how much difference GB makes
> in the real world (I think I read no where near as much as 10X
> better). However, I guess twice as fast could mean a backup takes half
> as long so could be worth it?

I know what you mean. It seems to be a Herculean task to persuade Gbit
ethernet to run at 50% of its potential speed. 30 to 40 percent seems to
be the more usual figure with all but bleeding edge kit. However, since
the problem is down to the connected PCs failing to live up to
expectations, this just means you're unlikely to suffer bottlenecking
with the Gbit switches you may be using, leaving you with capacity to
spare even if you have two or even three pairs of PCs transferring data.

As you say, even if you only see a doubling of transfer speed compared
to fast ethernet (100Mbps), it's a noticable improvement. I'm running a
3 year old Acer Aspire 3660 Notebook 24/7 simply to serve as a Freeview
PVR to resolve scheduling conflicts and I decided to fit a Gbit ethernet
card to improve on the built in fast ethernet performance to speed up
the transfers to my desktop PC where I top and tail and slice and dice
the captured TV programs.

I didn't even see a doubling, it was more like a 60 or 70 percent boost
but at least it was still a noticeable improvement (from just over half
a GB a minute to almost a GB per minute.

> <snip>

> > Once you go for an instant response 4 drive setup, the marginal power
> >saving between an especially low power consumption MoBo and an
> >undervolted, underclocked mainstream one becomes much less important.

> Good point.
> >
> > If I placed more value on the saving of some 30 odd quid a year on the
> >electricity bill over the integrity of and instant access to my data, I
> >could employ spindown power saving to the tune of some 30 watts. Since I
> >don't, that extra 5 or 10 watts used by my choice of MoBo over a
> >specialised low power one, becomes of little significance.

> True. Outside the W <> VA question I'm looking at ~45W(Atom) V 60W
> (AMD) with much more versatility in the AMD setup. However, I'll have
> to measure it again with a GB card to be sure ad if my up-time_needs
> don't warrant it being on 24/7 then the *difference* matters even
> less. (It *was* sometimes awkward that my hardware NAS wasn't on 24/7
> but because it wasn't big or flexible enough to do anything else ...
> it didn't need to be).

> I've got a 120W PSU in the old Flex PIII system I was playing with
> before and I might try the Atom board in there with the thoughts of it
> being a workstation to replace the Mini (then it wouldn't matter that
> it only had one 3.5" hdd location).
> >

> > The only other snag with this Gateway case is that the I/O shield area
> >isn't the standard letterbox cutout, the i/o holes have been directly
> >punched out of the rear panel to accomodate the original MoBo.

> I cut down a wireless card to put in the (Gigabyte) slimline Flex case
> and the Atom board has a single PCI slot. ;-)
> >
> > You'll either have to hack out the existing holes or create a standard
> >I/O shield letterbox aperture to accomodate a standard shield plate to
> >allow a replacement MoBo to be installed. Not a showstopper, just a
> >trivial impediment to upgrading the MoBo for someone prepared, if needs
> >be, to spend a fiver in Aldi on a pair of 'Tinsnips' ;-)

> I have a pair of nibblers that are brilliant for opening up fan holes
> (I normally cut the perforated fan positions out and replace hem with
> finger guards). The head of the cutter goes through one hole and you
> just chomp your way round, finishing off with a half round file. No
> chance of distortion but every chance of sending a cutting into yer
> PC's so do it elsewhere. ;-)

If my 'handiwork' is "Out of Sight"(tm), I don't bother cleaning up the
jagged edges of such a 'perforectomy'. Although I'm normally a bit of a
perfectionist with this sort of handiwork, it was enough to have removed
a _major_ fan intake obstruction and I left the hole with jagged edges
simply to make a statement[1]

> Mine are like this:

> [..]

That's a handy tool, nice! I bet the hole it produced looked a lot less
ragged than the one I've got hidden behind the front panel of my win2k
box. ;-)

> <snip>

> > I do have a 270 watt micro ATX PSU of the type intended for those SFF
> >cases which I could have fitted with very little modification to the
> >case, but, although its efficiency was noticably better than the usual
> >commodity ATX units, it wasn't quite as good as the original (but I keep
> >it in the "Server Spares" box just in case the original finally pops its
> >clogs and heads off for 'Silicon Heaven'(tm Red Dwarf) ;-)

> Can you easily / cheaply get std format PSU's in low power and high
> efficiency (80 Plus?) ratings? I haven't seen many?

I don't know. I've never tried to track down such a source of PSUs.
It's just that whenever I tested any prospective motherboards using
whatever standard ATX 400 watt PSU I have on my work bench to power test
setups, they seemed to use about ten watts more than when they were
actually fitted into that Gateway case with its original SFF 145W PSU.

> >

> >EndQuote:
> >
> > I've since replaced two of those 7200rpm drives with a couple of
> >Samsung's eco green 1TB units which seem to, as best as I can judge,
> >save a couple of watts apiece on the standard drives.

> I was thinking of doing similar if I stick with the Atom. For
> cheapness I have both setups in small/midi ATX (£22) cases that happen
> to have loads of drive bays. Seeing all the space around the Mini-ITX
> board and knowing I might be able to find a small / PSU and only
> needing room for two HDD's that it wouldn't take much to make a wooden
> case (probably using the metalwork from a case for the actual
> mountings) but wondered about RFI? Mind you, many cases have plastic
> front panels (but could be RFI treated) but what of the Perspex boxes?

> > I've also upped
> >the cpu clock multiplier a little to improve the Gbit throughput (6
> >instead of 4 for a 1.2GHz clock speed, afaicr - just over half rated
> >speed with a slight Vcore increase to, afaicr, 0.9v) and fitted a second
> >1GB ddr2 dimm to enable dual channel memory mode.

> I'm still interested in exploring a chip that supports Cool n Quiet
> and or a Mobile CPU.
> >
> > The net effect of all those changes means the server now consumes some
> >53 watts. I estimate that if all the drives were the eco green types,
> >this would probably drop to just below the 50 watt mark, which, for an
> >always available 4TB server, is pretty economic.

> Yup, I think that's about the best bang / watt.
> >
> > Write speeds, when the source files on this nearly 6 year old win2k box
> >aren't fragmented, just exceeds 2GB per minute (about 35MB/s) and read
> >speeds can exceed this by a small margin. I think a more up to date
> >system box would show even better results since I suspect I'm
> >encroaching on the Gbit speed limit of the win2k box but I don't often
> >see any decent spec systems _with_ Gbit ethernet included to test this
> >out.

> The think is, like your pot_luck to the CMOS settings thing I think if
> your focus has been best VFM / ease / low energy for max capacity then
> to specifically try to put network performance into the equation
> (beyond insuring it's working reasonably etc) might be a factor too
> far.

TBH, I was rather hoping that a board with built-in GBit would be able
to offer a significant improvement over one that was then some five
years old and relying on a netgear PCI adapter. The thing is, at the
time, the version of FreeNAS I was running didn't support the on-board
NIC so I had to disable it and plug another Netgear PCI adapter into a
PCI slot. The performance was around the 25 to 30MB/s mark with that
setup and enabling jumbo frames didn't help very much.

When the next version release of FreeNAS became available, it looked,
as best as I could figure, like it had support for the on-board NIC.
This proved to be the case but I saw very little improvement and there
seemed to be a bug in enabling jumbo frames (it simply wouldn't!). In
the end, disabling jumbo frames on my win2k box gave me the best
performance boost. A little bit more tweaking of CPU speed gave me yet
more speed, enough to the point that I was able to just exceed the magic
2GB per minute transfer rate I'm currently achieving.

As I've already said, I rather suspect that it's this almost 6 year old
win2k box that's bottlenecking the transfers rather than the FreeNAS
box. Since I'm currently working on a more recent system box that is
blessed by on-board Gbit (but had been cursed by a rather nasty trojan),
I'll be able to take advantage of its availability RSN to test whether
there is any foundation to my suspicion.

I've just tested that "No Name" Box, 2GHz clocked Athlon64 with 1GB ram
in a recent A-Open MoBo and a 160GB IDE hard drive with a fresh install
of winXP SP3, and the best read transfer speed just failed to beat the
worst rate of this rather venerable win2k box (V600DAP MoBo and a Barton
clocked at 1.83GHz). The write transfer speed on the winXP box was a
lousy quarter of its read speed (the win2k box gets very similar speeds
in _both_ directions).

Rather mixed results from the more recent box with built-in Gbit
ethernet. The problem is, I don't know how much of this is down to
winXP's deficiencies, the hard drive being IDE rather than SATA, or some
other factor (seems unlikely to be a CAT5/6 cabling problem somehow, but
I can eliminate this with another test, not right now though, it _is_
rather late, even for me!).

Surprisingly, it seems my win2k box is excelling itself, in spite of
its age! Ah well, this Gbit ethernet is a slippery beast to tame. I
recall seeing up to 40MB/s transfers when I connected this win2k box
directly to a 3GHz P4 'Power User' box (winXP) a customer had brought in
for repair a few years back so it pains me to still see such wild
variations in Gbit performance. Perhaps it speaks volumes for A-Open's
design qualities in this instance?

I guess I'll have to bide my time and repeat the test with any other
suitable PCs that come in for repair before I can make any further
conclusions.


> Cheers and thanks for the feedback. ;-)

> T i m

You're welcome!

[1] The statement being...

"Look, you Bastards! When it's out of sight, just use a fan shaped hole
and don't _EVER_ use perforations, not even when there is a risk of
personal injury to someone daft enough to poke their fingers into where
they don't belong! AND, if you must protect the stupid, use a wire
guard!"

Those cooling fans can shift considerable amounts of air if they don't
have to fight restrictions on intake and exhaust. IOW, _avoid_ any
unnecessary restrictions if you want a quiet and _effective_ cooling
solution. Unfortunately, most case designs fail to offer such
unrestricted ventillation paths and the cooling performance suffers as a
result (still, there's always the tinsnips and 10mm drill bit 'solution'
to remedy this ;-).