Atom schmatom; folks with a KillaWatt, how low can you get your system idle and load?

[dug777]

Load for gaming was about 115W (1.1vcore and EPU-6 running).

Saw some numbers for Atom machines that didn't seem THAT much better for idle, given how piss-weak it is.

Suspect someone with a 45nm C2D or an equivalent AMD chip and an IGP will win this one easily

Now I've replaced the 3450 with a 4850 my idle and light use (office/net/email/winamp) is around 110W, and my gaming load is around 230W

Of course, I can now play games

 

[aigomorla]

115W light idle lol...
wow thats sipping electricity...
My sammy is around 80-100W at full load..
http://i125.photobucket.com/al...aigomorla/IMG_1112.jpg

Its the only non watercooled computer i have at my disposal.
http://i125.photobucket.com/al...aigomorla/IMG_0013.jpg

Eva draws about 300W idle. :rofl:

lets not talk about load..

ugh i have a nasty hangover..

 

[frostedflakes]

My X2 3600+ w/6100 IGP would idle at about 41w with CnQ. I have a 7600GT in there now, though. Haven't bothered measuring idle power consumption with it installed, but it's probably closer to 60w.

I had an old Sempron64 system that I got down to 25w idle.

 

[pm]

My HTPC - which is a Pentium M 740, i915 chipset, 2GB RAM and a Radeon 4350 (and Lite-on Blu-ray, 1TB HD, ATI T Wonder) - idles at ~52W. With the hard drive powered-down, it's ~49W. With the video card in power-down mode, it's 47W. The Atom system that I had idled at 54W. At measurements are approximate - my kill-a-watt never seems to peg at a specific number, but even idling moves around a watt or so.

 

[dug777]

Thanks folks, keep them coming

Somewhat confirms my view that Atom is utterly pointless in a desktop, not exactly a particularly enlightening or astounding view I know

 

[ajaidevsingh]

My home Wifi server Pentium 3 "Under clocked Coppermine" ideals at 41W. I am useing a laptop 2.5" hdd and other than that it does not have extra video card, etc.

 

[hans007]

my amd 780g based , phenom 9550 machine is around 65 watts at idle.

thats with an earthwatts 500 using integrated video, with a sseagate 1.5TB 7200.11 and 4gb of crucial and 1 case fan.


the atom platform still sucks b ecause of the chipset and because much faster cpus realy dont use that much power anyway.

for example, a wolfdale 3M core like the e7200 at idle, is only using 3 watts. the atom at load uses 4 watts, and at idle lets say it uses .1 watts. even then its only a 2.9 watt difference at idle.

load an e7200 uses something like 18 watts according to xbit labs. i mean i have an atom netbook and its still plenty fine, but the e7200 can handle a lot more higher end computing for barely more power use especially when combined with the horrible chipset the atom uses compared to say an e7200 and a modern 65nm chipset like the geforce 9300 or g4x.

the atom might be ok once they put the chipset on the die when it goes to 32nm at least it will have a big enough power advantage at that point.

 

[LOUISSSSS]

yea i think the ATOM cpu is nothing special. ITS SO DAMN SLOW.

my Lenovo X200 laptop draws about 27 watts from the wall with normal usage (downloading, max brightness, 2-3 usb devices (cooler, mouse, flash drive), and playing music)
if you're REALLY trying to skimp out on some power draw between 15 and 27 watts you shouldn't be buying any computers.
if you're in it for "low temps and long battery," all laptops are made to work within spec (under ~72C.) my laptop idles at 35c and loads at about 45. 6 cell battery gives me about 5 hours of battery life on normal usage. 6 hours on conservative.

heres my laptop specs:
Lenovo X200
12.1in screen
Intel P8600 (2.4ghz)
4gb DDR3
Intel X4500 Chipset
5400rpm hdd
6 cell
about 4lbs w/ battery

 

[nyker96]

i use a E2140 M0 @3.1 //ip35-e//6200LE pcie card//1TB WD black drive idle about 65W, loaded about 110W.this is with speedstep.

 

[SlowSpyder]

I do not think my overclocked/overvolted Phenom and overclocked 4870 would do well in this competition.

Impressive numbers you guys are getting! Keep them coming, I like seeing them. One of these days I'll get a Kill-a-watt to see how component changes affect power, it intriques me.

 

[razor2025]

Originally posted by: LOUISSSSS
yea i think the ATOM cpu is nothing special. ITS SO DAMN SLOW.

That's because it's paired with crappy Northbridge/IGP. If Intel released it with low powered / mobile version of G45, it's draw would probably be ~25-30 watt. The Atom chip alone only needs 4-6 w.

 

[Liberator21]

So, how much would a maxed out power-hungry computer actually use at load? I'm talking a Skulltrail, 4 mid-range cards, couple hard drives, overclocked, etc.?

 

[Soulkeeper]

i can idle with 60Watts total system usage on my c2d e4300 OVERCLOCKED to 3GHz
it pulls around 120Watts under full load tho

this is with an matx motherboard, integrated intel graphics and ony 512MB of ddr2

 

[Gonad the Barbarian]

Ugh, I just checked my HTPC with an X2 4200 & 6800GT and got 104W at idle w/ CnQ enabled.

 

[dug777]

I could get my Athlon XP machine with a 6800 GS, 1.5GB DDR, 250GB WD, and just one case fan down to high 60W/low 70W idle when I downclocked it to 666Mhz, but the lowest vcore option for my A7V8X is 1.6v

Using s2kctl (a bus halt program) shaved about 10W off the previous idle figure.

 

[Zap]

A couple things...

The Kill-A-Watt is not 100% accurate. It has been shown that these and other consumer level (and even some professional level) power meters can be fooled by APFC into reporting an inaccurate reading. Most of the time inaccuracies result in it reporting close to the power being drawn. For instance, an 80W load would read as 80W instead of 100W (assuming 80% efficiency).

I believe SPCR showed that a 45nm C2D on a G45 chipset board used around as little power idling as a mini ITX Atom board. The "problem" with the Atom is the archaic 945GC chipset board that uses around 30-35W by itself. On the flip side, the mini ITX board runs around $75 and that's including both board and CPU. Cheapest 45nm chip is E5200 which runs more, plus the motherboard which will at least double the cost.

For some perspective my undervolted mobile P4 (on a desktop board) server idles around 75-80W with two HDDs and no 3D video card (Tseng Labs ET6000). I think the performance is probably little better than the Atom. I think I once measured someone's Prescott core P4 3.4GHz with an 8600 GTS video card and 4 HDDs at around 240W idle. I'm sure the same parts with a higher performance E5200 chip would idle at less than half.

I've often daydreamed about making my server perform better (not that I need more performance) and draw less power by replacing it with a G45 chipset board and E5200, and a Pico PSU. Underclock/undervolt the CPU and with a WD "green" HDD I think the idle would be around 35W and the load would be around 50W. Maybe.

 

[edplayer]

Originally posted by: Zap
A couple things...

The Kill-A-Watt is not 100% accurate. It has been shown that these and other consumer level (and even some professional level) power meters can be fooled by APFC into reporting an inaccurate reading. Most of the time inaccuracies result in it reporting close to the power being drawn. For instance, an 80W load would read as 80W instead of 100W (assuming 80% efficiency).

Can you post some more info on this?

 

[Zap]

Originally posted by: edplayer

Originally posted by: Zap
The Kill-A-Watt is not 100% accurate.

Can you post some more info on this?

here's a start

 

[dug777]

Originally posted by: Zap
A couple things...

The Kill-A-Watt is not 100% accurate. It has been shown that these and other consumer level (and even some professional level) power meters can be fooled by APFC into reporting an inaccurate reading. Most of the time inaccuracies result in it reporting close to the power being drawn. For instance, an 80W load would read as 80W instead of 100W (assuming 80% efficiency).

I believe SPCR showed that a 45nm C2D on a G45 chipset board used around as little power idling as a mini ITX Atom board. The "problem" with the Atom is the archaic 945GC chipset board that uses around 30-35W by itself. On the flip side, the mini ITX board runs around $75 and that's including both board and CPU. Cheapest 45nm chip is E5200 which runs more, plus the motherboard which will at least double the cost.

For some perspective my undervolted mobile P4 (on a desktop board) server idles around 75-80W with two HDDs and no 3D video card (Tseng Labs ET6000). I think the performance is probably little better than the Atom. I think I once measured someone's Prescott core P4 3.4GHz with an 8600 GTS video card and 4 HDDs at around 240W idle. I'm sure the same parts with a higher performance E5200 chip would idle at less than half.

I've often daydreamed about making my server perform better (not that I need more performance) and draw less power by replacing it with a G45 chipset board and E5200, and a Pico PSU. Underclock/undervolt the CPU and with a WD "green" HDD I think the idle would be around 35W and the load would be around 50W. Maybe.

I'm not actually using a KillaWatt, using a Power-Mate I borrowed from work.

Interesting stuff nonetheless, and I have no doubt it applies to this too.

 

[Acanthus]

The problem with atom isnt really atom itself. It is a fantastic design for its use. The atom however doesnt have a "low power" chipset or IGP to truely knock it down below undervolted machines.

 

[PingSpike]

Seems kind of pointless to stick an ultra low power processor in an ancient power sucking chipset.

 

[Zap]

Originally posted by: Acanthus
The problem with atom isnt really atom itself. It is a fantastic design for its use. The atom however doesnt have a "low power" chipset or IGP to truely knock it down below undervolted machines.

It does, but a couple things... first I'm not sure it is available yet. Second, it is really stripped down.

Poulsbo: An Unusually Revolutionary Chipset

Instead of re-using a desktop chipset, Intel architected Poulsbo from scratch. In order to conserve power things like SATA and some of the USB ports were ripped out of the design.
...
Only 2 channel audio is supported
...
In a highly atypical move for Intel, Poulsbo also supports non-standard 1.5V DDR2 memory.
...
Poulsbo requires roughly half the power of mainstream chipsets.

 

[Acanthus]

Originally posted by: Zap

Originally posted by: Acanthus
The problem with atom isnt really atom itself. It is a fantastic design for its use. The atom however doesnt have a "low power" chipset or IGP to truely knock it down below undervolted machines.

It does, but a couple things... first I'm not sure it is available yet. Second, it is really stripped down.

Poulsbo: An Unusually Revolutionary Chipset

Instead of re-using a desktop chipset, Intel architected Poulsbo from scratch. In order to conserve power things like SATA and some of the USB ports were ripped out of the design.
...
Only 2 channel audio is supported
...
In a highly atypical move for Intel, Poulsbo also supports non-standard 1.5V DDR2 memory.
...
Poulsbo requires roughly half the power of mainstream chipsets.

I dont believe it is out yet.

 

[imported_harlanpepper]

A mATX build I did a few months ago (E3110 at stock 3GHz, 4GB ram, old 7800GT card, 640gb drive, 1 DVD) consumed around 80w at idle and somewhere around 110w under load. Never measured it using the IGP (g35 P5E-VM HDMI).

 

[Idontcare]

Originally posted by: Zap

Originally posted by: edplayer

Originally posted by: Zap
The Kill-A-Watt is not 100% accurate.

Can you post some more info on this?

here's a start

FWIW I checked my Kill-A-Watt readings with those my power company's meter said I was pulling thru the meter, both readings were within a watt of each other.

Guess I just got a good kill-a-watt, or my power company uses craptastic metering themselves. Either way my kill-a-watt data reflects well what I end up paying for, so I find it value-add to make measurements with it.

In any event these kill-a-watts are far more likely to be accurate than the DTS based CPU temperatures everyone likes to report (and go into crisis mode over) around here.