Lowest power dual-core

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

rredman
Posts: 2
Joined: Wed Apr 16, 2008 9:14 pm
Location: WA

Lowest power dual-core

Post by rredman » Wed Apr 16, 2008 9:58 pm

AMD CPU
It could possibly be this one
- Athlon 64 X2 3800+ "Brisbane" Energy Efficient 2Ghz 35W 90nm
but you can't buy it anywhere, so it's got to be one of these
- Athlon X2 BE-2300 "Brisbane" 1.9Ghz 45W 65nm
- Sempron X2 2100 1.8Ghz 65W

Hard to find info on the Sempron X2 2100, I guess since it is so new. So, which one has the lowest power consumption, anyone know for sure?

AMD Motherboard
Which board draws the least power. Is there a definitive answer? It's really hard to get info on this since everyone has such different setups. Almost everyone uses a different motherboard it seems. I don't really have any "requirements" just integrated video, a few sata ports, and gigabit ethernet.

Intel CPU
Would have to be either of these
- Intel Pentium E2140 1.6Ghz 1MB cache 65W
- Intel Celeron E1200 1.6Ghz, 512KB cache 65W
All I could find was the xbitlabs shootout that compares E2160 with E1200, with just 1W diff at idle and 2.6W at load.

Intel Motherboards
Same problem as for AMD motherboards.

yuu
Posts: 132
Joined: Sun Aug 26, 2007 5:03 pm
Location: eu

Post by yuu » Thu Apr 17, 2008 2:42 am

http://img373.imageshack.us/img373/4884 ... ardak6.png

i have p31-ds3l and 2140 with undervolting cpu alone 20W are saved, so if it is 30W under load, think half of that.

idle is 100W with 8800gt

it is known p31-ds3l to have very low power usage, and the new EP31 is energy saver

http://www.thetechlounge.com/article/50 ... therboard/

winguy
Posts: 447
Joined: Tue Nov 02, 2004 9:31 am

Post by winguy » Thu Apr 17, 2008 4:09 am

I think the new AMD 4850e , 4450e and 4050e are the lowest power mainstream CPUs from AMD at the moment.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Thu Apr 17, 2008 5:05 am

yuu wrote:http://img373.imageshack.us/img373/4884 ... ardak6.png

i have p31-ds3l and 2140 with undervolting cpu alone 20W are saved, so if it is 30W under load, think half of that.

idle is 100W with 8800gt

it is known p31-ds3l to have very low power usage, and the new EP31 is energy saver

http://www.thetechlounge.com/article/50 ... therboard/
With an AMD X2 4000+ undervolted at stock speed, My system with 2 HDDs and an HD3870 Idles ~77W. I imagine the new E8xxx series undervolted would show some ridiculously low power usage.

ST
Posts: 67
Joined: Tue Apr 11, 2006 11:18 pm
Location: Silicon Valley

Post by ST » Thu Apr 17, 2008 8:55 am

You'll be surprised how minute the differences are between AMD's EE (BE) offerings and Intel's more powerful solutions:

http://techreport.com/articles.x/10508/1

Image

Image

This is not even accounting for the newest E7200 45nm Core 2 Duo's that will show up soon...

nutball
*Lifetime Patron*
Posts: 1304
Joined: Thu Apr 10, 2003 7:16 am
Location: en.gb.uk

Post by nutball » Thu Apr 17, 2008 10:22 am

ST wrote:You'll be surprised how minute the differences are between AMD's EE (BE) offerings and Intel's more powerful solutions:
What surprises me is that anyone could regard a system with an idle load of >100W as saying anything meaningful about which dual-core processor is the lowest power.

30W full-system idle, you have some evidence. Anything else is noise.

smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Post by smilingcrow » Thu Apr 17, 2008 3:12 pm

nutball wrote:What surprises me is that anyone could regard a system with an idle load of >100W as saying anything meaningful about which dual-core processor is the lowest power. 30W full-system idle, you have some evidence. Anything else is noise.
Since you talked specifically about processor power consumption I could argue it the other way and say that by using a platform with a high power consumption at idle it puts the system into a relatively flat spot of the power supply’s efficiency curve so that measurements aren’t skewed by this variable.
This way when you swap out multiple CPUs on the same platform more of the power difference is purely down to the CPU.

rredman
Posts: 2
Joined: Wed Apr 16, 2008 9:14 pm
Location: WA

Post by rredman » Thu Apr 17, 2008 8:34 pm

winguy wrote:I think the new AMD 4850e , 4450e and 4050e are the lowest power mainstream CPUs from AMD at the moment.
Show me where I can buy those, and then we can talk about them again. :)

I think nutball was really referring to the fact that the graphs ST quoted are all > 100W, and that doesn't help me one bit. I said LOW. 100W is not low.

Yuu says p31-ds3l and 2140 undervolted will be < 30W.
Anybody have anything else?

yuu
Posts: 132
Joined: Sun Aug 26, 2007 5:03 pm
Location: eu

Post by yuu » Fri Apr 18, 2008 12:45 am

in fact i'm wrong, p31 is not int. video, or it is disablesd, but g31

that one g31 is just like p31-ds3l, but lacks all solid caps ultra durable
http://www.gigabyte.com.tw/Products/Mot ... uctID=2725

frostedflakes
Posts: 1608
Joined: Tue Jan 04, 2005 4:02 pm
Location: United States

Post by frostedflakes » Fri Apr 18, 2008 1:48 am

rredman wrote:
winguy wrote:I think the new AMD 4850e , 4450e and 4050e are the lowest power mainstream CPUs from AMD at the moment.
Show me where I can buy those, and then we can talk about them again. :)

I think nutball was really referring to the fact that the graphs ST quoted are all > 100W, and that doesn't help me one bit. I said LOW. 100W is not low.

Yuu says p31-ds3l and 2140 undervolted will be < 30W.
Anybody have anything else?
The graph is fine for comparing relative power consumption, which is what you're ultimately interested in. And as smilingcrow pointed out, at these loads the power supply's efficiency curve is flatter, meaning more comparable results. For example, if a power supply is 60% efficient at 40w load and 80% efficient at 60w load, the AC power measurements for a 40w and 60w system would be 67w and 75w, respectively. This suggests only 8w difference in power consumption, when in actuality it's much larger. This is a pretty extreme example, but some power supplies do have very steep efficiency curves at low loads.

Anyways, what do you plan to do with the system? I don't know how the newer 45nm CPUs do, they are probably much better in regards to idle power consumption, but the general trend has been that an AMD will have lower idle power consumption with CnQ enabled, but higher power consumption at load. And Intel has a pretty significant advantage in performance per watt.

So if you are going for quiet and high-performance, and have the money to spare, definitely grab an Intel. For a basic system, though, you might as well save some money and stick with AMD. I went with them because, at the time I built my system, Intel dual-core chips were twice the price of the Brisbane and I didn't need the extra performance they offered.

Even with undervolting they probably won't be able to touch the performance per watt of the Intel cores, but you can achieve a very efficient system with Brisbane. My 3600+, for example, can undervolt to 0.975V at 1.9GHz, which is a roughly 50% drop in power consumption (probably makes it about a 25-30w CPU). To be honest I wouldn't even bother with the low-power chips, not worth the price premium IMO. Just grab a desktop Brisbane and undervolt/underclock it to achieve your desired power envelope. In theory Sempron should have lower power consumption because of less cache (fewer transistors). In practice I doubt this translates to a significant power savings, maybe a few watts at most.

I may have them posted on the forums, so you can try searching, but off the top of my head I can't remember what my load power consumption was with the undervolted Brisbane. However, idle power consumption was 41w I believe (before I got the 7600GT) with CnQ enabled.

najames
Posts: 20
Joined: Mon Feb 27, 2006 7:44 pm
Location: Florida

Post by najames » Sat Apr 19, 2008 6:48 pm

[[/quote]Show me where I can buy those, and then we can talk about them again. :)

http://www.newegg.com/Product/Product.a ... &Tpk=4450e

astrayan
Posts: 45
Joined: Fri Jun 17, 2005 5:54 am
Location: astraya

Post by astrayan » Tue Apr 22, 2008 6:33 am

frostedflakes wrote:This suggests only 8w difference in power consumption, when in actuality it's much larger. This is a pretty extreme example, but some power supplies do have very steep efficiency curves at low loads..
How do we know if most of the inefficiency of the PSU is not just a wattage offset, caused by a fixed overhead? Then, the only quandry would be what efficiency to multiply by, to get your figure. (Remember that they are bad at documenting the curve below 20% load.)

My Seasonic PSU registers 14W with the computer mobo "off", which seems to imply that about 11W is pissed away as an offset. In this case, the flatter part of the curve is where resistive effects are starting to win. This would mean that on the flat part of the curve, if you measure +50W, you are really only adding 45W, but on the steep bit, +50W might be 48W.


I'm getting a bit peeved with the boast that Intel has designed a superior chip to the Athlon, and that Core 2 45nm performs so well on a joules per instruction basis, and then Intel don't let anybody undervolt to 0.8V. It's like congratulating yourself for buying a hybrid car, and then towing a caravan everywhere. Until Intel has Athlon-style undervolting, they are not really in the race, are they?

yuu
Posts: 132
Joined: Sun Aug 26, 2007 5:03 pm
Location: eu

Post by yuu » Tue Apr 22, 2008 9:00 am

i dont get it, my Conroe cpu undervolts to 0.825 at least, thats without the droop of 0.5, really it is 0.775V, that plenty. any lower is not stable anyway

Image

astrayan
Posts: 45
Joined: Fri Jun 17, 2005 5:54 am
Location: astraya

Post by astrayan » Wed Apr 23, 2008 11:17 pm

OK, so how did you do that undervolt? Is that all software controlled, or part BIOS magic?

My understanding is that there is no software "VID" control for Duos below 0.95V - 1.15V, so lower voltages have to be kludged using static BIOS configurations. The lowest software voltage is dependent on various whims of the motherboard, and core version.

yuu
Posts: 132
Joined: Sun Aug 26, 2007 5:03 pm
Location: eu

Post by yuu » Thu Apr 24, 2008 7:43 am

bios, or gigabyte's easytune5 software can touch voltage, but software only lasts till restart, the bios is better way to do it.

this mbd is suitable for undervolting tests. it can go down to 0.5v but it refuses to boot lower than 0.825 with E2000. with restart can push down to 0.80. maybe i can do lower in windows since it is booting and ramping up the voltage issue

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Post by frank2003 » Thu Apr 24, 2008 9:18 am

I posted my configs in another thread here: viewtopic.php?t=31045&start=30

Here are the specs, reposted:

mobo: Biostar TA690G
CPU: AMD X2 Brisbane 4000+ underclocked to 1GHZ @ 0.784V (per CPUZ) idle; 1.4GHz @ .832V under full load.
Memory: 2x1GB, 800MHz
Disk: 3.5" WD 1TB, GreenPower model
PSU: PicoPSU-120, 80W brick
fans: stock CPU fan + 1x60mm case fan

Power Consumption: 29-30W idle; 44-48W playing 1080i HDTV recordings.

Note the components are all readily available desktop components (i.e. there's no cheating by using notebook parts :-)). Part of the credit goes to the use of the GP version of the drive. which consumes a few watts less than regular 3.5" drives. If I wanted to go overboard I could go passive cooling on the CPU and shave off another watt or so used by the fan. I could also reduce yet another 2 watts or so if I didn't use the NIC (but that's impractical since these days a standalone PC is pretty much useless).

I understand a mobo based on the nvidia 7050 chipset uses even less power. Since I don't own one I can't confirm if this is true.

Finally, the Biostar TA690G supports CPU voltage and multiplier controls in BIOS, for those who consider this feature important.

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Post by frank2003 » Thu Apr 24, 2008 10:07 am

frank2003 wrote:Part of the credit goes to the use of the GP version of the drive. which consumes a few watts less than regular 3.5" drives.
Correction: The major contributing factor the lower power consumption figures is the use of picoPSU which lowers the numbers by 25W-30W compared to a standard ATX PSU.

I've been taking pico for granted for a little too long...

drees
Posts: 157
Joined: Thu Jul 13, 2006 10:59 pm

Post by drees » Thu Apr 24, 2008 10:11 am

frank2003 wrote:Correction: The major contributing factor the lower power consumption figures is the use of picoPSU which lowers the numbers by 25W-30W compared to a standard ATX PSU.
I replaced my Seasonic Super Tornado 300 which isn't exactly efficient under 50 watts with a PicoPSU and power consumption in my system dropped by 8 watts at idle (from 50w to 42w) and 6 watts at load - so unless you have a really crappy PSU I think saying that a PicoPSU reduces power consumption by 25-30w is probably a bit generous.

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Post by frank2003 » Thu Apr 24, 2008 3:29 pm

My figures were based on my experience with 4 ATX PSUs. Maybe it was just my luck.

drees
Posts: 157
Joined: Thu Jul 13, 2006 10:59 pm

Post by drees » Fri Apr 25, 2008 12:08 am

frank2003 wrote:My figures were based on my experience with 4 ATX PSUs. Maybe it was just my luck.
Maybe my Super Tornado is just more efficient that I thought it was?

What were the four PSUs you replaced?

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Post by frank2003 » Fri Apr 25, 2008 2:29 am

I tried Antec TP-II 430W retail, Antec TP 380 (came with the original Sonata case), Q System 300W (bought from quietpcusa.com), and Enlight 180W (came with a cheap $15 Enlight 7396AM1 case).

astrayan
Posts: 45
Joined: Fri Jun 17, 2005 5:54 am
Location: astraya

Post by astrayan » Fri Apr 25, 2008 4:43 am

I can believe that a picoPSU would save 10-12W over a crappy PSU, but one thing you have to watch is that these watt meters aren't all that shit hot. They will give different accuracy readings on different PSUs. I think they pick up on resonance spikes on the wave, and this varies depending on whether the PSU has good filtering and power factor correction. When you look at the current/amps input on an oscilloscope, you will see what crap the input wave actually is, and you will marvel that a digital thingo can do a real-time integration of V x A, and come up with any figure.

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Post by frank2003 » Fri Apr 25, 2008 10:10 am

astrayan wrote:I can believe that a picoPSU would save 10-12W over a crappy PSU, but one thing you have to watch is that these watt meters aren't all that shit hot. They will give different accuracy readings on different PSUs. I think they pick up on resonance spikes on the wave, and this varies depending on whether the PSU has good filtering and power factor correction. When you look at the current/amps input on an oscilloscope, you will see what crap the input wave actually is, and you will marvel that a digital thingo can do a real-time integration of V x A, and come up with any figure.
I'm shocked, shocked to learn that my cheapie kill-a-watt meter might be reporting bogus numbers :) .

But seriously, one way to determine if the ATX PSU's overhead is for real would be to measure the difference in power draw between idle and full CPU load for each PSU. If this difference is about the same for both ATX PSU and Pico then the overhead is for real. If I have some time this weekend I will run this test and report back.

astrayan
Posts: 45
Joined: Fri Jun 17, 2005 5:54 am
Location: astraya

Post by astrayan » Fri Apr 25, 2008 8:19 pm

I'm going to be running an M2-ATX directly off 12V soon (96% efficiency), but I have to wait for an inline 20A mu45 ammeter. It's difficult measuring current and voltage at the same time. I'll be trying to measure a Brisbane on an M2A-VM-HDMI

I like those Greenpower WD drives. Normally when they call something 'Green', it takes more power; such as Seasonic Ecowatts, or Antec Earthwatts, or the P35 chipset. But a 500GB WD GP drive takes 3.3W, which is pretty awesome.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Sat Apr 26, 2008 2:49 am

Normally when they call something 'Green', it takes more power; such as Seasonic Ecowatts, or Antec Earthwatts,
how exactly does the Earthwatts take more power than a normal (non 80Plus) PSU?

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Post by frank2003 » Sat Apr 26, 2008 9:42 am

Just to give another idea of how low one could go with a desktop dual core system, here are the specs for another Biostar 690G system:

mobo: Biostar TA690G
CPU: AMD X2 Brisbane 4400+ underclocked to 1GHZ @ 0.784V
Memory: 512MB, PC2-4300
Disk: 2.5" WD notebook drive (SATA)
PSU: PicoPSU-120, 80W brick
fans: stock CPU fan + 1x80mm case fan

Power Consumption: 20W-21W idle with GbE disconnected; 23W idle with GbE connected.

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Post by frank2003 » Sat Apr 26, 2008 10:05 am

frank2003 wrote:one way to determine if the ATX PSU's overhead is for real would be to measure the difference in power draw between idle and full CPU load for each PSU.
I ran a quick test between the picoPSU and the Q Technology 300W PSU (this is a rather old PSU, purchased from quietpcusa.com some years ago to power a dual P-III system). Here are the numbers as measured by kill-a-watt meter (I also included volt-amp numbers):

PicoPSU
idle @ 1ghz, 27W (27VA)
orthos @ 1.8ghz, 45W (46VA)

ATX PSU (Q Technology 300W):
idle: 49W (51VA)
orthos @ 1.8ghz: 67W (70VA)

In terms of watts, the difference between idle and full load is 18W for both PSUs. One can draw the conclusion that 1. the kill-a-watt is consistent in reporting the same amount of power consumed for the same task, and 2. the ATX PSU operates at a point that's 22W above that of the picoPSU.

From memory I think the overhead numbers for other PSUs I mentioned earlier were around the same ball park or worse.

astrayan
Posts: 45
Joined: Fri Jun 17, 2005 5:54 am
Location: astraya

Post by astrayan » Sun Apr 27, 2008 4:05 am

jaganath wrote:how exactly does the Earthwatts take more power than a normal (non 80Plus) PSU?
With the Seasonic, it's easy to demonstrate how the Ecowatts takes more than older Seasonic PSUs, because they are reviewed on this site, and the Eco is pretty ordinary. However, I believe the Antec Eartstuff is no better performing than the NeoHE 380 http://www.antec.com/specs/NeoHE380_spe.html and the Trio430 http://www.antec.com/us/productDetails.php?ProdID=23430 , but what is alarming is that someone who is concerned about the "Earth" would be running a 380W computer (which is the smallest Earthstuff they make). Seasonic engineers reckon that meeting the 80+ standard was hard for them, so I think they reach this 20% load target by increasing the size of the PSU. If anyone really cared, they would be making 100W supplies, like the picos. The average office computer, which is what they should be aiming at, would be working at 7% load on a 380W PSU. So, I think higher efficiencies were achieved ages ago in smaller supplies. The 80+ standard is flawed.

Antec 430 (made by seasonic)
http://www.silentpcreview.com/article684-page4.html
Seasonic Eco
http://www.silentpcreview.com/article792-page4.html
Past Greats
http://www.silentpcreview.com/article601-page4.html

I'd say the Earthwatts was 56% efficient on the above Brisbane power draw.

smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Post by smilingcrow » Sun Apr 27, 2008 8:40 am

astrayan wrote:I'm getting a bit peeved with the boast that Intel has designed a superior chip to the Athlon, and that Core 2 45nm performs so well on a joules per instruction basis, and then Intel don't let anybody undervolt to 0.8V. It's like congratulating yourself for buying a hybrid car, and then towing a caravan everywhere. Until Intel has Athlon-style undervolting, they are not really in the race, are they?
Under-volting is not a mainstream hobby and reviews focus on what you can achieve at stock settings which is what the vast majority of PCs run at. So from an environmental perspective what matters most is how PCs built by the large OEMs perform. E.g. how does a Dell Inspiron 530 (Intel) compare with an Inspiron 531 (AMD)? I’d like to see the data for that.

With regard to under-volting as low as 0.8V, from what I’ve seen it seems to be the law of diminishing returns. How much do you save at idle when dropping VCore from 1V to 0.8V?

frank2003
Posts: 332
Joined: Sun Jun 24, 2007 3:35 am

Post by frank2003 » Sun Apr 27, 2008 11:03 am

For normal desktop usage I don't think it's worthwhile to tweak your power consumption to the lowest possible levels; I'm sure AMD's Cool 'n Quiet and Intel's SpeedStep (or whatever they call it these days, EIST?) are plenty good in terms of managing CPU utilization and power.

However, there's a class of machines where running at the lowest possible power at idle (or near idle) is important - machines such as PVR and home file servers that are on 24/7. For these machines a watt saved is 24 watt-hours saved per day.

That said, I think we can be far more effective in saving power by using a more efficient PSU.

Post Reply