Atom vs Pentium M: Power Consumption results

Info & chat about quiet prebuilt, small form factor and barebones systems, people's experiences with vendors thereof, etc.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
joefriday
Posts: 9
Joined: Sun Dec 17, 2006 11:49 pm

Atom vs Pentium M: Power Consumption results

Post by joefriday » Mon May 25, 2009 9:07 am

I purchased an Acer Aspire One gently used off CL for $200, and decided to test it's power consumption. It is the 8.9" screen, 160GB 5400 rpm conventional HDD, 1GB DDR2 ram, Atheros internal wifi, Atom 45nm single core processor with 800MHz 1.0volt idle, 1.6GHz 1.3 volt load, and intel 945GSE chipset model. It has a 3 cell 2200mAh battery and a 19volt 30 watt max AC adapter. I tested its idle power consumption as well as CPU load consumption using prime 95.
To compare, I used my old trusty laptop, a Dell Inspiron 1200 with the following specs: 14" conventional display, 30GB 4200 rpm hdd, 512MB DDR ram, Atheros internal wifi, Pentium M 130nm single core processor, 1.7GHz Banias core, with 600MHz 0.7160 volt idle, 1.7GHz 1.1960 volt load, Intel 910GL chipset. it has an 8 cell 4400mAh battery and a 19v 62watt max AC adapter. All measurement taken using a Kill-A-Watt power meter. Here's the comparison figures:

Idle power consumption, maximum screen brightness, low power states enabled. Wifi enabled, hdd spinning:
Acer: 15 watts
Dell: 18 watts

Idle power consumption, maximum screen dimness, low power states enabled, wifi enabled, hdd spinning:
Acer: 13 watts
Dell: 13 wats

Idle power consumption, screen off, low power states enabled, wifi enabled, hdd spinning
Acer: 12 watts
Dell: 11 watts

Idle power consumption screen off, low power sate, wifi disabled, hdd spining:
Acer: 12 watts
Dell: 9 watts

Idle power consumption screen off, low power state, wifi disabled, and hdd stopped:
Acer: 9 watts
Dell: 8 watts

Load power consumption, max screen dimness, low power state, wifi enables, hdd spinning:
Acer: 14 watt
Dell: 16 watt

Standby power consumption:
Acer: 1 watt
Dell: 1 watt

Idle power consumption, max power state, max screen dimness, wifi enabled, hdd spinning, fan off:
Acer: unable to force max power state at idle. 13 watts.
Dell: 14-15 watts

Load power consumption, max power state, maximum screen dimness, wifi enabled, hdd spinning, fan off:
Acer: 16 watts
Dell: 26 watts

Load power consumption, max power state, maximum screen dimness, wifi enabled, hdd spinning, fan on low
Acer: 17 watts
Dell: 27 watts

Load Power consumption, max power state, maximum screen dimness, wifi enabled, hdd spinning, fan on high:
Acer: 18 watts
Dell: 28 watts

Super Pi 1M times and power consumption: forced low power state, max screen dimness, wifi enabled, hdd spinning, auto fan:
Acer: 2minutes 58seconds, 14-15 watts (800MHz)
Dell: 2minutes 10seconds, 16 watts (600MHz)

Super Pi 1M times and power consumption: max power state, max screen dimness, wifi enabled, hdd spinning, auto fan:
Acer: 1minute 33seconds, 16 watts (1600MHz)
Dell: 59 seconds, 26 watts (1700Mhz)

Typical battery life
Acer: 3 hours
Dell: 6 hours

Conclusion: It's hard to say if the Atom netbook is really efficient or not. In this comparison, we see a full-size notebook, with a real screen size of 14", score comparable idle power consumption to a netbook with a tiny screen. Even more impressive is the fact that the old Dell notebook is using the older Pentium M Banias, a 130nm transistor design, almost 3x the size of the 45nm transistors found in the Intel Atom. The old Pentium M also has more transistors in total: 77million vs the Atom's 47 million.

When we look at a performance per watt perspective things become even more confusing. At low power states, the Dell's Pentium M bests the Atom in super pi by 48 seconds while consuming just 1.5 watts more power. However, if we let the Atom run at max 1.6GHz, it completes 1M super pi in 1 minute 33 seconds, besting the Pentium M's low power state score of 2minutes 10 seconds by 37 seconds, while both units consume the same amount of power(16 watts). Dell's power consumption sky rockets at max power settings with 26 watts, but beats the best score of the Atom by a sound 34 seconds. Performance per watt in watt/seconds is as follows, lower scores being more power efficient:
Acer, low power state: 2581
Acer, max power state: 1488
Dell, low power state: 2080
Dell, max power state: 1534

The results give some insight to the story. When both are at low power states, the Dell is more efficient given a performance per watt standpoint. However, at max power states, the Acer netbook as more power efficient, given the fact that it does not increase power consumption significantly when under load, thanks to the Atom processor. Finally, if we look at the actual power consumption figures and correlate them with the performance per watt figures calculated above, we come to the conclusion that the Acer Aspire One netbook is superior in performance per watt, as it can run at it's maximum performance state of 1.6Ghz while consuming only 16 watt under load, a power consumption figure comparable to the Dell's load power consumption while stuck at 600MHz, while producing much more performance than the Dell's low power state performance.
Hope you enjoyed the comparison.
Joe

Ch0z3n
Posts: 400
Joined: Thu Oct 02, 2008 5:48 am
Location: Orlando, FL

Post by Ch0z3n » Mon May 25, 2009 11:05 am

One of the major problems with netbooks currently is the horridly inefficient northbridge they use. If you look at non-laptop atom boards, the northbridge has a fan, the processor is passive.

Also, when you were doing your power measurements, did you have the battery removed?

On a side note, kill-a-watts are good to give you a general idea, but they have their faults when it comes to accuracy.

joefriday
Posts: 9
Joined: Sun Dec 17, 2006 11:49 pm

Post by joefriday » Mon May 25, 2009 12:26 pm

Ch0z3n wrote:One of the major problems with netbooks currently is the horridly inefficient northbridge they use. If you look at non-laptop atom boards, the northbridge has a fan, the processor is passive.

Also, when you were doing your power measurements, did you have the battery removed?

On a side note, kill-a-watts are good to give you a general idea, but they have their faults when it comes to accuracy.
The 945GSE chipset in the netbook IS NOT the same as that found on an ITX board. The 945GSE has a max TDP of just 6 watts, while the 945GC has a TDP of 22.2 watts. Kill-a-watts are accurate to 1-2 watts. The point of the article is not necessarily the actual number of the wattage, but the comparable difference between readings, which are always the same. The batteries were attached, and fully charged. Power consumption does not differ with battery present or removed. I've already tested that.

Ch0z3n
Posts: 400
Joined: Thu Oct 02, 2008 5:48 am
Location: Orlando, FL

Post by Ch0z3n » Mon May 25, 2009 3:49 pm

I understand it isn't the same, but that 'just 6w' is still more than twice the power consumption of the processor. You also have another 3.3w I/O controller.

Anyway, what your power consumption readings say to me is that there is a relatively high overhead with the current atom setups since there isn't a huge variation in power draw between the lowest and highest power modes.

Post Reply