soc power draw [rough numbers]

[monstercameron]

so I just compiled this quick list of soc power numbers, it must be stated that these all include platform power[ei the other components] and they all have different types and sizes of displays.

lower to higher: [perf]

rockchip 3188

idle                           - 1.4-5W  |  2.8-8W 

max load                   - 7W        | 11W

sources:

http://www.notebookcheck.com/Test-Bullman-Tab-8-Mini-AQQI-Tablet.99453.0.html

http://www.notebookcheck.com/Test-Bullman-Tab-9-AQQR-Retina-Tablet.93938.0.html

http://www.notebookcheck.com/Kurztest-Hyundai-MID-X900-Play-Tablet.96596.0.html

exynos 5250 [for refence]
 

idle                           - 4-9W

max load                  - 9W

sources:

http://www.notebookcheck.com/Test-Google-Nexus-10-Tablet.85166.0.html

snapdragon s800

idle                           - 1W           |  1-2W

max load                   - 4.3-8.4W |   4-6W

sources:

http://www.notebookcheck.com/Test-Google-Nexus-5-Smartphone.105040.0.html

http://www.notebookcheck.com/Test-Sony-Xperia-Z1-Smartphone.103772.0.html

nvidia tegra 4

idle                           -  2-4W        |   2.5-6W

max load                   -  7-16W     |    9-13W

sources:

http://www.notebookcheck.com/Test-Microsoft-Surface-2-Tablet.105059.0.html

http://www.notebookcheck.com/Test-Toshiba-eXcite-Pro-AT10LE-A-108-Tablet.98288.0.html

intel atom z3740

idle                           -  2-4W

max load                   - 11-12W

sources:

http://www.notebookcheck.com/Test-Asus-Transformer-Book-T100TA-C1-GR-Convertible.105688.0.html

amd a6-1450

idle                           - 4-7W

max load                   - 13W

sources:

http://www.notebookcheck.com/Test-Samsung-ATIV-Book-9-Lite-905S3G-K01DE-Subnotebook.101920.0.html

this just shows that unless there is a compromise on performance, the power usage will be higher than the omap [atleast under load], so the battery would probably need to me larger. We can argue over uarch but it seems powerdraw is no longer a major concern when comparing the two. The s800 has the best power numbers, while the a6-1450 has the highest performance and the z3740 is at a happy medium, tegra 4 aint half bad either.

 

[Exophase]

The S800 has the best power numbers because they tested a PHONE, unlike everything else there. This isn't even approaching an "SoC power draw" comparison. Unless you're probing the SoC rails in isolation, which they aren't, this is useless. I guarantee you that if you were able to look at power numbers only from the SoC that you'd be able to get the S800 to draw more than the Z3740 and most likely RK3188 and Exynos 5250.

These extreme utilization tests are of questionable value for power consumption to begin with. They're torture tests/power viruses where not only the CPU and GPU are pushed as hard as they can be but with the screen at max brightness and the wifi cranked up. They're upper limits, that's it.

 

[comradekingu]

Why do you account for performance on the amd chip when you forget to do so in comparison of OMAPs? Over the same period of time the newer chips compute more data, that saves battery life.

 

[_wb_]

The Pandora consumes 0.1W when idle (lid closed) or 0.6W when idle (lid open / display on at min brightness), 1-2W at normal load, 3W if you're stress-testing it.

I have no reason to assume that any of the above SoCs would use more power than OMAP3 for the same task.

Your idle numbers probably mostly reflect the power needed for the display (and maybe wifi/3G in the background). Your "max load" numbers reflect a highly improbably scenario of keeping all cores busy - that is probably only going to happen in benchmarks and stress tests.

There is absolutely nothing you can conclude from the numbers you listed.

I don't think the P2 will need a bigger battery, at least not because of the SoC. We just have to make sure that the display it gets is not too power-hungry, because it will be larger and higher resolution. Sure, the SoC will be able to use more power, but there's no need to make the device last 10 hours during a stress-test, it's enough if it lasts 10 hours under low/medium load (which corresponds to more processing power than what the P1 has anyway).