SoC Poll - What's most important for you?

Power or Battery - what's more important for you?

  • Power - I don't care about battery, as long as it's the FASTEST!

    Votes: 38 11.0%
  • Balanced - I don't mind a little shorter battery life if it has more power

    Votes: 199 57.5%
  • Battery - I'm pretty sure every current SoC is fast enough for my needs, but I want a long battery l

    Votes: 109 31.5%

  • Total voters
    346

You'd need to actively disable to cores dynamically when they are not needed, but I don't think Linux does that.
At least on x86 Linux actually does that if the CPU supports it (you can verify that with powertop).

On my PC the CPU is in C6 power state (which is the lowest power state my CPU supports) most of the time in Linux, which means the core is pretty much completely powered down through power gating.

It could very well work on ARM as well, but I haven't checked that.
 
Last edited by a moderator:
Yes, if you had a steady to high CPU usage, multiple cores would probably be more efficient. The problem is with webbrowsing that you have high peaks for a couple of seconds at a max, then it idles again.

When idling, four CPUs need more power than two or one.

So with everything that only has short peaks, less CPUs should be better.

You'd need to actively disable to cores dynamically when they are not needed, but I don't think Linux does that.
It depends on the power governor. I know for sure there are ones that will power gate cores that are frequently unused. Ones that don't do this aren't very good.
 
Last edited by a moderator:
It depends on the power governor. I know for sure there are ones that will power gate cores that are frequently unused. Ones that don't do this aren't very good.
Yes of course, you have to use the correct power governor.
 
ARM

Pro:

I think the PCB is easier to design, makes it cheaper to build and faster to get it working.

Reuse parts, software base and knowledge of the pandora (makes it cheaper and easier to produce).

Feels like the GPU with ARM is faster, could be wrong there. We need it for gaming, so the CPU/GPU combination must suit those needs.

I can't think of any cons! (that is just me, there could be plenty I can't come up with now)

Con:

X86

Pro:

PC OSes, software and games.

Build for the masses (even non games might be buying it creating bigger sales).

Con:

Leaving current pandora base behind.

Not sure what the new Intel chip requires, but the X86 use to come with bios and all such additional chips you do not really want on you tiny board.

People not interested in the community will get it without respect for ED and the project, might lead to additional stress/work/costs.

One thing that might be good to think about is: "What would the successor of the Pyra look like?".

If it might use the X86 the switch would mean a better software base in the future.

And in the end I doubt if the CPU architecture would really matter to me as I probably still want the device.

(looks like no-script kills your lay-out in the posts :p )
 
Last edited by a moderator:
Install cpufreq, i think the options state which cores are bundled to each other. Available governors and a whole heap of settings is available there.
 
Last edited by a moderator:
I am using cpufreq right now on the devboard - but then again, since it can't switch to 500 MHz because of a misconfiguration, who knows if it even tries to disable the core now ;)
 
I am using cpufreq right now on the devboard - but then again, since it can't switch to 500 MHz because of a misconfiguration, who knows if it even tries to disable the core now ;)
I can set mine to 500mhz manually on my omap5432 dev board.. However I haven't been playing with much outside of the base image I got from gta04 and yourself.


I can set it to 500mhz,1000mhz and 1500mhz on command manually, however nothing in between.. I haven't even looked into the governors yet.
 
Are you sure you can do that?

I can also set it and the userspace reports it's set to 500MHz, but both power top and my Multimeter that reads out live power consumption says it doesn't.

Which might well be that wrong voltage setting mentioned above - cpufrequtil THINKS it can set the CPU speed to 500, but OMAP doesn't do it because it's an unsupported frequency.
 
^ just going by what cpufreq-info is telling me.. I can setup an amp meter...
 
cat /proc/cpuinfo gives live setting, at least on x86 where you can see the frequency changing when doing things.
 
Well I'm seeing a slight difference between 500Mhz and 1Ghz and a greater jump once set to 1.5Ghz..

Just pushed a video up.. mind my voice I'm getting over a head cold.

http://www.youtube.com/watch?v=mh5kgmd1AT8
 
Last edited by a moderator:
Such a sexy voice. Have you ever considered to quit your job to become a singer ?
 
Don't be modest ^^.

Did you try to compare compilation time between OP and dev board (source copied in RAM as the SD accesses are not the same) ?

http://openbox.org/dist/openbox/openbox-3.5.0.tar.gz is small enough and does not require dependencies (at least on Slack).
Well I was only trying to prove setting it to 500Mhz actually did anything as discussed earlier with EvilDragon a few posts ago.. To test out Pandora vs omap5, I would have to cut up another power cable to properly hitch up the amp meter as it does have a different sized connector.. Also would need to run it without the battery for a similar comparison other wise it would just be running off the battery and drawing the charging amount.

Compile times is something I could try, but can't do it tonight.
 
Last edited by a moderator:
It's interesting enough to see (has it to be proven ?) the non-linearity of the CPU Hz/V .

Whenever you want. I just did it on the OP:

1GHz@1GHz

./configure

time make

real 5m15

user 4m45

sys 0m15

I'd be curious to see your make -j1 and make -j4

Of course, i know it's a non-benchmark (not the same compiler, libs), but it gives an idea.
 
Last edited by a moderator:
I remember one day, i had Joe Cocker's voice and kept singing "Unchain my heart" again and again. So much fun.
 
Back
Top