SoC: Back and forth!

[God Ginrai]

- Qualcomm SoCs don't have big.LITTLE. They do have the ability to clock the CPU cores independently, which ARM-designed CPUs don't. Except you can clock clusters of CPUs independently, ie the A15s vs the A7s in big.LITTLE.

That's actually pretty interesting.

Does that mean a Quad-Core A15 SoC will most probably use MORE CPU Power when running something simple like a SNES emulator than a Dual-Core A15 SoC, simply because all SoCs run at the same clock (and 4 is more than 2)

Or is Linux able to share one task between all four cores, so that the four-core system would need less clock for the same process as a dual-core system?

If that's the case, DualCore A15 or Qualcomm seems to be the better choice for us (battery-wise)

 I did read somewhere that it's one of the features of 805 to clock the cores differently.

Don't have the source any more.

Would be careful if we only can get the 800.

Read Exophase's response to EvilDragon. He explains how the big.LITTLE works.

EDIT:

The A80 sounds interesting, however I'm having trouble finding specs for this outside of the type and number of ARM cores and the fact that it uses a 6 series PowerVR GPU. If you could point me to where there are more detailed specs, I'd love to read up on this SoC more. (All I can seem to find are people talking about the specs of the OptimusBoard rather than the SoC)

-God Ginrai

It's still too early I guess, it's not even listed on their product page yet. But if they're really shipping the eval board within a month or so they'll probably have datasheets to go with it.

Well, 8 cores w/ true big.LITTLE and an SGX that is a series above what the OMAP 5 provides does sound enticing. I can only hope that the rest of the specs are equally impressive, and that they have Linux support.

-God Ginrai

 

[mcobit]

The 800/805 doesn't use big.little.

 

[Askarus]

- Qualcomm SoCs don't have big.LITTLE. They do have the ability to clock the CPU cores independently, which ARM-designed CPUs don't. Except you can clock clusters of CPUs independently, ie the A15s vs the A7s in big.LITTLE.

That's actually pretty interesting.


Does that mean a Quad-Core A15 SoC will most probably use MORE CPU Power when running something simple like a SNES emulator than a Dual-Core A15 SoC, simply because all SoCs run at the same clock (and 4 is more than 2)


Or is Linux able to share one task between all four cores, so that the four-core system would need less clock for the same process as a dual-core system?


If that's the case, DualCore A15 or Qualcomm seems to be the better choice for us (battery-wise)

 
I did read somewhere that it's one of the features of 805 to clock the cores differently.


Don't have the source any more.


Would be careful if we only can get the 800.

Read Exophase's response to EvilDragon. He explains how the big.LITTLE works.


EDIT:

The A80 sounds interesting, however I'm having trouble finding specs for this outside of the type and number of ARM cores and the fact that it uses a 6 series PowerVR GPU. If you could point me to where there are more detailed specs, I'd love to read up on this SoC more. (All I can seem to find are people talking about the specs of the OptimusBoard rather than the SoC)

-God Ginrai

It's still too early I guess, it's not even listed on their product page yet. But if they're really shipping the eval board within a month or so they'll probably have datasheets to go with it.

Well, 8 cores w/ true big.LITTLE and an SGX that is a series above what the OMAP 5 provides does sound enticing. I can only hope that the rest of the specs are equally impressive, and that they have Linux support.

-God Ginrai

The 800/805 doesn't use big.little.

And only the 805 is able to clock the cores at different speed each.

 

[God Ginrai]

The 800/805 doesn't use big.little.

I wasn't suggesting they did. I was correcting what looked to me like an assumption that you couldn't turn off cores in a big.LITTLE configuration.

-God Ginrai

 

[Exophase]

And only the 805 is able to clock the cores at different speed each.

Actually all multicore Qualcomm SoCs that use their custom CPUs have asynchronous DVFS (different clock and voltage for each core). Even the old dual core Scorpion SoCs from a few years back.

 

[Askarus]

And only the 805 is able to clock the cores at different speed each.

Actually all multicore Qualcomm SoCs that use their custom CPUs have asynchronous DVFS (different clock and voltage for each core). Even the old dual core Scorpion SoCs from a few years back.

Here's the source: http://vr-zone.com/articles/qualcomm-snapdragon-805-vs-snapdragon-800/64753.html?utm_source=rss

As illustrated by the chart, the Snapdragon 805 features the Krait 450 CPU, which now comes with a 2.5 GHz clock. The quad-core CPU is based on the AMRv7 instruction set, but the architecture has been tweaked by Qualcomm, which allows each core to be at a different clock based on the usage.

So I think I did understand it wrong than and the quote refers to every CPU, not only the new one.

 

[mcobit]

That almost sounds like the best way for single threading performance and this also is what most applications ported feom Pandora will be able to profit from.


big.little seems more like a powersaving function that is intended for phones where you need to keep the system alive and listening while it is in standby to me.


Might not be as relevant with the Pyra.

 

[Askarus]

In my oppinon big.LITTLE might bring a huge power advantage to the Pyra.

I've read of A7 cores that are slightly faster than the A8 (at least what concerns clockspeed).

Even if they are only as fast as Pandora cores they consume less power than A8.

So there are 2 advantages.

1: Things that run on Pandora will last longer on the Pyra.

2: If developers want to support the Pandora further they also can optimize their Programs to run well on the A7 cores. That will also give more battery life.

 

[mcobit]

That might depend on how efficient a single a15 clocked at low speed and volatage is compared to four a7.

 

[Askarus]

Why?

Can't you only use 1 A7 and shut down the others in bib.LITTLE?

 

[mcobit]

I thought the cores have to be clocked the same in big.little. You can only switch between the sets of cores.


But maybe I misunderstood that.


Either way the question still is valid.


How efficient is an a7 at the same speed as the a8 in the Pandora compared to an a15 at the same speed (not MHz)?

 

[bzar]

Aren't the cores individually disablable (funny word) even though their voltage/clock are changed in unison?

 

[Galaxis]

The Allwinner was announced after we already had secured the OMAP5...

Allwinner seems to be mostly an Android shop, from a first look - the Cubietruck board that uses one of those (http://cubieboard.org/) comes with Android preinstalled, for example. And all the Linux distributions for that board seem to use the same 3.4.75 kernel...

 

[sswam]

Edit: Looks like all Snapdragons have integrated GPS. I'm a little concerned about possible "mischievous behaviour"  

If its going to be Qualcomm please dont take the full package with integrated 3G.

The governments via the phone companies do track all cell phones, all the time, while they are turned on; and they record the history of every cell phone's movement. This is done using a technology called TruePosition LOCINT, not GPS. It uses "uplink time difference of arrival" (U-TDOA) to find your position from surrounding cell towers. U-TDOA is perhaps slightly less accurate than GPS, but more reliable in cities.

So yes, big brother is watching, and the GPS on the chip is irrelevant to that. The chip does not have to phone home in order for them to find you. If you use cellular radio at all, that is enough. Maybe you knew this already.


An article in wired: http://www.wired.com/dangerroom/2011/07/global-phone-tracking/all/

My blog post: http://sswam.com/2010/11/04/big-brother-can-track-cell-phones-accurately-en-masse-sans-gps/

TruePosition LOCINT site: http://www.trueposition.com/products/trueposition-locint/

 

[Null]

Aren't the cores individually disablable (funny word) even though their voltage/clock are changed in unison?

After having read http://m.hexus.net/tech/tech-explained/cpu/48693-arm-biglittle-processing/ it looks like they're not individually disableable. the processors are though.

 

[cwall]

Another thought, it is important to pick a SOC that is going to be produced for a while.   So who else is buying the SOC for long term products, so you know that it will be around next year when you want to do another round of boards.  After all Pyra will be small volume to whoever we buy from, and the chip maker will not be starting up another production run just on our re-order, although it would be nice if Pyra got to that point.    :rolleyes:

 

[Hồng Thất Công]

^ of course ED knows that. Have you been living under a rock

 

[ghostpatch]

The governments via the phone companies do track all cell phones, all the time, while they are turned on; and they record the history of every cell phone's movement. This is done using a technology called TruePosition LOCINT, not GPS. It uses "uplink time difference of arrival" (U-TDOA) to find your position from surrounding cell towers. U-TDOA is perhaps slightly less accurate than GPS, but more reliable in cities. So yes, big brother is watching, and the GPS on the chip is irrelevant to that. The chip does not have to phone home in order for them to find you. If you use cellular radio at all, that is enough. Maybe you knew this already.

yes, I did know.

1.Pyra is not a phone. 3G is and should be a seperate option.

2.It is a difference wether GPS is integrated in the SoC or optional on a seperate chip. Atackers assholes could switch it on and track you. No physical way to cut the circuit. I have enough stuff with GPS in it and I never use it -.-  .Its understandable that I don't want that huge unnecessary risk. There are tons of GPS sticks if needed.

3.Thats not a reason to give up, and make it easy for them to track you and collect data!

4.shockingly a smartphone is not one of our vital organs. We can live without it. Just leave it at home, switch to airplain mode or turn it off completly. Thats working for me. People will complain about you not beeing reachable 24/7 or delayed responses but I don't care ; )

 

[notaz]

2014 is a completely different world from 2007, criticising the team for following along with the market at the time is unfair.

I am not criticising the team for it and I am well aware that in 2007 ARM CPUs without any FPU were much more common, the Pandora still hasn't one of those. I was merely asking why that decision had been made. As it turned out, it actually was because of the graphics driver.

In 2007 armhf compilers did not exist, it has little to do with the driver..

- I'm not part of the Pandora team and I don't really matter compared to notaz. I'll port DraStic whatever is chosen and whatever other emulators I work on in the future, if that's where the concern it. Also, I don't prefer x86, if all other parts of the SoC are equal, that's just not always the case. I'd rather the thing stayed with ARM too if it meant avoiding losing notaz, because there's a much bigger risk that this won't work without him.

You're giving me way too much credit, there have been lots of negative feedback lately. I kind of wish ED chose x86 so I can finally retire.

 

[EvilDragon]

- I'm not part of the Pandora team and I don't really matter compared to notaz. I'll port DraStic whatever is chosen and whatever other emulators I work on in the future, if that's where the concern it. Also, I don't prefer x86, if all other parts of the SoC are equal, that's just not always the case. I'd rather the thing stayed with ARM too if it meant avoiding losing notaz, because there's a much bigger risk that this won't work without him.

You're giving me way too much credit, there have been lots of negative feedback lately. 

I don't think so.

Whenever there were issues with the Pandora, you helped as good as possible.

Of course, you couldn't fix Wifi (still probably some hardware issue), but you did improve it a lot.

Same for the touchscreen (which had a lot of noise).

And many many other optimizations and tweaks.

I wouldn't say the Pandora wouldn't have been possible without you - but who knows whether we would be running Kernel 3.2 with awesome standby time in that case.

I also haven't seen lots of negative feedback, to be honest - I've only seen heated-up discussions at the boards (which is normal, when different people have different opinions), but I don't see real negativity here, and definitely not towards yourself.

I kind of wish ED chose x86 so I can finally retire.

Whatever we'll choose - you should not be forced to help out.

You should do what you like to do - don't feel yourself under pressure from me.

If there's something you can't or don't want to do anymore, let me know and I'll try to find someone else to do that.

I'm really happy for your help, as you're doing a great job (you never give yourself credit for that though), and during the long years I've met you in IRC, you've become a close friend to me, but I will not force you to work on something!