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

But what could that be?

DreamCast runs at 50 - 70% speed on the Pandora already, so it should work with the Pyra running with Single Core.

PSP is similar, N64 as well.
Assuming Gamecube gets better support for the 4 A15 CPU,it not like I'm going to use it for only that. Probably play on a car/wall charger.
That said, I'm confused. Even phones last longer than that on these battery

tests.

How would you connect the two batteries together? You see how they are made right ?
That's a good point.
 
Last edited by a moderator:
But what could that be?


DreamCast runs at 50 - 70% speed on the Pandora already, so it should work with the Pyra running with Single Core.


PSP is similar, N64 as well.
Assuming Gamecube gets better support for the 4 A15 CPU, it not like I'm going to use it for only that. Probably play on a car/wall charger.

That said, I'm confused. Even phones last longer than that on these battery


tests.
Under full load?

Nope. My CortexA9 Smartphone drains the battery in 3 hours when playing games.

They only last longer when they're just waiting for phone calls (and the Pandora lasts over 100 hours just lying around).
 
Any amount of A15 cores up untill you draw 2A at 5V  (10W) total, is a good thing.

A bigger battery standard means it will take longer to charge, i think a 12 hour forecast on an actual most-cases usecase sounds splendid. Good tradeoff.

People dont use the available range on their teslas each day, just because there is a way to run it down quick doesnt mean you have to keep your foot to the floorpan just because its potential sportscar performance...

With range anxiety, all you end up doing is being less efficient, dragging around a unused battery pack most of the time.

Where are people for hours playing games and running computing in the remainder cpu and gpu cycles? There is a limit to how long i can play N64, i have to sleep at some point...

The pandora was great, this is better at the same things, and then _at least_ for hours whatever thing you can throw at it beyond that.  _And_ it has a hotswap feature.

Get the DC directly option to work and everyone is fine. Lets not stare ourselves blind at 2 hours runtime, a bigger battery takes longer to charge, and makes the device bigger, and then you end up carrying batteries for two platforms, nobody wins there.

Keep options open. Looking for other batteries doesnt hurt, but accomodate running from DC alone and the hotswap as much as possible at the other end. If the test units for some reason should been unsatisfactory with stock battery, one can always redesign the case at that point.
 
Last edited by a moderator:
I don't get it, how can 2 cortex a15 cores at 1.5 ghz drain pandora's battery in about 2 hours? That's horrifying.  :eek:

Does cortex a15 generally have worse perf/watt than a8 on the highest clock? 

I want to know if all cortex a15 based socs use that much power at 1.5ghz per core. 

Still we can use the little cores for light emulation, such as psx, snes, mega drive, and so on. 
 
Nope. My CortexA9 Smartphone drains the battery in 3 hours when playing games.
My HTC Desire Z with it's 800Mhz Snapdragon and extended battery which I only need to charge once every 2 or 3 days under regular use is dead in under two hours if I'm gaming on it.
I don't get it, how can 2 cortex a15 cores at 1.5 ghz drain pandora's battery in about 2 hours?
One A15 at 1.5Ghz is about three DM3730s at 1Ghz. The OMAP5 running full throttle is like 6 Pandoras running full steam off the same battery. One Pandora drains the battery in 6 hours, two would drain it in 3, four in 1.5 hours, and six in just 1 hour. It's basically doubling the battery life, OR getting twice as much work out of it in the same battery life.
 
I don't get it, how can 2 cortex a15 cores at 1.5 ghz drain pandora's battery in about 2 hours? That's horrifying.  :eek:

Does cortex a15 generally have worse perf/watt than a8 on the highest clock?
Nope, at much better perf/W.

MHz is not MHz.

The CortexA9 has about 50% more performance running at the same clock speed as the CortexA8.

Same step from the CortexA15 compared to the A9.

So if you have Pandora running at 1,2GHz, a Cortex A9 would need approx. 800MHz for the same task.

And a CortexA15 would need approx. 533MHz for the same task.

So a CortexA15 with 500 - 550 MHz should have approximately the speed of a 1,2GHz Pandora.

That's one of the main mistakes most people do - comparing MHz with each other. That only works if you have the same processor technology (i.e. A8, A9 and A15).

One A15 at 1.5Ghz is about three DM3730s at 1Ghz. The OMAP5 running full throttle is like 6 Pandoras running full steam off the same battery. One Pandora drains the battery in 6 hours, two would drain it in 3, four in 1.5 hours, and six in just 1 hour. It's basically doubling the battery life, OR getting twice as much work out of it in the same battery life.
There's just one thing we shouldn't forget:

Both cores share the same memory. So a DualCore doesn't give double performance :D

Now we're going complicated ;)
 
Last edited by a moderator:
^ I see, ED, then I don't get why do big little have those little cortex a7 cores?

We can just clock a15 at a lower clock, and save us a ton of battery life. 

Does A7 have better perf/watt than an underclocked A15?

I guess now it's up to you to decide how much power is best for us, and if we could use an even larger battery. I'm personally fine with any cortex a15 cores then, they are all good enough. 
 
I would prefer a slimmer profile to a bigger battery, but if a capacity increase can be done while reducing the size of the Pandora even just a little, I'd be all for it.
 
Just for the information:

I'm currently working on testing various power usages of the OMAP5 EVM, etc.

PLEASE NOTE!

THESE ARE JUST QUICK TESTS I DID MYSELF!

Please don't take them as 100% correct, I might have made a mistake here!

Also, why I generally know what I'm doing, I'm no expert!

If my assumptions and calculations are correct, then it would run about 12 hours under full load when limited to the speed of a 1GHz Pandora running at 1,2GHz.

That's what I expected upfront - the Perf/W is better, so it should be better at same speed.

When idling, power consumption goes down, should be similar to what the Pandora does right now (needs some more work on the Kernel, same for the Pandora when we moved from 2.6 kernel to 3.2).

However, when running both cores at full load with 1,5GHz, consumption goes up... Battery would last a bit more than 2 hours.

That would be like running six 1,2GHz Pandoras with full load though (pretty unlikely scenario, I guess - what app needs the power of six Pandora at the same time for a longer period?)

It made me think of two things though...

1. Does a Quad-Core A15 even make sense? That would lower the battery life to 1 hour... big.LITTLE wouldn't help here. Full Load is full load, it would use the A15 and not the A7.

2. Should I try to increase the battery size a bit? It could be possible and get us more capacity, but the battery won't fit anymore in the Pandora (well, except for if we made it thicker and offered a new battery cover together with it).

Some more things I found in the Internet:

The Snapdragon 800 seems to be capped to using 4W max on tablets and 2W max on smartphones.

So while it's more powerful, it's limited to not use too much power.

The Intel z3770 has 2W SDP (typical use scenario).

According to Intel, that's running at low speed (1,4GHz) with normal work being done (no games, etc.)

2W would mean approx. 3W with all other components needed in a mobile system, and that would mean 5 hours battery life with a Pandora battery.

So approx. 5 hours with 1,4GHz normal usage.

Whatever SoC we choose - while they certainly all are better than the current Pandora in terms of Perf/W, they are all able to drain the battery within hours when being run with high-demanding tasks.
The one thing the Pandora is renowned for is long battery life. If Nova 3 and Reicast run fine on a dual-core at full fps in 1080p, then I don't see the necessity for a higher CPU performance or additional cores. The question to ask is are we future proofing or are we seeking bragging rights here?  
 
1. Does a Quad-Core A15 even make sense? That would lower the battery life to 1 hour... big.LITTLE wouldn't help here. Full Load is full load, it would use the A15 and not the A7.
These days, SoCs will usually let you run one core or two cores at higher clock speeds than all four cores. Yet it's still worthwhile to have more cores.

Hypothetical scenario partially/vaguely inspired by reality: say you're running DraStic on dual core Cortex-A15, where at 2GHz for same game it takes 3ms to do everything but render 3D. Now let's say 3D takes 4ms, so 7ms total.

Okay, now let's say that 3D was rendered at 4x resolution like PCSX-reARMed can now. And 3D now takes 16ms, so 19ms total. That's over the 16.67ms to be realtime speed, but 3D can be threaded, so it's now 11ms on one core and 8ms on the other core. So you run the two cores at 1.3GHz, and the second core sleeps for like 27% of the time.

Let's expand that to four cores, now it's 7ms on one core, 4ms on the other three cores. So you can run the four cores at around 840MHz, and three of them sleep for 43% of the time.

So which one uses more power? perf/W isn't linear and you're spending more time sleeping in the latter, so it's entirely possible that you use less power in the quad core scenario. But I'm not going to try to actually concoct some numbers to defend that precisely. It's also possible at that point that A7 cores could handle it too, depending on how high they clock. And there are going to be at least some games that are much more well threaded than an emulator like DraStic can be.

At any rate, you shouldn't assume all Cortex-A15 will have the same perf/W core for all cores. Look at nVidia's claims for Tegra 4 vs Tegra K1:

http://www.pcper.com/image/view/35618?return=node%2F59259

Both of these SoCs are made on TSMC 28nm processes. nVidia explains the improvement as being due to a move to using the HPM variant of the process, improvements ARM made in a newer revision of Cortex-A15 (that I can confirm OMAP5 isn't using), and improvements in nVidia's physical implementation of the CPU. From the numbers you're giving, it sounds like OMAP5's peak power consumption is a lot like Tegra 4's. I'm not saying that something like A80 isn't going to also be more like Tegra 4, but right now we don't really know.

OMAP5 may just be seeing availability now but it was supposed to be out in 2012. There's a good chance TI didn't do it for an HKMG process (like TSMC 28LP, or maybe a UMC 28nm process..); they do say it's 28nm but don't specify beyond that. So they could be at a power efficiency disadvantage.

2. Should I try to increase the battery size a bit? It could be possible and get us more capacity, but the battery won't fit anymore in the Pandora (well, except for if we made it thicker and offered a new battery cover together with it).
Is there any way a bigger battery could be optionally supported with an alternate battery cover?

The Snapdragon 800 seems to be capped to using 4W max on tablets and 2W max on smartphones.

So while it's more powerful, it's limited to not use too much power.
I doubt that there aren't tablets and phones that let it use more, perhaps substantially more than that. Those could just be recommendations based around some throttle point. I think I remember those numbers being higher anyway, like 5W for tablets.

The Intel z3770 has 2W SDP (typical use scenario).

According to Intel, that's running at low speed (1,4GHz) with normal work being done (no games, etc.)
Based on various numbers and measurements I've seen I'd estimate around 1W for 2.4GHz, which should outperform an OMAP5 A15 core at 1.7GHz while using much less power. Of course you can't run all four cores at 2.4GHz while using <= 2W, much less with the rest of the SoC involved.

2W would mean approx. 3W with all other components needed in a mobile system, and that would mean 5 hours battery life with a Pandora battery.

So approx. 5 hours with 1,4GHz normal usage.
No, not normal usage. 2W for all four cores running at 1.4GHz while the cooling meets some requirement (this isn't just a different way of specifying max performance - higher temperature reduced power efficiency, there's a non-linear feedback effect at play). Normal usage doesn't dictate running all four cores at even 1.4GHz. It can probably run one core at over 2GHz and still stay below 2W for the whole SoC.

Whatever SoC we choose - while they certainly all are better than the current Pandora in terms of Perf/W, they are all able to drain the battery within hours when being run with high-demanding tasks.
Indeed. It's all thanks to the advent of tablets and the increasing size of smartphones and phablets bringing increased battery capacity and thermal mass.
 
Last edited by a moderator:
My off the grid usecase is walking around listening to flacs, for which the m4 cores will lower the drain to almost nothing so long as i dont open up to change my playlist.
The M4 cores don't even use the same Instruction Set. Their use cases are limited.

-God Ginrai
 
Last edited by a moderator:
Apparently, Dolphin can only use up to 3 cores.

If A80 does what it meant to do, that seem like the best case

scenario.
 
Last edited by a moderator:
Not necessarily.

If we can make the battery cover thinner with more robust plastic and make the lid thinner (the LCD is thinner than before), we could be able to double the thickness of the battery which should be double the battery time.

That could give you about 24 hours battery time when running the speed of a 1GHz Pandora ,)

Something we need to check, but I can't say anything about that yet.
That would be AMAZING. Exactly what I want, I mean with this, Pyra will definitly be perfect and the ultimate portable device of all time.

I personally don't care about Pyra's thick, 3mm more or 3mm less is nothing, especially as the design will be better.

Any amount of A15 cores up untill you draw 2A at 5V  (10W) total, is a good thing.

A bigger battery standard means it will take longer to charge, i think a 12 hour forecast on an actual most-cases usecase sounds splendid. Good tradeoff.


With range anxiety, all you end up doing is being less efficient, dragging around a unused battery pack most of the time.

Where are people for hours playing games and running computing in the remainder cpu and gpu cycles? There is a limit to how long i can play N64, i have to sleep at some point...

Get the DC directly option to work and everyone is fine. Lets not stare ourselves blind at 2 hours runtime, a bigger battery takes longer to charge, and makes the device bigger, and then you end up carrying batteries for two platforms, nobody wins there.
I bet there is a way to reduce the time charge.

Your arguments are available if you go back at home everyday, but I occasionally don't go home for 2-3 days, sometimes unexpectly. The prospect to have a full pc in my hands which I don't have to care about battery is amazing.
 
Last edited by a moderator:
Sure, same as on the Pandora: Set the maximum allowed CPU frequency   But don't forget: It means FULL LOAD all the time with BOTH cores at maximum frequency. That's about 6x 1,2GHz Pandora with full load. Under full load, a Pandora has about 6 - 7 hours battery as well.
Then is this really an issue at all ? Most of the time the Pyra will not be under full load. 
That's exactly what I think.

But it doesn't really make much sense having even more A15 cores then, as they will give you more power but even less battery...

Well, except for when you're using it at home with an AC.
ED, yes. HOME USE is super important. This turns the Pyra into a full computer once you plug it on a full screen. I think this is very relevant when running plugged in, so A15 is a good choice when you are in home desktop mode. 

There are some interesting things that can happen when plugged in for example an Ubuntu system will be able to "converge" meaning that a powerful enough Pyra could be made into a full desktop machine with a lighter mobile setting and apps to go along with it.  Who knows what Canonical is going to be able to pull off here.

So Evil Dragon whatever you do please choose a SOC that has a very high chance of supporting Mir and overall more importantly Wayland a few years from now.  I think that it will be a major fail if it doesn't.  I suggest that you add proper Wayland support in the driver as a hard requirement to your SOC spreadsheet.
 
I'd like to see the higher max performance in Pyra, as long as I can enable a battery-saving profile to (pretty much) guarantee a >=10h battery life. Applications can then be tuned to these two performance profiles (TURBO mode like in ye olde PCs? :D ).
 
ED, I see this "2h under full load" as very problematic. Yes, the current software does not require this many resources, but games usually put the available "more" directly into fps (when thinking about native OGL games). So developers would have to implement framelimiting to make sure not too many resources are used. Honestly I don't see this happen. For emulation this might somehow work since there you can try to only emulate until "real speed" of the emulated device, but maybe give it a testrun on your desktop system and check how CPU usage behaves there when running some emulator, if it really tops out before making full use of at least one CPU core or not.
 
ED, I see this "2h under full load" as very problematic. Yes, the current software does not require this many resources, but games usually put the available "more" directly into fps (when thinking about native OGL games). So developers would have to implement framelimiting to make sure not too many resources are used. Honestly I don't see this happen
Ed already mentioned this could be handled on the OS side by limiting the resources allocated when on battery.  
 
Back
Top