the open pandora 2 must use an amd apu!


you know, I'd be all for an x86 cpu IF it can come close to the pandora's cpu power consumption.

http://boards.openpandora.org/index.php/topic/10738-pandora-power-consumption-breakdown/
Same power consumption at heavy load or same (or better) power consumption while providing the same performance a 600MHz (or 1GHz if you prefer) Pandora can? Because frankly I think Medfield can already do that - it has strong voltage scaling and power management in general and is on a much more advanced process technology than even the DM3730 in the 1GHz Pandoras. Of course newer ARM SoCs should meet this criteria too.

I think people get lost because they look at power consumption of devices running at full tilt, with sites like Anandtech showing power consumption in a bunch of benchmarks. I am however glad that someone is doing this, because for a long time that's the one thing no one did - and I suspected that Intel wasn't nearly as competitive at full load. They actually do a pretty good job, although the fight is far from over.

What would really be best for a new Pandora is something with big.LITTLE, IMO. A single Cortex-A7, if sufficiently clocked (say, 800MHz or so?) should provide about the same performance as at least a 600MHz Pandora in most tasks. That'll also use much less power. A big.LITTLE SoC would be nice, but if OPT is going for OMAP5 it won't have that. We'll still have to see what the power consumption on that is like. Exynos 5 has higher peak consumption than I was expecting, and the IPC improvement on Cortex-A15 wasn't as high as I expected either (IPC improvement can actually save power)
 
Last edited by a moderator:
you know, I'd be all for an x86 cpu IF it can come close to the pandora's cpu power consumption.

http://boards.openpandora.org/index.php/topic/10738-pandora-power-consumption-breakdown/
Same power consumption at heavy load or same (or better) power consumption while providing the same performance a 600MHz (or 1GHz if you prefer) Pandora can? Because frankly I think Medfield can already do that - it has strong voltage scaling and power management in general and is on a much more advanced process technology than even the DM3730 in the 1GHz Pandoras. Of course newer ARM SoCs should meet this criteria too.

I think people get lost because they look at power consumption of devices running at full tilt, with sites like Anandtech showing power consumption in a bunch of benchmarks. I am however glad that someone is doing this, because for a long time that's the one thing no one did - and I suspected that Intel wasn't nearly as competitive at full load. They actually do a pretty good job, although the fight is far from over.

What would really be best for a new Pandora is something with big.LITTLE, IMO. A single Cortex-A7, if sufficiently clocked (say, 800MHz or so?) should provide about the same performance as at least a 600MHz Pandora in most tasks. That'll also use much less power. A big.LITTLE SoC would be nice, but if OPT is going for OMAP5 it won't have that. We'll still have to see what the power consumption on that is like. Exynos 5 has higher peak consumption than I was expecting, and the IPC improvement on Cortex-A15 wasn't as high as I expected either (IPC improvement can actually save power)
wasnt it you that said batteries were cheap? so why not just offset higher power draw with large batteries? the ipad 4 is a good example.
 
Last edited by a moderator:
Cheap is relative, that only makes sense in some context..

It's all a tradeoff. Add bigger batteries and you make the device larger and heavier. But it's not just that. The higher the peak power consumption of the device, the more cooling it has to have in order to prevent it from overheating (note that actually being comfortable to hold is an even stronger requirement since the device may get uncomfortable long before it overheats). You don't want it to have a big heatsink like the Shield, much less do you want a fan, and an all metal chassis that acts as a heat spreader is expensive.
 
Cheap is relative, that only makes sense in some context..

It's all a tradeoff. Add bigger batteries and you make the device larger and heavier. But it's not just that. The higher the peak power consumption of the device, the more cooling it has to have in order to prevent it from overheating (note that actually being comfortable to hold is an even stronger requirement since the device may get uncomfortable long before it overheats). You don't want it to have a big heatsink like the Shield, much less do you want a fan, and an all metal chassis that acts as a heat spreader is expensive.
in any case what do you think about the a6-1450 apu(tamash)?
 
in any case what do you think about the a6-1450 apu(tamash)?
You sure are asking me that question a lot! I don't want to say a lot about chips that aren't released where you don't have much to go on beyond marketing because marketing can often be very misleading. In this case they don't seem to be saying anything about CPU frequency or even GPU frequency, nor have they said anything concrete about power consumption (I have seen some source claim 5.9W). I've watched the Jaguar presentation and they've made some nice improvements but I doubt it'll be encroaching on Clover Trail's perf/W levels for CPU. The GPU could possibly be industry leading both in perf and perf/W but AMD needs to learn they can't sell products with strong GPUs alone..
 
in any case what do you think about the a6-1450 apu(tamash)?
You sure are asking me that question a lot! I don't want to say a lot about chips that aren't released where you don't have much to go on beyond marketing because marketing can often be very misleading. In this case they don't seem to be saying anything about CPU frequency or even GPU frequency, nor have they said anything concrete about power consumption (I have seen some source claim 5.9W). I've watched the Jaguar presentation and they've made some nice improvements but I doubt it'll be encroaching on Clover Trail's perf/W levels for CPU. The GPU could possibly be industry leading both in perf and perf/W but AMD needs to learn they can't sell products with strong GPUs alone..
it is a similar case to the ipad isnt it, apple focuses on gpu over cpu(dual swift core instead of quad) and it has worked out very well for apple.
 
it is a similar case to the ipad isnt it, apple focuses on gpu over cpu(dual swift core instead of quad) and it has worked out very well for apple.
I don't agree with this one bit, Swift is a highly accomplished design and Apple has put far more resources into developing it than they have into licensing PowerVR cores. It doesn't have quad cores because Apple (rightfully) feels that they can't get enough benefit from quad cores to justify the die area needed. Others are focusing on quad cores in ARM SoCs more out of marketing appeal than real benefit, and Apple doesn't play by the same marketing rules.

In AMD's case it's not that they're not trying, it's just that their marketing makes GPU out to look more critical than it really is.
 
it is a similar case to the ipad isnt it, apple focuses on gpu over cpu(dual swift core instead of quad) and it has worked out very well for apple.
I don't agree with this one bit, Swift is a highly accomplished design and Apple has put far more resources into developing it than they have into licensing PowerVR cores. It doesn't have quad cores because Apple (rightfully) feels that they can't get enough benefit from quad cores to justify the die area needed. Others are focusing on quad cores in ARM SoCs more out of marketing appeal than real benefit, and Apple doesn't play by the same marketing rules.

In AMD's case it's not that they're not trying, it's just that their marketing makes GPU out to look more critical than it really is.
that makes sense, but isnt AMDs overall strategy to move people to the gpus(with opencl) so that they can easily move to HSA?
 
that makes sense, but isnt AMDs overall strategy to move people to the gpus(with opencl) so that they can easily move to HSA?
Yes and this is exactly what I'm referring to, AMD is putting too high hopes and expectations into GPU integration. People aren't migrating to GPU compute nearly as much as they need them to be.


Even if it could or even does end up being this way in the future you can't win technology now on the basis that it needs software to catch up, especially if you don't have the market presence to strongly influence software.
 
Last edited by a moderator:
that makes sense, but isnt AMDs overall strategy to move people to the gpus(with opencl) so that they can easily move to HSA?
Yes and this is exactly what I'm referring to, AMD is putting too high hopes and expectations into GPU integration. People aren't migrating to GPU compute nearly as much as they need them to be.


Even if it could or even does end up being this way in the future you can't win technology now on the basis that it needs software to catch up, especially if you don't have the market presence to strongly influence software.
from what I understand you are saying, amd is on shaky ground, pressure from the top(intel) and pressure from below(lack of industry motivation to use opencl), could hsa put amd out of business even if it is an excellent product(assuming the infrastructure still isnt there)?
 
from what I understand you are saying, amd is on shaky ground, pressure from the top(intel) and pressure from below(lack of industry motivation to use opencl), could hsa put amd out of business even if it is an excellent product(assuming the infrastructure still isnt there)?
If AMD can't make HSA succeed then we'll never be able to tell if it was an excellent product or not. Actually I'd go one step further and say that if it doesn't succeed it wasn't an excellent product.

AMD's hands are pretty much tied though. They bet the farm on GPUs long ago (starting with a huge overpayment for ATi) and for better or worse this is the path they're stuck with.
 
from what I understand you are saying, amd is on shaky ground, pressure from the top(intel) and pressure from below(lack of industry motivation to use opencl), could hsa put amd out of business even if it is an excellent product(assuming the infrastructure still isnt there)?
If AMD can't make HSA succeed then we'll never be able to tell if it was an excellent product or not. Actually I'd go one step further and say that if it doesn't succeed it wasn't an excellent product.

AMD's hands are pretty much tied though. They bet the farm on GPUs long ago (starting with a huge overpayment for ATi) and for better or worse this is the path they're stuck with.
I'm running one of the older ones and it's quite nice. Zacate e-350, and I love it. I think the real question is whether or not OpenCL can take off quickly enough to provide a difference. Personally, I refuse to buy Intel chips just on principle, unfortunately, I'm clearly in the minority.

The way things are moving is to HSA, it's just a question of who's going to be left to design and produce the chips.
 
I would love to see this chip in the Pandora 2, AMD's APUs really are great chips. I'm using a Quad-Core A6 in my laptop and it works quite well, even with a 64-bit OS, and being able to run FULL operating systems like Windows on the Pandora would be absolutely amazing.The only problem is taking care of all that heat... 
 
Last edited by a moderator:
Linux is a "FULLer" operating system than Windows by a long shot, however yes it would be nice to run Windows / x86 games on it if it ran well.

I predict we will stay with ARM, which makes me happy.
 
Linux is a "FULLer" operating system than Windows by a long shot, however yes it would be nice to run Windows / x86 games on it if it ran well.

I predict we will stay with ARM, which makes me happy.
When I said "full" i was mainly referring to one written for a standard architecture-- like x86. The GNU/Linux that the Pandora runs isn't a fully-fledged OS in that sense.

But I agree, Linux has far more to it than Windows. 
 
actually the GNU/Linux that Pandora runs is a portable operating system, that runs on dozens of different CPU's.  It makes windows look like a bad joke.
 
Last edited by a moderator:
Much of the stuff available as a pnd was originally written with desktop (mostly x86) linux users in mind

SL4P is very similar to slackware for Desktop PCs.

Firefox, GIMP, Blender, VLC, are all primarily desktop applications, and all run on pandoras

As Dredd says, GNU/Linux itself is beautifully portable
 
Back
Top