How about a next gen Atom for the Pandora 2 ?

How about a next gen Atom for the Pandora 2 ?

  • Great idea !

    Votes: 4 8.2%
  • Only if it gets the performances of a 2 GHZ Core 2 Duo and at least 5 hours runtime

    Votes: 8 16.3%
  • Either way, NO !

    Votes: 37 75.5%

  • Total voters
    49

i hear so many reasons from exophase why this and that shouldnt be used. how about an advice on what TO use around 2013 in pandora form?
 
K, you got a point that they will most likely be Tegra 4 and 5.
But if the three is 10x the power of the tegra 2, why would the 3 be 50 times faster then the 2?


If the Tegra 4 is 10 x faster then the three, then it would would be a factor 100 faster then the 2.

Did you not read where I said Tegra 3 is nowhere close to 10x as powerful as Tegra 2 under any reasonable metric? Its CPU performance is at best about 3x, in practice a lot less since you can't easily utilize all of those threads with full performance scaling and because they didn't increase the memory bandwidth nearly enough (or the L2 cache at all).. Its GPU is also at best around 3x as powerful when you combine clock speed and execution resources. This 10x number is totally baffling, and along with other questionable or outright fraudulent claims nVidia has made, such as their comparisons with Core 2, it makes me take what they say without much confidence.

i hear so many reasons from exophase why this and that shouldnt be used. how about an advice on what TO use around 2013 in pandora form?

How about what they already said they probably will use, which is OMAP5 or Exynos 5?
 
Last edited by a moderator:
Voted for 'Only if it gets the performances of a 2 GHZ Core 2 Duo and at least 5 hours runtime'.


Bloody hell, a pig just flew past my window!
 
Somebody might have told you this already BUT. Tegra is a SOC.. which includes a Cortex-A9 CPU much like the OMAPS. The GPU is nvidia based ULP.


T30L/T30/T33 - Each version is slightly better than the last.


When the Pandora 2 is being prepared, they'll find the best SOC which is available at the time... saying the current Tegra will be outdated is pointless because if it's the best they can get. it'll still be the chip they use. The next best thing is already around the corner
 
Last edited by a moderator:
but hteir are a bunch of atom skus...e series 2.7W tdp, n series n2600 3.5W tdp and amd gt 16R 2W tdp...arm may have better sleep and idle power consumption but when push comes to shove tpds get high and battery life is that much better than x86 parts.
 
but hteir are a bunch of atom skus...e series 2.7W tdp, n series n2600 3.5W tdp and amd gt 16R 2W tdp...arm may have better sleep and idle power consumption but when push comes to shove tpds get high and battery life is that much better than x86 parts.

When Pandora lasts 10 hours that means it's consuming only around 1.5W. For the entire thing - not just the SoC, but the memory and display are large additional consumers. The SoC would have to be using well under 1W.


Looking at your other x86 examples the perf/W is just not that good. Your Tunnel Creek 2.7W example is just a 600MHz single core Atom which is pathetic. AMD's G-16TR is actually 4.5W TDP (the 2.3W "average" claim could mean anything, I never take those claims seriously) and is similarly highly gimped, running only 1 Bobcat core at 615MHz. N2600 at least has some decent capability but 3.5W is just too high. And I can see a 1.5-2GHz dual or quad core Cortex-A15 SoC easily beating it, in fact 1.5GHz dual or quad Cortex-A9 SoCs probably beat it when not running heavily x86 biased benches like Javascript ones.


The only x86 SoC that could fit power consumption is Medfield and right now that's just one core with a base speed of 1.3GHz and a pretty poor GPU, again I'd much rather wait for Cortex-A15 SoCs. Sure Silvermont (Merrifield SoCs) could be a good fit but we're talking mid-2013 before those are released. You can't plan a Pandora 2 release late 2013 around an SoC that has only been ready for a few months, that just doesn't work.
 
Last edited by a moderator:
but hteir are a bunch of atom skus...e series 2.7W tdp, n series n2600 3.5W tdp and amd gt 16R 2W tdp...arm may have better sleep and idle power consumption but when push comes to shove tpds get high and battery life is that much better than x86 parts.

When Pandora lasts 10 hours that means it's consuming only around 1.5W. For the entire thing - not just the SoC, but the memory and display are large additional consumers. The SoC would have to be using well under 1W.


Looking at your other x86 examples the perf/W is just not that good. Your Tunnel Creek 2.7W example is just a 600MHz single core Atom which is pathetic. AMD's G-16TR is actually 4.5W TDP (the 2.3W "average" claim could mean anything, I never take those claims seriously) and is similarly highly gimped, running only 1 Bobcat core at 615MHz. N2600 at least has some decent capability but 3.5W is just too high. And I can see a 1.5-2GHz dual or quad core Cortex-A15 SoC easily beating it, in fact 1.5GHz dual or quad Cortex-A9 SoCs probably beat it when not running heavily x86 biased benches like Javascript ones.


The only x86 SoC that could fit power consumption is Medfield and right now that's just one core with a base speed of 1.3GHz and a pretty poor GPU, again I'd much rather wait for Cortex-A15 SoCs. Sure Silvermont (Merrifield SoCs) could be a good fit but we're talking mid-2013 before those are released. You can't plan a Pandora 2 release late 2013 around an SoC that has only been ready for a few months, that just doesn't work.

IF you get specs and samples in time i dont see how you could not plan ahead of a soc release, most company's do that to stay ahead of competition.


But i agree with waiting for the cortex, only question remains, which manufacturer, tegra, omap, snapdragon etch.
 
Last edited by a moderator:
IF you get specs and samples in time i don't see how you could not plan ahead of a soc release, most company's do that to stay ahead of competition.

It's pretty obvious that NOBODY out there is considering x86 for small mobile devices. That's also why Microsoft is developing W8 for ARM. It's just the best choice at the moment.


ARM has a proven track record in power consumption/power.


Besides, it takes time to design/build/qualify a system, you don't get samples just a few months before release. Why do you think Apple doesn't release a new iPad every 3 months?
 
IF you get specs and samples in time i dont see how you could not plan ahead of a soc release, most company's do that to stay ahead of competition.

Go look at phones and tablets, look at when they release now look at when the SoC manufacturer says they were available. "Specs" are not good enough and pre-release engineering samples are often not really available that far ahead of release and when they are often have various errata. If they're available to the manufacturer in question to begin with; they're a very limited resource and small players will have a harder time getting them.
 
Indeed, this question has to be asked, but:


x86 is crap, Atom is crap, Intel is crap.

I have to agree, but the boost in software support might really worth it assuming they could prevent it from melting and get a decent batterylife without costing an arm and a leg...
 
With some next gen ARM cpu you can probably use an x86 emulator and still have a reasonable amount of horsepower left after all the emulator overhead. The problem of "software support" is that there's a huge amount of old non-open source x86 binaries out there; my guess is that at some point in the near future, most of the new software will also have ARM binaries that might be made to run without needing emulation. Things would of course be much simpler if everything would be open source.


Anyway, I don't think Intel will produce anything sufficiently low-power anytime soon.
 
IF you get specs and samples in time i don't see how you could not plan ahead of a soc release, most company's do that to stay ahead of competition.

It's pretty obvious that NOBODY out there is considering x86 for small mobile devices. That's also why Microsoft is developing W8 for ARM. It's just the best choice at the moment.


ARM has a proven track record in power consumption/power.
Ahem!


vs05-01_10-23-34x1020_gallery_post_large_verge_medium_landscape.jpg
 
Yes, there are Medfield phones.

The only x86 SoC that could fit power consumption is Medfield and right now that's just one core with a base speed of 1.3GHz and a pretty poor GPU, again I'd much rather wait for Cortex-A15 SoCs. Sure Silvermont (Merrifield SoCs) could be a good fit but we're talking mid-2013 before those are released. You can't plan a Pandora 2 release late 2013 around an SoC that has only been ready for a few months, that just doesn't work.

The further question I'd ask is, why use Medfield now? What's so great about it? No one is running proper Windows on those phones, I'm not sure you even can. From where I stand it's not the current top of the line in CPU capabilities and it's much further behind in GPU capabilities. Yeah there are a few places where there's still a big performance deficit for ARM over x86, like Javascript, but I don't think you'll find that's the case for most of the stuff people are running on a device like Pandora, which is decently optimized for ARM.


You could probably get WINE going on it, I'd consider that the only real plus side. Somewhat diminished when qemu+wine gains more steam.
 
Last edited by a moderator:
IF you get specs and samples in time i don't see how you could not plan ahead of a soc release, most company's do that to stay ahead of competition.

It's pretty obvious that NOBODY out there is considering x86 for small mobile devices. That's also why Microsoft is developing W8 for ARM. It's just the best choice at the moment.


ARM has a proven track record in power consumption/power.
Ahem!


vs05-01_10-23-34x1020_gallery_post_large_verge_medium_landscape.jpg

And what device is that? Specs?
 
The state of mobile device benchmarking is very poor right now. Most performance benchmarks are purely synthetic, and worse a lot of Javascript stuff like Sunspider. Most power consumption benchmarks only test periods of low CPU utilization like cell phone talk and casual web browsing, and standby time. Most graphics benchmarks only test pre-baked scenes and not real games (while these are good to some extent for testing GPU capability they do nothing to test the typical CPU load of a real game).


Pandora owners will surely want to know long it can run and how well it can run while playing games and emulators, both of which can eat a lot of CPU.


It never should have been a question on anyone's mind that Intel can make Atom SoCs that don't use a lot of power when you're barely using them. But we don't really know what they're like when you are. The closest we have is Intel's claims, which is about 750mW at 1.6GHz w/o Hyper Threading. This number could be alright, but that's assuming it's really true - when we saw Moorestown in a few devices the battery life was horrifying and fell way below what you'd expect from Intel's values.
 
Their x86 architecture was fusked up from the very beginning so it's simply hopeless.
 
I can kinda see wanting to run windows and mainstream software on a Pandora - but No, ARM makes so much more sense!
 
Back
Top