Overview of SoC options


And the drivers do get commits from people who don't work at Intel, you can get an idea from the e-mail addresses.. http://cgit.freedesktop.org/mesa/mesa/log/ None of the last 10 people to commit have Intel e-mail addresses. One of them is from AMD though!
Yes but take a look at those non-Intel commits themselves - it's mostly maintenance work like compiler support, API changes, macro renames and so on. It's what I mentioned as Linux graphics stack support.

On the other hand there is significant sponsored work on it by Google and even curiously AMD, that is of course good for them.
 
Last edited by a moderator:
Jebe there isn't an SoC on this list that won't consume > 2W under heavy load, and the most power hungry ones there are all being put in phones. There are known cases of games draining phones with 8+Wh batteries in well under 2 hours.
Without a lot more information that doesn't really demonstrate anything to counter my point.  LEDs are quite efficient in producing light from electricity, so while they do generate waste heat a good portion of their power consumption is being turned into light.  The actual LCD part of an LCD should be similar.  Speakers similarly will be have a significant amount of their power use turned into pressure waves, and antenna have a significant part of their power use turned into EM signals.  All of which, as you know, contribute to power consumption in addition to the SoC.

If you have direct evidence of a phone being able to handle a sustained 2 Watt thermal load, and how they're making that work then go ahead and bring it forward.  Or for that matter data on the Z3770 that isn't related to SDP which by Intel's own admission cannot be assumed to be where they set the TDP governor aka cTDP, let alone turbo throttling potential and should be assumed to be rated lower then both figures.

There also the issue that it's not uncommon to have metal cased phones like the iPhone 5S where case conduction turns the entire back of the phone into a heat spreader which also, like most Apple products, conducts heat to the user using them as a de facto heat sink.  Needless to say double digit liters of blood plasma can handle 2 Watts with the very high specific heat associated with what makes up the majority of blood plasma.  Their laptops however have some known issues with using that trick of using the user as a heat sink associated with the higher thermal load of those systems.

I doubt OpenPandora GmbH is in the position to be using precision milling machines with fancy aluminum case designs.  If you want to bring some more specific examples where you can bring in the SoC power numbers then we can look at that.  Otherwise thermodynamics isn't something where you can cheat the reaper so I'm not sure what you're trying to prove here.

A Pandora is better equipped to handle heat dissipation than a phone because its display isn't sandwiched against the rest of the thing.
The waste heat of the display should be relatively minor compared  to the waste heat associated with discharging through a large battery.  You're ignoring one of the biggest players while obsessing on something that's not necessarily even a major player.  Heck it might actually be serving as a sink in some cases given how they're constructed normally, LED efficiency, and that they have a clean radiation window to the outside world.

Z3770 has a base speed of 1.46GHz. Passively cooled tablets can run all four cores at the max turbo of 2.4GHz for fairly lengthy periods of time.
This is where I have to ask what your reasoning is that this makes a meaningful case.

The Z3770 is basically designed for 10.1" tablets and when push comes to shove can be crammed into 8" systems.  Using the iPad 4 as a reference it has a back which if flat would have an area that's 18 times my theoretical 5cm square wafer, which is large for a handheld, and has even more of an advantage as it's curved nature means there's even more area.  Naturally with case conduction getting the heat to the outside and there being quite a number of grams of Aluminum involved meaning it also takes a while to raise the temperature significantly.  To radiate 2 Watts of heat with even that flat back if my diseased state isn't making me mess things up should be possible at well below room temperature.

@ Jebe : You certainly seem to have a good grasp of thermal dynamics, if you could comment on the practicality of using no cooling solution for a current gen SOC in this thread it would be much appreciated.


Could you explain a little more as to why in your opinion a passively cooled SOC would work fine in a tablet / phone but not in a handheld mini computer?
I'm sick right now, I'll look it over during the weekend.  I don't necessarily agree the question is valid in the first place given there's not exactly a wealth of Tablets and Smart Phones demonstrating the ability to mach or outmuscle the Shield on a sustained basis rather then throttling in a short period of time, or anything else to substantiate the other inherent assumptions in that question.
 
Last edited by a moderator:
Power numbers under full load:

Galaxy S4: 5.1W http://www.notebookcheck.net/Review-Samsung-Galaxy-S4-GT-I9505-Smartphone.92829.0.html

Xperia Z: 6.6W http://www.notebookcheck.net/Review-Sony-Xperia-Z-Smartphone.89509.0.html

Lumia 920: 5.4W http://www.notebookcheck.net/Review-Nokia-Lumia-920-Smartphone.86872.0.html

HTC One: 4W http://www.notebookcheck.net/Review-HTC-One-Smartphone.90716.0.html

Nexus 4: 6.5W http://www.notebookcheck.com/Test-Google-Nexus-4-Smartphone.85941.0.html

You can compare the full load vs idle numbers to get a fairly rough idea of how much the SoC alone uses, which is: 3.9W, 2.9W, 1.5W, 2.9W, 5.1W

If a Pandora can't handle the cooling necessary for a 2+W SoC then they can't handle a current SoC. Doesn't matter if it's from Intel or not. It doesn't matter how much LED lit displays dissipate in phones, because no matter what it'll be more than is relevant on Pandora, which is 0W (and it's hardly insignificant, even red LEDs are < 40% lighting efficacy). I'm not "ignoring" the battery because phones are sandwiched behind it just the same, and having a lower capacity battery doesn't if the charge is the same or higher. If anything, distributing the battery over a wider surface area only helps.

I don't think you get what I was saying with the Z3770's ability to sustain 4 cores at max turbo. It uses ~2.5W while doing this. The reason why this is relevant is because it doesn't need to do 4x2.4GHz to be useful in a Pandora-like device. Other current phones will generally throttle if you try to run all of the cores at max clock speed for more than a short duration. And they get away with this because outside of benchmarking you rarely want to run four cores at full clock speed in a phone. I guarantee that it won't be much different in Pandora. Just having one of those BayTrail cores run at the full 2.4GHz is enough for most uses, although it can probably manage two without blowing Pandora's power budget. So what if there are more cores there and you can't run them at full clock for very long? That's not a problem if you're constantly power gating them because they're not used heavily. And if you use 100% CPU time on all of them you're still going to be able to achieve higher perf/W than you would be able to with just two cores even if you have to lower clock to keep the same power target, because perf/W isn't anywhere close to linear - having more cores is more efficient. GPUs thrive on this principle.

If you hate BayTrail for using four cores then strike off all of the other four core SoCs from consideration, which is most of the ones on this list. If you hate BayTrail because you think the perf/W/core is inferior to the competition (especially those using A15s) then you're just wrong.
 
Last edited by a moderator:
I wonder are there any way to increase Pandora's exposure?
Even if the last 500 would not go into production soon, I doubt Ed could reach sales figures that would give him a foot into the door of one of the companies that currently have no interest in supplying him.
 
Any data on the failure rate of gadgety smartphones after warranty period has expired?

Asking because I've often wondered whether those things are just designed to "live fast, die hot" :).

At least kids seem to be changing to new models faster than they change their socks.

(And I'm quite afraid in some cases you can take that one literally.)

I'm not outright saying running apps which drain the battery in about the same time it takes to charge it fully would somehow diminish the life expectancy of said devices, I've just wondered whether it's really healthy for them in the long run. I definitely would not count on Apple to invest on longevity in any of their products...
 
I still think we could transport a significant amount of heat to the front edge of the device (center 3/4 that doesn't touch hands) and radiate it to the air through a passive fined grill.
 
I still think we could transport a significant amount of heat to the front edge of the device (center 3/4 that doesn't touch hands) and radiate it to the air through a passive fined grill.
I was just kidding.  I'm happy as long as I get pass 55 :)  so a little EMF radiation in my body won't kill me before that :)
 
Last edited by a moderator:
Going by the numbers from your own post, the only Smart Phone with an implied thermal load north of 1.5 Watts associated with the SoC are the Xperia Z and Lumia 920.  Something that needless to say doesn't substantiate your premise any modern SoC must be housed in a device that can deal with over 2 Watts of sustained thermal dissipation.  However checking the links under a critical reading of the data only leaves the Sony Xperia Z as a candidate for a potential SoC that's north of 1.5 Watts, and if that back of the napkin calc is right it would be in the territory the Shield's SoC is in. 

I'm all ears for why you think nVidia is overeacting with their decision to use passive/active convection cooling with a finned heat sink, and you actually playing up Sony's approach and verifying the implication is actually reality.

If a Pandora can't handle the cooling necessary for a 2+W SoC then they can't handle a current SoC.
Your own data rejects this claim.  As does Intel about their own product line, for that matter.  The Z3770 is not slated for phones, Merrifield is.  Are really effectively accusing them/OEMs of being incompetent for not cramming parts designed for 10.1" tablets into Smart Phones?  Because that's effectively what you're outlining.

Doesn't matter if it's from Intel or not. It doesn't matter how much LED lit displays dissipate in phones, because no matter what it'll be more than is relevant on Pandora, which is 0W (and it's hardly insignificant, even red LEDs are < 40% lighting efficacy).
...  In case you failed to notice most LCD panels are wrapped in insulators except for the part that's emitting light, and they're usually cool to the touch despite that.  Similarly a number of the phones you're bringing up are insulator wrapped except for that screen.  Exactly how are you imaging the heat gets out, with what in your little world is a major heat generator?  Magic?

I take it the mental brickwall prevents you from contemplating the LEDs actually have the potential to act as a cooler, and the case consumption screen's backing plate has 68.9 sq cm worth of area to play with.  The window used to emit visible light photons not exactly being a wall to IR photons.  Due to the nature of the scaling of thermal radiation with regards to temperature, assuming ideal black body conditions and no reradiation, with temperature lands the temperatures at below freezing for 2 Watts, and well over the implied SoC consumption at the given idle temperature.

Obviously the backing plate is not a blackbody radiator, and the real world doesn't approximate conditions where re-radiation can be ignored, but you should be able to see the difference.  As I've said if you have solid trustworthy data on the Z3770 present it with citation, please. 

You're not exactly in a position to expect me to take your word on it when you're invoking heat dissipation fairies this week with questionable scholarship where you have cited.  Especially when your response to questioning the wisdom of trying to cram an SoC designed for 10.1" tablets into a ~5" handheld, rather then one the vendor designed for ~5" handsets, is to try to move goal posts and try to pretend ~5" handsets use 10.1" tablet SoCs, and the screen you're playing as evil heat producer of DooM is the only way you could have tried to substantiate that notion.  Yeah, okay...
 
Last edited by a moderator:
I didn't say that a phone housing HAD to deal with 2W, but in this case the entire phone uses 6.6W so of course it's handling > 2W. I know you think that a large percentage of that 6.6W is light and not heat but that really isn't true at all. That 1.5W number for the Xperia Z is suspicious, when compared against other phones there with the same or similar SoC.. namely, the idle value it gives is just really high. Which makes it seem like it's not really true idle (ie, some crap is running in the background despite their efforts against it). Mind you, this is max idle vs max load so that could be due to fairly intermittent spikes. What YOU were saying is that a Pandora would melt under > 2W and I'm not seeing how that's true. I don't know what you're going on about with blackbody radiation and insulation

Intel will make Merrifield for phones, yes. There can be a lot of reasons for this, for example we don't even know what embedded modem options it'll have. Pandora isn't even a phone. And you keep bringing up 10.1" tablets, as if BayTrail-T hasn't already been announced to be in smaller tablets. Like these: http://www.zdnet.com/acer-announces-iconia-w4-8-inch-tablet-with-intel-bay-trail-windows-8-1-7000022095/ http://www.zdnet.com/lenovo-unveils-miix-2-8-inch-tablet-bay-trail-and-office-for-299-7000022027/ http://www.geek.com/microsoft/windows-8-1-on-a-7-inch-bay-trail-tablet-is-a-peek-into-the-future-1557789/

Think way back, this was more or less Pandora form-factor and similar purpose: http://www.engadget.com/gallery/ocosmos-ocs-1-oak-trail-gaming-handheld/ Using Oak Trail, intended for tablets instead of phones (Moorestown was meant for phones). BayTrail-T is far more power efficient than Oak Trail was. This device never saw the light of day, but Intel had to have been willing to supply Oak Trail for it. Which is why I think there's a chance they'd be willing to supply BayTrail-T for Pandora. But it really doesn't matter, the question is if it makes sense and all data points to it. A bunch of ARM SoCs are getting put in both phones and tablets alike. They just have different max clock settings depending on where they are. Tegra 4 is a great example of this. nVidia uses a finned heatsink and fan for Shield and no I don't think they're overreacting. At the same time I don't think Xiomi and ZTE are underreacting by putting Tegra 4 in phones that most certainly do not have either of the above. They do it by having somewhat more modest peak clock settings. If the same Tegra 4 model is allowed to scale to accommodate different levels of cooling why exactly can't BayTrail-T do it? With configurable TDP settings and turbo boost Intel is well prepared for scaling, just like they were when the same Ivy Bridge SKUs were put in both high end tablets and laptops.

So you say a mental brickwall prevents me from contemplating that the display is acting as a cooler? Just because you found a research article about LEDs having > 100% efficacy doesn't mean it has any bearing on what we'd be looking at here... which is under 40% like I said. The very second sentence of your article shows it has no relevancy beyond extremely low power conditions.

Yet again I'll post the Anandtech article that says 2.5W for a heavy CPU load with all four cores at 2.4GHz (Cinebench) http://www.anandtech.com/show/7314/intel-baytrail-preview-intel-atom-z3770-tested/2
 
This is interesting to play with, but I'm pretty sure I'm doing it wrong.

http://www.daycounter.com/Calculators/Heat-Sink-Temperature-Calculator.phtml

Does anyone have a good understanding in how to use that or have another heat sink sizer?

If we use a heat spreader to move the air to a radiator/sink on the outside edge of the case, how big of a heat sink (radiator) would we need to handle 3W while keeping the heat sink temperature < 43*C in an 25*C ambient room?
 
On the other hand there is significant sponsored work on it by Google and even curiously AMD, that is of course good for them.
Intel is the driving force behind Mesa, they've got a whole horde of developers working on all that stuff (which seems to often cause regressions with certain older Intel GPUs, however). Even though they still refuse to make use of Gallium, many improvements are being ported rather quickly over to Gallium, which heavily benefits the other free drivers.

As far as I can tell AMD's radeon developers seem to be generally less bound to their company (there are actually not that many people around, anyway. One of the largest contributors still comes from Red Hat), and improvements to Mesa help everyone involved. After all, it's not like there's any serious competition between those two on the GPU market.

Google once had a lot of plans with low power Intel solutions - they used to develop a Gallium Intel driver. That work was just recently picked up again by some other guys, IIRC. And from what I've heard some time ago, Google actually gives their programmers some paid time to work on any open source project they want to - besides all the tracking and spying they still somewhat try to remain the "good guys".

Overall I've got pretty much the impression that those guys are much more open minded than one would think, they are actually being paid to support the idea of free software instead of following financial motives.
 
Last edited by a moderator:
Hi,

Does anybody think about allwinner ICs? if you compare all the ARM based SOCs they seems to be the more linux friendly, the sunxi devs are very active, even allwinner is considering supporting officially linux. In my experience the SOC performances come more from the software support than the spec. For example I don't see much differences between my Odroid-U2 and an A20-OlinuXino(openpandora and pandaborad comparison is also interesting). And if I want to compile the kernel, U-boot... it's more stable and easy on the allwinner side. I would add that if you want to keep it portable using 8 core SOC is quite useless specially if the support is bad and you always underclock them. My last point, the allwinner SOCs are cheap https://www.olimex.com/Products/Components/IC/A31/ and very integrated, you need few components around.

Best regards,

Alexis.
 
^ You're a bit late to the party i'm afraid, ED's going with a OMAP5 SOC.
I hope it's not because I'm not even sure TI isn't really wanting to sell anymore OMAP SOCs, it seems they moved to others family, one of my colleague used to work with the french office (developing OMAP) told me that they didn't want to continue OMAP and now it's more clear, the French office is closed.
 
Just my 2 cents:

We need a SoC that can offer all the features we want while not killing the battery.  

Pandora 2 needs to have features and benefits to beat out devices like the Nvidia Shield and Archos GamePad 2.  Running a full desktop Linux, having a keyboard, and being somewhat pocket-able wasn't enough to sell the device especially at the price point it is being retailed for.  It has to offer much more than just that this time around.  ED and the guys need to really put things together in a very cost effective manner.  A 600 dollar Pandora 2 isn't going to cut it.  Look at how the $349 price tag for Nvidia Shield was received, which soon was dropped to $299 due to consumers reactions.  To sell a product it is about cost versus features.  To give you an example: Pandora versus Nvidia Shield....while the Pandora has a full desktop environment and keyboard how many people could justify those features over the much faster cutting edge and 1/3 price Shield? No one could.

Point I am trying to make is if the SoC in the Pandora 2 isn't fast enough to compete with some of the quad-core phones then most consumers aren't going to give the P2 a second look unless it cost $249 or something cheap like that.

My personal Wish list;

Quad-Core SoC

Micro HDMI or Display Port

2 or More GB of RAM

16GB Built in Storage with 100% compatible Android 4.3+ and Full Desktop Linux dual-boot capability

At least 1 SDXC compatible slot

1 USB 2.0 or 3.0 

Charge Via micro USB

The device has to be pocket-able or you open yourself up to compete against the Shield, Archos Gamepad 2, JXD 7800b and whatever else has been released. From a price point perspective I don't see the P2 being able to compete with those devices....but please prove me wrong.  
 
Back
Top