GPD Win (x86 Computer / Palmtop)

[EvilDragon]

I'm sure, future (and more powerful ) devices like Pyra & Co. will need at least a metal case. Modern Smart Phones benefit alot from this, even without a active Fan. And these Phones are already much faster than the OMAP5. AFAIK the new iPhone SoC's are real beasts.

The metal case won't solve the heating issue, it will just delay the time it takes to heat up fully.
So instead of 10 Minutes running full speed, you get maybe 30 Minutes.

It's not a fix, it's just an improvement.

More modern SoCs help, of course, as they produce less heat for more cpu power than older ones, so they can do more demanding things for less heat (and power usage).

But still, if you use their full power for a longer time, they will heat up.

And while CPUs have become faster and faster during last year, the heat issue has become worse, and a lot modern CPUs are being throttled after a few minutes of full usage already, to prevent overheating.

 

[levi]

Yeah, calling the thing you put on a CPU in a desktop machine (for example) a heatsink is a bit of bad english. It's really a radiator, but nobody calls it that (and even calling radiators radiators is a misnomer, cos they don't radiate heat, they heat air). These sort of solutions in these handhelds are more like sinks for heat if you don't evacuate the heat via fans or holes or what have you, and that can be a useful thing to have if your power usage fluctuates having only brief spikes of power usage, but for any real power usage you need some way to convect or radiate heat out of the unit.

You know all that of course, I'm mainly moaning about stupid english terms for things.

 

[Exophase]

However, I'd say that NO SoC that's as powerful or more powerful than the A15 can run on full speed for a longer period of time without getting really warm.

One Cortex-A72 core at 2.3+GHz on TSMC's 16FF process uses less power than the Pandora's Cortex-A8 at 1GHz. Running them at the equivalent of 2x1.5GHz Cortex-A15 performance levels would probably also use less power than Pandora's CPU.

Intel's 14nm Atoms are also likely far more efficient (on a perf/W basis) than Cortex-A15 in OMAP5.

Yes, pretty much every phone out there is using SoCs that can't sustain anything close to full clock speed when all the cores are loaded, but this is with four or even more cores. That's a lot different from 2 cores and they're at much higher than OMAP5 levels.

 

[EvilDragon]

Yes, that's what I said. New CPUs are more efficient and provide more power for the same heat, but they'll all heat up when they are driven to the max.

 

[Exophase]

Yes, that's what I said. New CPUs are more efficient and provide more power for the same heat, but they'll all heat up when they are driven to the max.

Okay but what are you getting at? Pretty much every CPU today is thermally constrained, and that probably includes the one in your desktop or laptop. The most favorable way to put it is that they let you run beyond what they can sustain because it's better to have that added power for burst loads. The less favorable way to put it is that it helps with marketing and benchmarks.

But "can it run all cores at X MHz indefinitely without getting too hot in this form factor" is a pretty meaningless statement.. I don't care about whether it can run at the maximum burst setting indefinitely, I care about what the actual sustained performance is. And your statement, that Cortex-A15 level stuff can't run at full speed, strikes me as misleading. If you take "full speed" to mean its rated maximum with all cores enabled that's true for a lot of things today that are weaker than A15 like A53 SoCs. But if "the A15" means the two that are in OMAP5 then that's not really true.

 

[Blue Shadow]

Okay but what are you getting at? Pretty much every CPU today is thermally constrained, and that probably includes the one in your desktop or laptop.

I guess his point simply is that many users are not aware of that fact and think that if a cpu gets too hot that would be a hardware issue that shouldn't solved in software. At least that's how I understand his postings.

 

[fusion_power]

The metal case won't solve the heating issue, it will just delay the time it takes to heat up fully.
So instead of 10 Minutes running full speed, you get maybe 30 Minutes.

Isn't that already what's all about? Everything small improvement helps there and nobody really wants to run a mobild evice for hours under full load. At least I hope so. It should be fine for moderate emulation and browsing and watching a Video. Smartphones are still mainly communication and social-app devices, not hardcore-gaming-devices. I guess that's one reason why real gaming Handhelds have compared weaker Chips (besides their age) or don't exist at all anymore. (PS Vita, 3DS, etc...) And look how great games can look on such devices, without needing active cooling and with compared weak hardware. It's all about the Software at the end.
It could help to make the (Pyra) programs more efficient, for example to keep the Chips cooler. Stuff like limiting FPS, reducing resolution and details for demanding 3D games etc. can help alot. Especialy when it comes to "straight" ports from the PC. I limit my PC games often to 60FPS even when my GeForce GPU can do much more. Results in much cooler temps and a quieter PC for sure.
There could be some "avoiding overheat" guidelines for demanding Pyra Software etc. that help to keep the device healthy.

At the end it may be unavoidable that the general philosophy changes here. Not just making/porting demanding Software without any limits, but more looking how this software can run on the Device for a long time without producing to much heat. It brings nothing to have a beautiful 1:1 PC port of an graphical demanding game that may look and run great (on the Pyra) but after 15 Minutes the device shuts down during overheat.

Yes, that's what I said. New CPUs are more efficient and provide more power for the same heat, but they'll all heat up when they are driven to the max.

And that's the reason to generaly avoid driving them to the max - at all costs if needed. I'm sure OMAP5 can still do fine without going to it's limits.

 

[Klumpen]

Isn't that already what's all about? Everything small improvement helps there and nobody really wants to run a mobild evice for hours under full load. At least I hope so.

There was a plan to sell a dock to easily transform the Pyra into a normal desktop PC, so I guess a cooling system built into the dock is mandatory now.

 

[EvilDragon]

I guess his point simply is that many users are not aware of that fact and think that if a cpu gets too hot that would be a hardware issue that shouldn't solved in software. At least that's how I understand his postings.

Exactly. Many users will see impressive videos of games / emulators (like on some tablets) and expect to be able to play these for hours.
Videos only show a few minutes though...

And most normal users expect a system which overheats to be broken.

I have seen so many users complaining that some mobile phones are using false advertising, as they throttle the CPU speed after a few minutes full power usage...

 

[_jr_]

A system that overheats when used *is* broken. That there was no SoC available that better fits the thermal budget of a Pandora like device is sad, but that's the way it is. I'm happy as long as the Pyra will not use more power than the Pandora for similar tasks, especially when idling (which is what my Pandora is doing most of the time, e.g. while editing source code, reading emails or using Anki).

 

[EvilDragon]

Then every mobile device without fan is broken in your opinion?

 

[Exophase]

Then every mobile device without fan is broken in your opinion?

It should really limit clocks before it gets uncomfortably hot, not after.

But it can be challenging to instrument this, and "uncomfortable" is sort of subjective.

 

[_jr_]

Then every mobile device without fan is broken in your opinion?

Unless it is like the Pandora which is pretty difficult to overheat (even charging while running the stress test didn't do it for me) or it is specifically designed in a way that the user cannot trigger overheating behavior even if the system would overheat under permanent load conditions (e.g. a dumb phone including the 'smarter' ones with some productivity apps). I don't agree with the 'reserve processing power for short demand peaks' marketing excuse, especially not in the case of the Pandora or Pyra that are explicitly marketed as mobile computers, where e.g. make -j <number larger than 1> while downloading large files or running an emulator for hours is fully intended and advertised usage. Sustained performance is the only relevant metric for the Pyra and if it isn't the full speed the SoC is capable of then marketing with the top speed simply is false advertising (which wouldn't prevent me from buying one, but it would annoy me just like talking about a 4200mAh battery on https://www.openpandora.org/index.html and delivering a 4000mAh one - according to the information printed on the battery itself - did annoy me).

 

[EvilDragon]

It should really limit clocks before it gets uncomfortably hot, not after.

But it can be challenging to instrument this, and "uncomfortable" is sort of subjective.

We were talking about not limiting the CPU

 

[EvilDragon]

I don't agree with the 'reserve processing power for short demand peaks' marketing excuse, especially not in the case of the Pandora or Pyra that are explicitly marketed as mobile computers, where e.g. make -j <number larger than 1> while downloading large files or running an emulator for hours is fully intended and advertised usage..

So, and how do you think we should fix that?
Use the OMAP3 for the Pandora or produce our own SoC?

Because there is NO SoC available otherwise that would work like that.

Besides, there is no problem with the peak usage, there is only a problem with users understanding it properly.

Every device has a limit of what it can run for an extended period of time without heating up.
You could either limit the SoC to that speed (so that it will never heat up) or you can allow the SoC to be run at full speed, but throttling it in case it gets too hot.

The second solution has no drawbacks compared to the first solution, it just has the improvement that it allows more power for short peaks, which helps with games that sometimes need an additional boost of power but only for a few seconds.

You should NEVER believe that when some advertisement of a mobile device tells you the CPU can run at 5GHz that it will do so for an extended period of time.
But compared to a different device that can run at max of 2GHz, it would still be better (in case the devices have the same limit for running extended periods of time without heating up).

Besides, telling an exact speed which a device can run with without heating up is not easily possible, it also heavily depends on the code.
Some functions need more CPU power than others, even if the run at the same clock speed.

It could be that one emulator can run a CPU with 1,2GHz without heating up while a different one heats up the CPU even when you limit it to 1GHz.

So, what would your solution be?
Sell devices telling they can run between 800MHz and 2GHz, depending on what you do?

 

[fusion_power]

There was a plan to sell a dock to easily transform the Pyra into a normal desktop PC, so I guess a cooling system built into the dock is mandatory now.

One possibility. I've also seen a solution with Thunderbolt 3 and ultra-mobile Laptops. You can plug in an external graphic card "box" (still rare but they come) to have have the full GPU power and still a cool Notebook.

I have seen so many users complaining that some mobile phones are using false advertising, as they throttle the CPU speed after a few minutes full power usage...

"Samsung VR" for example, 15-20 Minutes VR experience is enough until the phone throttles.

 

[ZXDunny]

"Samsung VR" for example, 15-20 Minutes VR experience is enough until the phone throttles.

Really? I've just had a four-hour session on Herobound and it's just warm - no warning about the temperature and the only reason I stopped was the phone battery dying.

 

[Chip]

It should really limit clocks before it gets uncomfortably hot, not after.

Every modern chip will throttle before it damages itself. All the chip cares about is its own internal temperature. A poor case layout with no path for the heat to escape will heat up internally and throttle long before you feel any temperature change on the outside. A device which shunts heat directly to the case could feel very hot, while the chip itself is well within its safe temp range.

If a device is getting hot on the outside, that's actually kind of a good thing (from the hardware's perspective). That means that a lot of the heat being generated is being transferred to the case and dissipating out into the air. It might make it uncomfortable to hold, but just as with a lot of bad stuff, heat is "better out than in". I've mentioned before that handheld devices like this with powerful processors would benefit greatly from using the case itself as a heatsink. Sure, you might get 1st degree burns from extended gaming sessions, but you'd never have to worry about thermal throttling

I'm really curious how a mainboard-behind-the-display setup with the entire top cover being a heatsink in direct contact with the SoC would function. ED claims the heat would damage the LCD panel, but I'm not so sure. Smartphones situate their chip on a board directly behind the panel, and you don't often hear of heat-related damage. Hell, my nexus 4 would get almost too hot to hold on occasion (charging wirelessly on a hot dashboard while serving as GPS and streaming music). Yet the display never suffered. I have to think that a 2W SoC backed by a ~13 square-inch slab of aluminum would stay plenty cool enough not to cook the panel. And it would certainly avoid throttling.

 

[vcoleiro1]

Another update from GPD today on Indiegogo. Pretty much just what the have finalised. You see the three stage switch for xinput, mouse mode and dinput


Dear backers.



This is the interface and L/R buttons rendering of final mould. Please see the pic blow:

And the final printing rendering of keyboard,gamepad and function mode. Please check the pic blow:

 

[ekianjo]

Ouch ,these shoulder buttons are going to hurt bad.