Moving along

[asimov-solensan]

Grat news!

Although only 10 minutes at full speed is not that great. And I don't expect the heatsink to make a huge difference. I feel pretty curious about stable speeds for a single core vs both running.

 

[elvissteinjr]

I'm now split on whether I want to get a Pyra before ED runs out of those clicky slider sticks (did he ever get a new supplier for those to continue sourcing them? I feel like I've asked this multiple times and forgotten the answer every time...)

iirc it's not that the other nubs are not gonna click, it's just that the manufacturer discontinued the ones with longer travel distance.

 

[Grench]

Grat news!

Although only 10 minutes at full speed is not that great. And I don't expect the heatsink to make a huge difference. I feel pretty curious about stable speeds for a single core vs both running.

For an SoC without a heatsink, that isn't bad at all. Modern SoC based systems aren't really designed to run at full speed on all cores all the time. Rather they are designed to peak for momentary loads then throttle back and keep running for longer jobs.

So, the part in this news that bothers me isn't that it only ran for 10 minutes at full speed, but that at 8 minutes it didn't drop to 1/2 speed (or whatever) and keep going. I wonder if those protocols/protections were deliberately left off or turned off?

 

[CommanderB]

For an SoC without a heatsink, that isn't bad at all. Modern SoC based systems aren't really designed to run at full speed on all cores all the time. Rather they are designed to peak for momentary loads then throttle back and keep running for longer jobs.

So, the part in this news that bothers me isn't that it only ran for 10 minutes at full speed, but that at 8 minutes it didn't drop to 1/2 speed (or whatever) and keep going. I wonder if those protocols/protections were deliberately left off or turned off?

I think we should wait for the tests inside of the actual Pyra prototype.

AFAIK the port/implementation of a more advanced throttling mechanism should be no problem as modern Linux already provides interfaces & mechanisms for this by default.

Gesendet von meinem SM-N9005 mit Tapatalk

 

[levi]

I'm now split on whether I want to get a Pyra before ED runs out of those clicky slider sticks (did he ever get a new supplier for those to continue sourcing them? I feel like I've asked this multiple times and forgotten the answer every time...)

IIRC he's got stock for thousands of Pyra. If by some fluke the Pyra becomes a big hit, or there are multiple revisions of the mainboard that everyone actually wants (although in that case he may well be able to harvest and reuse nubs from old boards I guess) then he might need to source more, but it's better to wait and see if that happens rather than spending any more money, especially just now.

 

[EvilDragon]

It's not just 10 minutes at full speed... Memtester both stresses the RAM and CPU to max.
This is unlikely to happen in normal usage, and that's without a heatsink.

Sent from my Moto Z using Tapatalk

 

[ashcloud]

So the Pyra is supposed to crash when compiling stuff on it for 10 minutes? I sure hope that's getting fixed yet.
Also why would the heatsink matter? Does the RAM overheat? The CPU should throttle down before it gets too hot.

 

[EvilDragon]

The Pyra is supposed to throttle down when it gets too hot, like any other mobile device out in the world which uses a powerful CPU.
Sometimes I do wonder what weird ideas arise

And of course memory can overheat. It actually heats up quite a lot when you stress it. The more memory you have, the hotter it gets.

That's why you don't have that effect with the 2GB RAM version.

Sent from my Moto Z using Tapatalk

 

[PowerGod]

The Pyra is supposed to throttle down when it gets too hot, like any other mobile device out in the world which uses a powerful CPU.
Sometimes I do wonder what weird ideas arise

It wasn't actually too weird as a question, if someone is used to x86 devices it is normal to see them suddenly shutting off when too hot... I had 2 portable PCs and had the same issue with both... so it's always important to explain the differences in architecture

 

[matzesu]

So when I play some PlayStation 1 Games on my 4gb Pyra at about 30 degrees Celsius outside, how long can I play??
I don’t know how to compile things, so it’s for me more a Gaming and a bit Office Device, and my now PC..

 

[levi]

And of course memory can overheat. It actually heats up quite a lot when you stress it. The more memory you have, the hotter it gets.

That's why you don't have that effect with the 2GB RAM version.

To try to clarify, this will definitely happen with memtester, because it tests all the RAM and with more RAM gates on each chip, the density is higher. If you were to do something that only used 2GB of RAM (which would probably be split across both 2GB RAM chips under Linux I think) then there will probably be much less difference between the units.

Regarding PS1 emulation, the PS1 only had 3MB (that's Mega, a thousandth - more or less - of a Giga) of RAM. Now, emulating one will use more RAM that that to keep track of bits of state and things, but it still should use well under 1GB I'd have thought. Now, it will also stress the GPU which I assume memtester doesn't actually touch, but it's unlikely to stress the CPU that hard, so it should last much longer than 10 minutes before overheating. And once the system it configured properly, it'll throttle before it overheats anyway, so your emulation might get a bit slower after perhaps half an hour (to make an educated guess) but the Pyra can probably emulate PS1 in it's sleep practically, so you may not actually notice a change. If you were to notice a change, the game would probably get a little laggier and stutter a bit and you might need to back of graphical enhancements a little. It's unlikely to stop you playing completely.

And you lot with x86 systems that power off if they overheat have weird systems. My Pentium-D build server throttles back to half max GHz once the core temperature reaches 60 degrees or so, running memtester on 75% of my RAM (leaving a little free because I'm also running linux to monitor core temperature and CPU clock). I don't think my old Pentium 3 server ever throttled, but I'd imagine Pentium 4 chips must have throttled because very few people had adequate coolers when those beasts launched.

 

[asimov-solensan]

I guess the system just freeze instead of throttling because such feature isn't implemented yet. That's why ED asked for developers among other things.

Anyhow I think we all here are hesitant to see a full prototype running and see what can we consider as max speed.

 

[levi]

Yes, I forget the exact name, but you need a little script or something to check the temperature and set the CPU frequency if it gets to close to a defined frequency. I assume they're running the ARM equivalent of memtest86+ which runs from the grub menu, so linux isn't even loaded yet. I'd kind of expect debian to ship set up to throttle already, but we'd probably need to tweak that because it may well be configured to throttle well before it actually needs to.

IIRC the Pandora used the on-demand...uuh...I want to say scheduler. That just clocks the CPU down when you're not using it, as as soon as it notices you're trying to stress the slow cpu, it ramps up the clock frequency and doesn't even look at the system temperatures at all. For the Pyra I guess we want something like that that can also monitor core and RAM temperatures and back off if it's cooking.

 

[EvilDragon]

The OMAP does throttle, but not very useful (once it reaches a certain temperature it goes down to the lowest speed until it has cooled down again.

This doesn't make much sense. It should dynamically throttle to keep the heat below that temperature, but running as fast as possible.

The memory is what causes the freeze right now, as it has no throttling and produces errors.
That will be fixed with the heatsink and also, it's a nonrealistic scenario, as you never stress the memory that much.

Sent from my Moto Z using Tapatalk

 

[fusion_power]

The Pyra is supposed to throttle down when it gets too hot, like any other mobile device out in the world which uses a powerful CPU.
Sometimes I do wonder what weird ideas arise

Not if the cooling solution is big enough. Of course most devices spare that part out to make it all smaller and thinner. Good thing is, the Pyra is not small or thin at all. I've heared good things about these new vapour chamber coolers, pretty thin and efficient, for instance.

 

[EvilDragon]

They're not thin. We got 0.4mm between the CPU board and the case. Those are three times as thick

Sent from my Moto Z using Tapatalk

 

[fusion_power]

They're not thin. We got 0.4mm between the CPU board and the case. Those are three times as thick

I know. It was just an example. And you know you can transport heat from one spot to another so the heatsink could be at a different place than the heat source.

 

[Swordfish II]

Hey yeah, we are three year after the preorder lets redesign the heat sink for another three years.

Refresh my memory (ha pun there). How did things work changing the original tested and proven, 2gb setup? How much headache did that add?

I have said it before and will keep saying it. Stop the scope creep and get the product out. Competition keeps popping up and beating the Pyra to market (gpd win, gpd pocket, gpd win 2, Gemini PDA)

 

[ImmaculateChimp]

... Regarding PS1 emulation, the PS1 only had 3MB (that's Mega, a thousandth - more or less - of a Giga) of RAM. Now, emulating one will use more RAM that that to keep track of bits of state and things, but it still should use well under 1GB I'd have thought. Now, it will also stress the GPU which I assume memtester doesn't actually touch, but it's unlikely to stress the CPU that hard, so it should last much longer than 10 minutes before overheating. And once the system it configured properly, it'll throttle before it overheats anyway, so your emulation might get a bit slower after perhaps half an hour (to make an educated guess) but the Pyra can probably emulate PS1 in it's sleep practically, so you may not actually notice a change. If you were to notice a change, the game would probably get a little laggier and stutter a bit and you might need to back of graphical enhancements a little. It's unlikely to stop you playing completely.

..I'm only saying this out of deep concern (and probably more than a little from a lack of technical knowledge/understanding..), not a desire to sabotage, but I find this to sound more than a little bit worrisome... I would've hoped that the Pyra would be more than capable of handling something like PS1 emulation without thermal throttling & lagging & stuttering & whatever under any conditions, short of attempting to marathon the entirety of something like FF9 or Metal Gear Solid from within the wastes of one of the world's hottest deserts..during daytime..

When you mention these "performance issues" seemingly in relation to graphical enhancements, are you saying we should only be concerned about running into any of those various problems if we are attempting to force a PS1 emulator on the Pyra to turn these old PS1 games into some sort of *ULTRA HIGH-DEFINITION* showpiece with *eNhAnCeD eFfEcTs*, or should we expect to have these sort of difficulties even when merely trying to have the games look a little closer to their intended look from the original console (or even, possibly, no "alterations" at all)?

Which brings up another question I was curious about: the possibility of adding "CRT-style" filters to various emulators like SNES and PS1 so they can look "more authentic" when played on some type of screen that isn't CRT...and I suppose the level of performance impact one should expect from availing use of one of them..

Also, sadly, I'd think if the Pyra won't be able to run PS1 games without "thermal issues", we should forget about even trying to run Dreamcast stuff..

 

[Silent-Hunter]

The Pandora runs PS1 fullspeed without thottling or overheating. The Pyra prototype already ran Dreamcast emulation at full speed I believe. It'll be fine!