Micro Pyra news from the.. dragons mouth

[Wally]

Hello,

Just thought i'd share some news just to keep our brains from exploding due to not having any fun news:

1. Next batch of Emails for preorderers will be sent this week, rest of orders have been finished (Bar 1 or 2)

2. Samples were finished last week for the new copper coloured case but havae not been collected yet.

3. Nikolaus is testing the RK3568 CPU on an Odroid Board and appears to be 20% faster than the OMAP5432 but that's only testing the CPU), This CPU is cooler than our existing CPU yet with performance gains and producing much less heat. This is basically the beginning of the new CPU board but there's still a long way to go.

Hopefully that keeps you all from exploding

 

[iggunr]

new color case? nice

 

[Farox]

Thanks @Wally very interesting news.

 

[hns]

3. Nikolaus is testing the RK3568 CPU on an Odroid Board and appears to be 20% faster than the OMAP5432 but that's only testing the CPU), This CPU is cooler than our existing CPU yet with performance gains and producing much less heat. This is basically the beginning of the new CPU board but there's still a long way to go.

Well, it is only clear from a 20 minutes test that it runs 20% faster for some multi-threaded prime number calculations (well, on 4 cores at 2 GHz instead of 2 cores at 1.5GHz so in fact each single core is much slower).
From the data collected so far I would not yet draw any conlusion which one is cooler in a handheld device.
The Odroid M1 (with big heat sink removed but at open air where the above test was made) goes up to 78°C inside the chip as reported by the Linux temperature monitors and >50°C outside (measured with a thermal camera) within 10 minutes.
For the OMAP5 a comparable experiment has still to be done (the Pyra itself is not a valid comparison as it runs in a closed case and the OMAP5 stacked inside the sandwich of printed circuit boards etc.).
It is said that the RG503 stays much cooler than the Pyra, but I don't know what they are doing in software (e.g. downclocking or better power management or just run one core) or other tricks. And it has only 1GB memory and memory chips also produce heat.
So drawing a conclusion that it is cooler at the same performance level is far too early in my view.
And yes it is a long way to go.

 

[Wally]

Thanks for the clarification, was just quoting EvilDragon here

 

[fantomid]

Well, it is only clear from a 20 minutes test that it runs 20% faster for some multi-threaded prime number calculations (well, on 4 cores at 2 GHz instead of 2 cores at 1.5GHz so in fact each single core is much slower).
From the data collected so far I would not yet draw any conlusion which one is cooler in a handheld device.
The Odroid M1 (with big heat sink removed but at open air where the above test was made) goes up to 78°C inside the chip as reported by the Linux temperature monitors and >50°C outside (measured with a thermal camera) within 10 minutes.
For the OMAP5 a comparable experiment has still to be done (the Pyra itself is not a valid comparison as it runs in a closed case and the OMAP5 stacked inside the sandwich of printed circuit boards etc.).
It is said that the RG503 stays much cooler than the Pyra, but I don't know what they are doing in software (e.g. downclocking or better power management or just run one core) or other tricks. And it has only 1GB memory and memory chips also produce heat.
So drawing a conclusion that it is cooler at the same performance level is far too early in my view.
And yes it is a long way to go.

Thanks for details.

Could we consider that the RK3568 will be the successor of the OMAP5 ?

 

[matzesu]

And how to change the CPU Board whitouth screwing up other incriedience of the Pyra?

 

[TrashyMG]

And how to change the CPU Board whitouth screwing up other incriedience of the Pyra?

I'd first worry about getting a CPU board designed and built and out to customers before worrying about replacing your CPU board with a currently non-existing CPU board.

 

[matzesu]

Thanks a lot for bringing the updates over from Discord to the Boards, (and for your Porting work at the start of the Pyra deliefery)
Even if its on the actuall Pyra Mainboard gets a bit hotter than on the Developer Board and its not that many times faster, its still lot better than the Omap5, allthough i found the Omap5 allready quite usefull as a beginners CPU..

 

[comradekingu]

Ultimate CPU-speed isn't holding back anything sensible at this resolution (even if it is lower res and CPU matters more)?
Better GPU is needed to make it a better portable plug-in console? Memory and storage are the never enough things to have.
Synthetic benchmarks with full core-utilization seems only indicative of not overshooting what amount of heat can be dissipated.

Longevity (not the new-and-shiny business it isn't really in) is usually achieved by having something that is easy to maintain, at a sensible power-envelope. More likely in a field that isn't automotive and industry, where nobody wants to touch what works.

Newer features, ala 64-bit addressing and new OpenGL ES to stay compatible with various software and releases is the kicker.
Moreover, the same speed at lower heat anywhere means better battery-life.
Whatever the baseline with no load amounts to IMO matters more, as having the device powered on and idle varies greatly if any little thing uses more than it has to.

Like for like, using something more devices and projects are going with helps as a joint effort, and avoids some possible semi-unique problems. That many of these are Linux-based projects can be a lot more valuable than whatever specs something has.

 

[λ the β-Redex Reducer]

Thanks a lot for bringing the updates over from Discord to the Boards, (and for your Porting work at the start of the Pyra deliefery)
Even if its on the actuall Pyra Mainboard gets a bit hotter than on the Developer Board and its not that many times faster, its still lot better than the Omap5, allthough i found the Omap5 allready quite usefull as a beginners CPU..

N900 has OMAP5 and runs fine with it. Sometimes the problem is not slow hardware but a game just not being retro enough.

 

[TrashyMG]

N900 has OMAP5 and runs fine with it. Sometimes the problem is not slow hardware but a game just not being retro enough.

N900 has an OMAP3, like the Pandora.

 

[Zwerg01]

Well, it is only clear from a 20 minutes test that it runs 20% faster for some multi-threaded prime number calculations (well, on 4 cores at 2 GHz instead of 2 cores at 1.5GHz so in fact each single core is much slower).
From the data collected so far I would not yet draw any conlusion which one is cooler in a handheld device.
The Odroid M1 (with big heat sink removed but at open air where the above test was made) goes up to 78°C inside the chip as reported by the Linux temperature monitors and >50°C outside (measured with a thermal camera) within 10 minutes.
For the OMAP5 a comparable experiment has still to be done (the Pyra itself is not a valid comparison as it runs in a closed case and the OMAP5 stacked inside the sandwich of printed circuit boards etc.).
It is said that the RG503 stays much cooler than the Pyra, but I don't know what they are doing in software (e.g. downclocking or better power management or just run one core) or other tricks. And it has only 1GB memory and memory chips also produce heat.
So drawing a conclusion that it is cooler at the same performance level is far too early in my view.
And yes it is a long way to go.

Isn't the OMAP5 32-Bit and the RK3566 64-Bit?

 

[matzesu]

Yep, Omap5 is 32 and RK6566 is 64 Bit, but other than that, my knowledge is limited..

 

[EvilDragon]

Oh, heh, I did expect Wally to post a bit longer news... I think what I sent him on Discord was more text than he posted here and I thought he'd grab the information and writes them up a bit more detailled (similar to what I did back with the Pandora when Craig fed me with news )) But heck, it's better than me not posting anything at all

Luckily, I have a bit of time right now, so here's some more information.

1. I didn't manage to finish assembling the still-left-to-ship Pyras last week, but most are out there. That means that the next batch of emails will go out this week, didn't manage it last week, sorry about that.
2. I've picked up the case samples. The copper coating is made by the same company that produces the chrome coating, so it looks gorgeous and is very scratch resistant as well. The black one also has been changed. It is VERY matte (but also resistant to scratches) and has a nice feel. It similar to soft touch - but without rubber and therefore doesn't get sticky over time. It's also used in Porsche cars
3. Ookay, about the OMAP5, RK3568, heating and so on. Yes, hns only made small tests, HOWEVER, I can give some real life experience. The Pyra gets pretty warm in your hand after around 10 minutes even if you're just playing simple games that even the Pandora could run without any issues. The RK3566 (which is almosts the same CPU, just with 200MHz less than the RK3568) is used in the RG503 (as mentioned above). It doesn't have ANY kind of cooling on the CPU. Not even some copper or similar. And it's sitting directly above the battery. That said, I can easily play Dreamcast emulations without it getting very hot.
   So from this real life experience, the RK3568 stays cool even with more taxing tasks - compared to the OMAP5 which seems up to heat up even with simple tasks.
   BUT: And this is one big thing. The Pyra stays cooler compiling stuff than playing games! So one guess is, that the missing audio hardware driver leads to a very high CPU usage whenever audio is played - so even with simple games, the CPU is taxed. Converting every audio to 96kHz audio is not a simple task for a CPU. Another sign that points that way is the fact that we often have audio buffer underruns (which usually happen if there's not enough CPU available for playing audio). hns and some others are currently heavily working on getting the audio ABE working - so this could also fix the heat-up of the Pyra. But still, moving to a 64bit CPU in the future that's probably better supported by the community and which runs 20% faster while staying cooler would be nice.
   What can be said for certain is that it would need less battery, simply because it uses LPDDR instead of DDR and because the manufacturing process is more recent.
   So we'll see where that leads

I hope that's a better quick summary and explanation. I'll go back assembling the remaining Pyras now

 

[rSl]

thanks for the in-depth update and woohoo for blackbody porsche pyras (in the hour of chaos).
looks like there is a real chance of getting even cooler pyras while we are still kind of alife.

 

[PowerGod]

Shame to @Wally for not reporting everything !!

 

[oskda]

I am not excited about performance increment with future RK3568 SOC because in single-thread it will offer less performance than actual OMAP5 (at least if no audio playing being involved), but it will be a great update:

-More support and drivers
-Less heat and more battery life
-64 bits ARM ISA
-More cores (good for some cases)
-No Spectre bugs: there ia no out of order execution (there was OoO on OMAP5 Cortex-A15)

And for me a some very big plusses: this SOC or it's variant, RK3566, are used on other Linux computers I want to own and use:

-PineNote: it is a tablet-conputer running Linux with a big e-ink display (more than 10 inches): https://www.pine64.org/pinenote/

-Pocket MNT Reform will have 4 options for CPU board (it is in a daughter board): https://liliputing.com/mnt-pocket-reform-is-7-inch-mini-laptop-with-a-modular-open-hardware-design/
and one of them contains RK3566

It is nice to have the same SOC on them.

 

[Git1]

Awesome news Ed, thanks. Great to hear you're catching up on stuff.

 

[EvilDragon]

Shame to @Wally for not reporting everything !!

View attachment 38512

Naaah, we don't want to blame Wally here
I'm happy he did post some news, it was still better then me posting nothing and he also doesn't have too much time on his hands as well

So... shame only on me