I'm out of ideas for news-titles


If it means you can start shipping by the end of the month vs. having to wait another month or something, then I'd certainly be fine with a 2GB (I am hoping you will be selling CPU board upgrades in a year or two …)
 
For mass production, I have already various solutions in mind...

1. If it turns out the CPU board layout needs to be fixed to get the 4GB running, we would have the problem that we already have 500 CPU boards lying at GC. Throwing them away would be a waste of 8000 EUR, so I would ask all everyone who preordered if he would be fine with a 2GB unit. If we get enough who would be fine with that, we could order the 2GB RAM chips and populate right away.

2. If the new RAM Chips work (but the old won't), we'd need to check the price difference and whether we could sell or give back the already delivered 4GB RAM chips. This could cause quite a loss (10k - 20k EUR), but thanks to my shop running well, we would be able to afford that (it would still hurt a lot).

It could also be a mixture of both.

But let's first try to get the RAM to work before we're doing more speculations!
After all, it runs stable on ONE existing CPU board, so whatever is keeping it to run stable on all of them can't be much.
Maybe it's really just the higher amount of noise coming from the older CPU Board revision...

We keep working on that - and you can be sure I'll shout it out loud once the RAM is working!

This is currently my main concern - as everything else is working already.

Honestly I would be Ok with #1, since I don't have hundreds of tabs open at once, autocad running, and I tend to use LXDE or XFCE, etc.

What happened with the wifi antenna? I remember someone had mentioned the antenna needed to be swapped due to issues. I also remember Golden Deli mentioned some issue with the (eagle) layout?
 
Last edited:
Wifi ?
Very important for me as I plan to use the Pyra on travels where hotel/camping wifi reception might be quite bad.
That's not an issue, just something that needs to be measured properly. With two diversity antennas, you shouldn't have any issues.
I wonder how the SW side of things are going and when that is likely to be ready .
Oh, coming along nicely. Battery readout is working, touchscreen calibration, keyboard incl. Shortcuts and the special keys are working...
Like on the Pandora, it will mature over the years, but it's working nicely already.
If it means you can start shipping by the end of the month vs. having to wait another month or something, then I'd certainly be fine with a 2GB (I am hoping you will be selling CPU board upgrades in a year or two …)
Well, once I give the greenlight, it'll take about 6 weeks and then production can start.

Sent from my XT1650-03 using Tapatalk
 
I didn't have a single free day for five weeks now,

Please rest.

1. The cases

FormAction is currently in the process of updating the 3D data to represent the current mould.
Which means I can check everything upfront, without having to wait for samples.

Also, the special material for mobile devices is on the way and should arrive end of this week (let's hope they're right).

Good news you can make some checks upfront, but you still need to test samples, specially if the material is different
from before. Some stress test or mechanical checking would be in order.

But you might remember that I considered a second company as another option, one with which I have already worked together and produced SNES, MegaDrive, NES and N64-shells:
Yes, we are keeping this open as an option. They've got a very interesting material as well: A mixture of plastic and carbon, which increases the durability of the plastic a lot (on their own tests, the normal plastic broke with 400kg force whereas the plastic/carbon-mix hold up to 1,5t!)

And someone from the community has contacted me: He's working for a plastic production company in Germany. I'll send them a Pyra case as well and will ask for a quotation.

I have a feeling changing company now would delay it further and introduce new problems that may slip through if checked in a hurry.
But I understand the scheduling problems with the company you're using now. I guess asking for some quotes can't hurt.

This mainly depends on the material arriving and the case quality of the next (and hopefully final) sample we'll receive from FormAction.

There's hope this works at last.

2. The cooling

The cooling consists of an off-the-shelf heatsink hidden below the USB port and some special 0.2mm thin heatpipe-material from 3M to transport the heat from the CPU to the heatsink.
The only thing we can improve here is to use two heatsinks (another one hidden below the second USB port) to keep it cool for an even longer period of time.

Good that the colling works. I have some questions. Excuse my ignorance.
  • Does the heatsink near the USB port have some impact on any attached USB devices ? Some USB devices get hot on their own.
  • is it possible to plug some external (removable) heatsink in the USB port (maybe for use in a table, not handheld, and far enogh that other ports are not blocked). Or would that put too much heat through the USB port and so stress it and eventually lead to contacts unsoldering or any other problem?
  • At some point I seem to recall the SD cages were also going to help cooling. Or was it that they needed cooling ?
When running the unit at full load with 1.5GHz, it heats up A LOT more.
The CPU reaches over 90°C within a few seconds - and then immediately throttles.

I don't know if this is possible, but have you tested some way with some bigger heatsink or pipe or some other cooling that is much more efficient than
what fits in the case, just to suggest whether there's a limit in the heat spreader (the sheet of heatpipe-material getting heat from CPU to heatsink) or the heatsink itself?

I don't know how feasible this coul be, and I don't think it has priority, but can the Pyra work plugged without a battery (and without battery cover)?
Could some cooling device be designed, for putting in the battery compartment when the Pyra is plugged to mains, so that
it improves heat dissipation and the CPU can work at 1,5GHz for longer when on a table near a plug?

It could just take heat out to external sinks or it could be powered and use fans or whatever. (Or have liquid pipes to heat a keetle or whatever :eek:p).
It could be combined with some form of docking station if convenient, although I tend to prefer one device for one function.

I think there are docking stations for laptops with fans, so it would be more or less the same but used without the battery to get
closer to the heat source. It could even daisy chain and allow the battery to be plugged externally and forward the contacts to the pyra,
if people want a bulkier desktop pyra combo unplugged.

It might need case modification to get some better thermal connection to the heat spreader or something... (holes in the plastic between battery and
cpu board for metal contacts in the cooling device to go through) The difficulty is you don't
want heat to go to the battery compartment when used handheld with battery (as designed) so any thermal connection should be
removable or replaceable with thermal insulation. Sounds complex.

The maximum speed you can run it with full CPU load for longer periods of time is therefore something between 1 and 1.5GHz.

And it depends on the weather...

And I'll work on a CPU Upgrade board as soon as we are finished here.

Let's finish this one first.

There are multiple ways: We could replace those with Tantalum capacitors - but we're not really fans of that (because of ethical reasons).

Glad to hear of ethical concerns. I wish I could help with that but I'm bad at electronics.
I don't think there are tantalum capacitors made from conflict free tantalum. I guess most providers just source
materials from different sources and don't give that kind of fair trade guarantee, but I don't really know.
So I'm hoping it can be fixed with non-tantalum capacitors.

Whether his patches work well with the CPU boards is something we're testing right now.

I don't remember how many 4GB prototypes are out there.

So with all the tests, the following scenarios are possible:

1. There's too much noise on the power traces or the power traces aren't wide enough which would cause a power loss.

Would that require a new PCB or is it somehow fixable by soldering a cable on top of it or something ?
I'm guessing power traces won't be accessible for soldering anything, maybe in a middle layer or whatever :(
I mean for testing whether this is the case, for production I guess you need a new PCB, since I can hardly imagine
pick and place machines soldering a cable between legs of different chips, specially "legless" BGA chips :)

2. The EMIF (Memory timing) settings aren't setup properly yet. We're still working on those (it's pretty complex!)

Good luck. That'd be best.

3. The memory chip itself is bad (which is unlikely, as it is being used in other devices as well).

And can't they just be poorly soldered? Maybe not a missed contact nor a short, just insufficient contact area sometimes ?
That might also explain why one board works and others don't ?

If the new RAM also doesn't work, the next step would be to populate some of the new optimized CPU boards with the 4GB RAM chips.

Can't you just test the new PCBs, a few with the new 4GB chips and others with the 4GB you already have ?

In fact if there have been changes, testing the old PCB is not as useful. We're interested in how well the new ones work.
Everything you can test with the already existing prototypes is fine, but if you're going to build new prototypes and have
new PCBs, isn't it worth to test if these new PCBs work with the RAM you already have in stock?

Maybe try in order:
1- new PCB with old RAM (if this works we're happy, but the test sounds more expensive than 2 so maybe exchange 1 and 2?).
2- old PCB with old RAM (and new EMIF settings).
2- new PCB with new RAM (good but you'll lose money)
3- old PCB with new RAM (you'll lose more money and whatever the new PCB fixes, which I'm not sure it was required after all, or nice to have).

1. If it turns out the CPU board layout needs to be fixed to get the 4GB running, we would have the problem that we already have 500 CPU boards lying at GC. Throwing them away would be a waste of 8000 EUR, so I would ask all everyone who preordered if he would be fine with a 2GB unit. If we get enough who would be fine with that, we could order the 2GB RAM chips and populate right away.

So the new 500 PCBs you have are the new optimized version (theoretically less noisy), they just doesn't have the hypothethical yet to be discovered hardware fix for 4GB RAM, right ?
Or you have 500 PCBs with the unoptimized (theoretically more noisy) previous design ?

I want 4GB of RAM, preferably with timings as fast as possible, I can wait for it. If eventually the only option is 2GB I guess I'd cancel.
Anyway I don't think I'm in the first 500 preorders, so it might not even matter.

2. If the new RAM Chips work (but the old won't), we'd need to check the price difference and whether we could sell or give back the already delivered 4GB RAM chips. This could cause quite a loss (10k - 20k EUR), but thanks to my shop running well, we would be able to afford that (it would still hurt a lot).

I think you are already overstressed, but I don't know if RAM prices can be tracked somehow (by someone more or less used to these markets?) so that
they can help ED know when is the best moment to sell minimising loses (I guess prices fluctuate? currency certainly does).
 
  • Like
Reactions: rSl
Wifi ?
Very important for me as I plan to use the Pyra on travels where hotel/camping wifi reception might be quite bad.

Successfully tested the Pyra's Wi-Fi at DebConf17. It worked, showed good speeds, but signal strength dropped off sharply with distance from the access point. It seems like, the transmit or receive SNR is worse than it should be.
 
One thing with tantalum caps is, that they're unreliable compared to ceramic capacitors.

By the way: there is not "the ceramic capacitor".
There are many different types of ceramic capacitor.
The cheaper ones have a tendency to whine.
There are sorts which stay silent.
 
Thanks ED for the news, it’s always a pleasure to read the in-depth story of the making of the Pyra.

I hope the 2GB first, 4GB later scenario works for everyone. I’d be waiting for the 4GB version though, but I’m pretty sure lots of people have use cases that don’t mandate 4GB and ordered this version because of the negligible additional cost.

Having lots of “free” RAM gives space for the Kernel to work faster, using cache and defragmenting the RAM less often for example (what Linux calls “free” RAM is full of cache, as it can be deallocated immediately should the applications need more memory). Cache is very important and modern filesystems make use of it to accelerate random writes and limit fragmentation, which is important on flash storage. People who want to use Eclipse or other development tools or any tools that stress the RAM and/or storage will always benefit more RAM, even if they don’t use all the RAM. Also swapping on flash hurts it and should be avoided.
 
Noooo !!! That's too short, I need circa 2 more months (tm) to end gta San Andreas, Half Life 2 and Doom 3 on the GPD Win..

But nothing Stopps me to play whit the Gpd even if I have my Pyra, so let's fix the issues and full steam ahead too Produktion
 
I hope the 2GB first, 4GB later scenario works for everyone. I’d be waiting for the 4GB version though, but I’m pretty sure lots of people have use cases that don’t mandate 4GB and ordered this version because of the negligible additional cost.

It works fine for me: I'll be happy to wait as long as it takes, and I don't mind other people getting their 2GB models before me.

I just hope that there aren't so many of us holding out for 4GB that it delays things for everybody.
 
Good news you can make some checks upfront, but you still need to test samples, specially if the material is different
from before. Some stress test or mechanical checking would be in order.

Yes, of course, don't worry about that!
But if I had the 3D data before, I wouldve known before the shoulder buttons won't work.

I have a feeling changing company now would delay it further and introduce new problems that may slip through if checked in a hurry.
But I understand the scheduling problems with the company you're using now. I guess asking for some quotes can't hurt.

Not really. The moulds can be transported there within a few days and they've proven to be a lot more reliable.
They also have 16 machines instead of only two, so the production would be a lot faster.

It's not an unknown company we're talking about - I already moved some moulds to them already.

Good that the colling works. I have some questions. Excuse my ignorance.
Does the heatsink near the USB port have some impact on any attached USB devices ? Some USB devices get hot on their own.

Not really, It's sitting below the USB port, not directly touching it.

is it possible to plug some external (removable) heatsink in the USB port (maybe for use in a table, not handheld, and far enogh that other ports are not blocked). Or would that put too much heat through the USB port and so stress it and eventually lead to contacts unsoldering or any other problem?

That would only work if the heatsink would be directly connected to the USB port.

At some point I seem to recall the SD cages were also going to help cooling. Or was it that they needed cooling ?

There's a huge aluminium plate between the PCB and battery compartment, and they're also connected to the SD Card slots.
So they help a bit of cooling, but not the CPU but rather help to spread the heat inside so that the battery won't run hot.

I don't know if this is possible, but have you tested some way with some bigger heatsink or pipe or some other cooling that is much more efficient than
what fits in the case, just to suggest whether there's a limit in the heat spreader (the sheet of heatpipe-material getting heat from CPU to heatsink) or the heatsink itself?

We can't use a thicker one, but we can add more width and also use two heatsinks.
Whether that would help (as the current one is already covering the CPU fully) is not known yet.

I don't know how feasible this coul be, and I don't think it has priority, but can the Pyra work plugged without a battery (and without battery cover)?
Could some cooling device be designed, for putting in the battery compartment when the Pyra is plugged to mains, so that
it improves heat dissipation and the CPU can work at 1,5GHz for longer when on a table near a plug?

The Pyra can be used without a battery, yes.
However, your idea is not feasible - as it would mean we would need to put heatpipes directly from the CPU to the battery compartment if you want to cool it there...
And when the battery is in there, it would directly heat up the battery - which is not a good idea :D


And it depends on the weather...

Not really - unless your weather would heat it up to more than 90°C :)

I don't remember how many 4GB prototypes are out there.

zmatt has two, Tony has one, Nikolaus has one.
I can build a few more though.

Would that require a new PCB or is it somehow fixable by soldering a cable on top of it or something ?
I'm guessing power traces won't be accessible for soldering anything, maybe in a middle layer or whatever :(
I mean for testing whether this is the case, for production I guess you need a new PCB, since I can hardly imagine
pick and place machines soldering a cable between legs of different chips, specially "legless" BGA chips :)

No, it can't be handsoldered. You'd need a new PCB layout.

And can't they just be poorly soldered? Maybe not a missed contact nor a short, just insufficient contact area sometimes ?
That might also explain why one board works and others don't ?

Unlikely, as then the memory would show other issues.

Can't you just test the new PCBs, a few with the new 4GB chips and others with the 4GB you already have ?

Yes, but populating a few of them would cost about 2000 - 3000 EUR.
So we'll try to work out everything else before.

Maybe try in order:
1- new PCB with old RAM (if this works we're happy, but the test sounds more expensive than 2 so maybe exchange 1 and 2?).
2- old PCB with old RAM (and new EMIF settings).
2- new PCB with new RAM (good but you'll lose money)
3- old PCB with new RAM (you'll lose more money and whatever the new PCB fixes, which I'm not sure it was required after all, or nice to have).

Well, our order is 2, 3, 1&2 at the same time.
Because 1&2 means a new population with the machine, and that triggers the above mentioned costs of 2000 - 3000 EUR.

So the new 500 PCBs you have are the new optimized version (theoretically less noisy), they just doesn't have the hypothethical yet to be discovered hardware fix for 4GB RAM, right ?
Or you have 500 PCBs with the unoptimized (theoretically more noisy) previous design ?

The 500 we have are the optimized ones. The one we're using right now are the unoptimized.

I think you are already overstressed, but I don't know if RAM prices can be tracked somehow (by someone more or less used to these markets?) so that
they can help ED know when is the best moment to sell minimising loses (I guess prices fluctuate? currency certainly does).

Naaaah. It's not like buying off-the-shelf RAMs. They don't fluctuate that much and usually have a leadtime of 6 - 8 weeks.
 
I'd take a capped to 2GB 4GB board if that would be an option. If the problem ends up to be on the software side, great! I'll update that and have 4GB board.
If it ends up to be a problem on the hardware side it would be nice to be able to trade in the old 4GB CPU board for an improved 4GB board, after which Ed can sell the old 4GB board as a 2GB board...
Something like that would have my preference. No boards thrown away, faster Pyra delivery, everybody 'appy
 
Noooo !!! That's too short, I need circa 2 more months (tm) to end gta San Andreas, Half Life 2 and Doom 3 on the GPD Win..
Don't worry, you will have plenty of time to finish those while waiting, I'm sure. If not, then you are free to move to the back of the queue :)
 
The 500 we have are the optimized ones. The one we're using right now are the unoptimized.
I thought the 500 run of the optimized boards was supposed to help fix the issues that are still being chased on the unoptimized version.

Wouldn't the next step still be to populate some of the 500 to see if it fixes any of the outstanding issues?

Yes - somewhat expensive, but at some point it becomes a logical step if the 4GB issues are going to be overcome.

With regards to the 'unoptimized' boards. You mentioned that there are several populated with 4GB being used to chase down the related issues. How many 2GB unoptimized boards are in developer hands? How are they working?
 
I chose the 4GB because I could. Plus I'm a bit worried about sharing memory with the video chip. I can see a game requiring 256 megs of VRAM and 1.5G of memory. This doesn't leave much for the system.
Anyway, I doubt such a game would make it to the first gen Pyra. And I think I'd gladly take 2GBs if it comes cheaper and earlier, and wait for an upgrade.

IMHO, futureproofing the current CPUPCB won't be by with maxing out its usable RAM, but rather its usable clock speed. After all...

xlhbq.jpg


Love that pic. I know this quote is out of context, I just find it funny.
 
Not really. The moulds can be transported there within a few days and they've proven to be a lot more reliable.
They also have 16 machines instead of only two, so the production would be a lot faster.

It's not an unknown company we're talking about - I already moved some moulds to them already.

Well, you know this business better than I, of course. I was fearing some informaiton lost somewhere,
some hidden incompatibility, they giving little priority to small orders (guessing yours is small, which I already
don't know because I don't know what's big or small in that maret) who knows?.

But if you already have the experience, I'd better shut up.

Not really, It's sitting below the USB port, not directly touching it.
Good to hear.

That would only work if the heatsink would be directly connected to the USB port.

Ok, but it sounds safer as it is. So all good.


There's a huge aluminium plate between the PCB and battery compartment, and they're also connected to the SD Card slots.
So they help a bit of cooling, but not the CPU but rather help to spread the heat inside so that the battery won't run hot.

Err... ok. I didn't remember this. I thought there was only this heatpipe thing touching the CPU and extending until near the USB.
So this video (near 7:45) is not up to date as to the mounting procedure, there are two more plates to add now. Correct ?

We can't use a thicker one, but we can add more width and also use two heatsinks.
Whether that would help (as the current one is already covering the CPU fully) is not known yet.

Now I don't know whether I wasn't clear, I was clear explaining a nonsensical idea and so equally hard to understand or both.
I was just thinking that if the current solution does not remove enough heat for the CPU to work more often at 1,5GHz maybe
one could try to diagnose whether it is because the heatsink is small or because the heatpipe can't move enough heat.
So one possible solution would be to measure with a bigger heatsink (a bigger heatpipe won't fit, I guess), even if you can't
close the case properly with the bigger heatsink. the goal is not to use that bigger heatsink, just to show that the heatpipe
can move enough heat to it, so there's some use in thinking about two small heatsinks or some other solution. If the bigger heatsink
does not make a difference then it is the heatpipe that can't move more heat and there's no point adding a second heatsink or
a better one or so because the bottleneck is the heatpipe.
Assume you have a bigger heatsink (just a PC CPU heatsink, for instance). You measure with the bigger heatsink (maybe
with the case open since it won't fit inside it. Then you measure with the case equally open but the current small heatsink.
If the two measures are similar, the current heatsink may be enough and you can leave it like now or at most try to improve the heatppe.
If the measure of the bigger heatsink is much better (lower heat near CPU) than the measure of the smaller heatsink then
you can think of adding a second heatsink, or replace the current one by something else, or case modification to vent near the
heatsink, or whatever.

But now you say you need a bigger heatpipe for 2 heatsinks so Iguess it is not worth it.

The Pyra can be used without a battery, yes.
However, your idea is not feasible - as it would mean we would need to put heatpipes directly from the CPU to the battery compartment if you want to cool it there...
And when the battery is in there, it would directly heat up the battery - which is not a good idea :D
That's the difficulty I said, the goal is the opposite with or without the battery.
You need something that brings heat to the battery compartment when you use the cooling device
and insulates it when using the battery.
I've looked at images again and I don't find how to do that.



zmatt has two, Tony has one, Nikolaus has one.
I can build a few more though.

Thanks. So 4 4GB prototypes.
And none works perfect with the default settings, one works fine with somewhat weird settings,
and one or three don't work with any known settings (I'm not sure there've been tests with all known settings in all prototypes).
Or I just didn't understand the situation?
This makes it difficult to tell whether there's some kind of yield issue that makes some trouble in 3 boards,
or the hardware design could be improved somehow and some of the chips Tony has just happen to be a little more tolerant
or all known settings aren't optimal and some unknown one would just make them all work...
Which is essentially what you've been saying from the start, yes.

Unlikely, as then the memory would show other issues.
Ah, ok.

Yes, but populating a few of them would cost about 2000 - 3000 EUR.
So we'll try to work out everything else before.

Makes sense. On the other hand if the issue is so subtle that the same design with the same population run produced one
working and 3 not working boards with the same EMIF settings, then maybe the optimization in the new boards might
change some subtle condition and make it work with more normal settings or even make the problem worse (unlikely).

I mean in some way there's some waste in the work on the unoptimized PCB prototypes since the new PCBs is what
will ship (if made to work with the amount of RAM 500 people accept). In theory the work on the old PCB prototypes is
useful because the new PCB doesn't change significantly and changes for good. But in theory 4GB should work in the old
PCBs, so there's missing data somewhere in this theory.

Well, our order is 2, 3, 1&2 at the same time.
Because 1&2 means a new population with the machine, and that triggers the above mentioned costs of 2000 - 3000 EUR.

Sorry, it looks I can't count to 4. I gave two options numbered 2. I meant:

1- new PCB with old RAM (if this works we're happy, but the test sounds more expensive than 2 so maybe exchange 1 and 2?).
2- old PCB with old RAM (and new EMIF settings).
3- new PCB with new RAM (good but you'll lose money)
4- old PCB with new RAM (you'll lose more money and whatever the new PCB fixes, which I'm not sure it was required after all, or nice to have).

So I guess you mean your order is 2,4, 1||3 in parallel. ?
Can you exchange RAM chips on an old PCB (for 4) without a population run ?
 
Back
Top