Oookay, quite a few posts to reply to, let's start
My expectation was that seven years after placing my preorder, I would have a Pyra in my hands.
Yes, that was my expectation as well. If I would've known upfront that it will become another trip to hell and (hopefully) back, I wouldn't even have started it. So we're on the same side here.
Yes, I knew there was risk. Yes, it was my decision to pay in bitcoin. Yes, a few Pyras have shipped, so it's not completely vapourware. Let me change my assessment from "vapourware" to "in development hell".
That is correct, and something I had hoped we wouldn't run into - but it happened and therefore I've thought about various solutions and how to move forward.
We paid you money because we thought you could deliver a product. I am upset because you are not taking responsibility for the lack of progress. Infrequent updates where you say "things have gone wrong, but I'm still doing stuff -- please keep waiting" is simply not good enough.
Well, basically I took responsibility from the very beginning - by not locking your payment to the Pyra, as it is usually the case with other crowdfunded projects.
Whenever you think it takes too long and it will never happen, you could simply use the voucher code and buy something else, regardless whether it's a GPD handheld, games or joysticks.
You could also ask for a refund. This takes a bit longer (as I'm still way behind with my eMails), but there was never a problem receiving a refund.
So... what are you expecting?
Taking responsibility for anything that goes wrong does mean you're acknowledging that you owe someone something.
And that's what I'm doing, because...
- I'm working on the assembly - albeit it's slow we're moving forward. That's for everyone who is still interested.
- I'm trying to keep everyone in the loop - but not that much happens. Companies are now even slower than before (so case coating took 2 months!), Fixes are happening, but also take a while (like the ongoing AESS Fix). They also do cost money, so there's a bit of a hold whenever I don't have enough spare money left.
- Everyone can use the voucher code for something else, so I'm taking responsibility in the amount of money I do owe you.
So what kind of responsibility are you expecting?
ED, you need to have a very honest conversation with us about exactly WHEN you think you can fulfill remaining preorders.
And how would that work? I could do that approximately for the first 500 units, as we have the produced boards for these as well and the new cases are currently being coated.
However, we don't know yet how many will have issues, so it's still a guess.
For the next 500, it could only a wild guess, as it depends on the availability of parts.
Heck, I'm trying to start the production of the next 700 Retrode2 for a while now. 2018 this was pretty easy. After placing the order, I could expect to receive the assembled units around 3 - 4 months later.
Since 2021, some parts either have a leadtime of up to 30 months (yes, really!) or need to be bought for a way higher price from a broker (which would double the production costs of the Retrode2).
Production this time is horrible - and it's not only us. I never had problems receiving more finished units from GPD, but I have not really received a significant amount of GPD Win Max 2 since January, as production simply can't keep up.
The Pyra has a LOT more parts compared to the Retrode2, and is a lot more complex. So yeah, no one could do any estimations for those at the moment.
You need to set a deadline, or else admit you cannot deliver.
... and then? What would be different?
The vouchers can already be used to buy something else.
Do you expect me to throw away all the populated PCBs and parts (worth over 120k EUR) if I can't meet that deadline?
It's up to everyone how long he/she still wants to wait, so everyone can set up his own deadline.
My personal deadline is "when I'm dead". And I certainly can't give you an exact date when I'll be dead
You cannot keep extending the project indefinitely. I genuinely want you to succeed, and I want a Pyra in my hands that I can use and show to my friends. But I have come to the conclusion that if this were possible, you would have done it by now.
Yes, so you set your own deadline. What keeps you from using the voucher code for something else or get a refund.
But that doesn't have to be the same for everyone else.
The obvious problem is manpower. You simply don't have the human resources this project requires. You need to raise some capital, hire some engineers, and devote more of your own time to this thing.
Besides the problem that good engineers are hard to find here in Germany, you are aware that this would raise the costs around 10k - 20k per month?
The Pyra will never-ever make that much money, even if we could deliver 1000 a month. It's still a niche product!
The problems are the one I stated, and yes, that's what lead to lacking manpower:
- Assembling a Pandora took 8 minutes, assembling a proper Pyra case would take 10 minutes (we've tested that with printed cases)
- Due to the fucked up case / mould, assembly of one unit takes around 30 - 60 minutes, as it needs a lot of manual labour work.
- The Pandora was assembled directly at the manufacturing facility (Global Componenes). With a rate of 80 EUR per hour, assembly of one Pandora did cost 10 EUR.
- The same was planned for the Pyra - but that wasn't possible for two reasons:
- If assembly takes up to 60 minutes, the assembly costs for one unit would be around 120 EUR now (due to raised wages)). For 1000 Pyras, that would be 120k EUR!
- As assembly is not simply assembling parts but manually CNC'ing them and testing if it does work now, it's not something that can be reliably done by your average person
That's the reason I do that myself whenever I find the time. Paying 120k just for assembly is not viable - and it wouldn't really be possible anyways.
The new cases are improved and lower the amount of manual labour work, so I should be able to do it in 20 minutes instead 60, but it still cannot be done by any random person.
But to create a proper case that can be assembled by anyone would need a complete redesign.
So yes, as I said before: The biggest issue is the case. Would the cases be as easy to assemble as the Pandora was, then I would receive 250 units per week (that's what GC did with the Pandora), and I would not be lacking manpower as I wouldn't have to do anything myself.
The only reason this is not possible is because of the case. Is that understandable?
TL;DR You cannot keep asking us to wait.
I never did. I always left the decision up to anyone how long he/she wants to wait.
It is more important to have more DMIPS or to have better support, new kernels, drivers, proper sleep/suspend mode, less heat inside?
A mixture of both. We spent around 150k to get the OMAP5 CPU boards to the state they are now.
Designing a new CPU PCB would be at least 50k as a guess.
So the next CPU board should really be an upgrade. If the board itself wouldn't be a lot better, it would be worth the money to spend on it's development!
NXP i.MX 8M has less performance using only one core but:
1/ It is 64-bit. I suspect you can't compare DMIPS from 32 bit to 64 bit CPU (of course if you are using 64-bit mode. Note these modes on ARM are separated, i.e. it is not like on Intel x86 where each ISA extending from x to 2x bits add up more instructions to the set mixing all them; on ARM 64 bit is a new ISA different and incompatible from 32 bit old ISA, you can't mix code, CPU is on 32 bit or on 64 bit mode and run very different software on each side).
Yes, 64bit is the only thing we're really missing on the OMAP5. I agree to that, that's basically one of the main reasons I'm looking into possible candidates.
2/ Do you remember the headache it was to support 4GB because that is the theoretical limit for a 32 bit system. No problem on 64 bit CPU mode.
That was a bit different - the kernel was missing LPAE as TI never officially made a 4GB board of the OMAP5. Usually, vendors do include all the code for that in the bootloader / kernel, and that was also planned for the OMAP5, but was only half-finished.
3/ With all cores it is a bit faster but more efficient.
Not necessarily. That depends on the slowest speed the cores can run, whether these have proper sleep modes. It also depends ont he nm the SoC is produced and whether it's designed to be used in a low-power-device.
The same SoC is often available as consumer and automotive version, and the automotive version needs a lot more power and is less efficient.
4/ Sleep/suspend mode is a MUST on a pocket computer. I have a lot of pocket computers since 80's and all of them were great on suspend lasting a lot of time on a charge or battery, some of them months, a few even more than a year. Even my Android smartphone if suspended (instant on but wifi/etc off) last a month. In a pocket computer (this is not only a console but a pocket computer) that factor is a must: it is needed it to be instantly on and ready from suspend when you pick up form your pocket. Even 14 hours (when Nikolaus finish on AESS) is ridiculous: you can pick up from your pocket to use it and... no battery :O
Well, there's no suspend on ARM devices, the cores go to a low-power sleep mode where they're still waiting for something to happen (otherwise a smartphone couldn't receive calls or messages anymore).
The OMAP5 can do that, it's just not yet properly configured in software.
The Pandora was no different - it was able to work for about 10 hours with the lid closed. After adding that, it easily worked for 25 hours in sleep mode.
The problem is, that notaz doesn't really have time anymore. He was basically the one awesome developer who made the Pandora as great as it was.
We're missing someone like him on the Pyra. We have a few guys helping out, but there is also more stuff to test / fix on the OMAP5 than on the Pandora.
If we get SmartReflex to work properly incl. RAM low-power mode, that would solve
I know this is hard/bug problem with OMAP5 (mostly used on automotive industry where suspend is not required), and there is no solution for it.
Well, yes and no. The OMAP5 also crashed regularly in powersaving, and that has been fixed a few weeks ago.
Also, TI ran into issues using the TILER as well - so they simply marked that as a hardware issue and their released workaround slowed things down horribly (rotating the screen slowed things down to 15fps - from around 200).
That was the reason we looked into the rotator chip.. until someone from the community took a look and fixed the problem.
So just because TI marks something as broken doesn't mean it really is. It still needs someone to look at though.
There is sense working on a new CPU board if offer much improvement on mentioned aspects. I see them much more important than raw performance.
Yes, that's why I generally like the i.MX and was looking forward to it. But it didn't bring enough improvements compared to the OMAP5 - development of a new PCB will still cost quite a lot!
When Librem 5 and MNT choose that NXP SOC they did for those reasons: good support, LTS and long term availability.
Yes, had we started when they started I would've chosen that one as well. When we started development the i.MX6 had been released and NXP mentioned the next SoC will probably only have an automotive version released.
Back then, the consumer version of the OMAP5 was still planned, no one knew back then we will only get an automotive version...
And yes, it took a few years until a consumer version of the i.MX8 has been released.
You've just explained exactly why he needs crowdfunding while assuming he can't.It wouldn't raise funds for new orders since that wouldn't be part of the crowdfunding (or if it was, it might not be the most popular option as you suggested due to mistrust). The funding would be for those with existing orders to get a guaranteed assembly and shipping date. He doesn't need to promise anyone else a new order. Those with pre-orders were going to be asked for more money upon shipping any way, so crowdfunding just allows ED to get that money early and charge a bit more to cover the costs. Surely I can't be the only one waiting on a pre-order who would pay more to get a guaranteed assembly/shipping date that actually gets met.
I personally dislike crowdfundings which try to raise more and more money later on, unless it's really assured these help stabilize the production. But that can't be assured here, there's so much stuff going on.
Did the donation page go away?
What I can give doesn't have enough 0's but I'm happy to kick in a bit.
The link at the bottom of the shop page to https://dragonbox.de/en/home/donation-to-the-shop.html
says "The product is no longer available"
Yes, I had to update the shop system to a new version and haven't implemented the "pay any amount" plugin yet. That's why it's not available at the moment.
Well we all know that production costs went up from before corona, so why not simply ask for more money for existing preorders? Problem is that the G7 just agreed for more sanctions, meaning the price of oil will raise again and again. With the oil everything will get more expensive.
The result is the more time we wait the higher the price will become, if the calculated loss is nearly 300€ right now it won't stay below 300€ for much longer. That much is certain.
It least create the poll mentioned above. No one could have foreseen the economic situation with the high inflation (at least in europe) and component shortage.
Hm, but I would first like to wait until I've tested how well assembly works with the new cases when I receive them back from coating. I want to be able to assemble units in a reasonable speed before I do something like that.