Analyzing 4GB RAM


Almost 18 months since preorders have started. Does anyone feel responsible?

Yes a lot of people and things are responsible. Greece company, RAM modules that wont work, keymats, everyone else and at least the chinese new year.
 
I'm holding out for the Pyra because, I really believe once its out there, this scene will produce some real gems. That happened with the Pandora, the GP2X, the Wiz and all that beforehand .. so its really the *only* reason I'm still hanging out for the Pyra. I believe in ED, and his gang of misfits, and the journey is the reward as well.

That said, the world is moving on, folks. I have to say that my favourite gaming platform right now, is the GPD Pocket paired with my (still working great) iControlpads... this has become the new reigning pocketable rig, alas...
 
Sounds like Nikolaus is making good progresses on zooming in on the RAM issue (4GB RAM surprise):
https://www.pyra-handheld.com/pipermail/kernel/2017-October/thread.html

From my quick scan of the discussion it sbeems to be a hardware problem needing some fixes to the PCB.

Yes, for me this is realy good news. The problem has to be exactly located now. Then the pcb has to be reworked and new pcbs have to be produced and tested. If everything goes well, this will only take 6 month. With series production we can have a pyra in only 1 year from now.
 
Yes, for me this is realy good news. The problem has to be exactly located now. Then the pcb has to be reworked and new pcbs have to be produced and tested. If everything goes well, this will only take 6 month. With series production we can have a pyra in only 1 year from now.
...If you really think it's going to take them that long, you're crazy, sir.

Once they have it narrowed down for sure, about all that'd be left is seeing how they'd fix it, and if it really does require a new board (i.e; can't be solved with capacitors or whatnot), to try to get any money back from printing the old boards.
 
Ok, so what is your estimation for a bugfree prototype and the shipment of the final product?
 
I read through it:

TLDR: Old traces on the boards that were fine for 2GB, experience noise/drop. However external wires do not fix the situation. So the issue is still not definitive.

This is frustrating as it seems like everything is good for 2GB, and scope creep has sprung up to bite the Pyra in the ass.
 
Ok, so what is your estimation for a bugfree prototype and the shipment of the final product?
Once they got the EXACT cause narrowed down and a solution in mind, all that should be needed is either retrofitting of the old boards, or (more likely based on what Nikolaus has been saying) a new board with redesigned traces. Proto that, test that, if it's all good, make it.

Cases can be made whenever, presumably once ED moves the molds.

I'd say 3-6 months, and that 6 is rather generous. I think it'll be more like 3-4 realistically.
 
Once they got the EXACT cause narrowed down and a solution in mind, all that should be needed is either retrofitting of the old boards, or (more likely based on what Nikolaus has been saying) a new board with redesigned traces. Proto that, test that, if it's all good, make it.

Cases can be made whenever, presumably once ED moves the molds.

I'd say 3-6 months, and that 6 is rather generous. I think it'll be more like 3-4 realistically.

Hmm My money is on the following:

4 Weeks to finally dial in the issue
6-9 Weeks to redesign (keep in mind there are around 22 layers to the PCB)
6 Weeks for another prototype run (this was what the others took)
3-4 Weeks of testing
6 Weeks for a production run

Total: 25 weeks best case, or about 6 months (rounded). Further issues will drive the timeline even further. IMO probably another year, which would be bad with the Psion and GPD Win 2 coming.
 
Hmm My money is on the following:

4 Weeks to finally dial in the issue
6-9 Weeks to redesign (keep in mind there are around 22 layers to the PCB)
6 Weeks for another prototype run (this was what the others took)
3-4 Weeks of testing
6 Weeks for a production run

Total: 25 weeks best case, or about 6 months (rounded). Further issues will drive the timeline even further. IMO probably another year, which would be bad with the Psion and GPD Win 2 coming.
Well it is still possible that EvilDragon will choose the 2GB solution for the already produced boards. Then we could see first Pyra's way faster in the wild

Gesendet von meinem SM-N9005 mit Tapatalk
 
TLDR: Old traces on the boards that were fine for 2GB, experience noise/drop. However external wires do not fix the situation. So the issue is still not definitive.

This is frustrating as it seems like everything is good for 2GB, and scope creep has sprung up to bite the Pyra in the ass.
I'd actually view these as a blessing in disguise. Yes, it was good enough for the 2GB models to have no issue, but if the increased demands of the 4GB chips made it fail like that, then we're talking a circuit that was just barely robust enough.

This redesign will probably benefit both 2 and 4GB Pyras. Of course, if some people don't care, then the 2GB ones can be fulfilled with the old boards, but then hopefully there is no situation that stresses noise on the RAM.
 
Last edited:
I'd actually view these as a blessing in disguise. Yes, it was good enough for the 2GB models to have no issue, but if the increased demands of the he 4GB chips made it fail like that, then we're talking a circuit that was just barely robust enough.

This redesign will probably benefit both 2 and 4GB Pyras. Of course, if some people don't care, then the 2GB ones can be fulfilled with the old boards, but then hopefully there is no situation that stresses noise on the RAM.

Traces should only be "just robust enough." If traces are too thick you would run into other problems. Plus keep in mind the following: Traces are not a wear item, voltage and amps are controlled running through them.
 
Traces should only be "just robust enough." If traces are too thick you would run into other problems. Plus keep in mind the following: Traces are not a wear item, voltage and amps are controlled running through them.
Yes, and there is a difference between "just robust enough" and "just barely robust enough."

That extra word - or its lack thereof - makes all the difference. That's the point I'm making.
 
Yes, for me this is realy good news. The problem has to be exactly located now. Then the pcb has to be reworked and new pcbs have to be produced and tested. If everything goes well, this will only take 6 month. With series production we can have a pyra in only 1 year from now.
Indeed. Beginning of 2019 sounds feasible. If everything goes well then somewhere in the third or fourth quarter of 2018.
I highly doubt it will be any quicker than that. But as always I would love to be proven wrong on this one.
  • Oktober ~ mid November: localize RAM error
  • November ~ December: design fix
  • January/February: produce new boards and test
  • March/April: rework for the final to be discovered issues
  • May: Final version, CE testing and final pre-production tests
  • June/July/Aug: Start Production and delay due to holidays.
  • Sep: Assembly, testing and shipping of first units
  • Oct/Nov/Dec: Final units in queue arrive
Here I assume that work on the case will be done in parallel. I am sure I have forgotten something.
There will probably be a couple of forest fires, volcanic eruptions and localized nuclear war that will add a bit of delay.

Well, this is just my opinion and guesses, I am sure many others have more optimistic estimates and a clearer picture of what needs to be done. I do think that there is no reason whatsoever to assume it will be the optimistic case.
 
Traces should only be "just robust enough." If traces are too thick you would run into other problems. Plus keep in mind the following: Traces are not a wear item, voltage and amps are controlled running through them.
A memtest run is just sample - it cannot tell you how frequent rare bit failures are. I wouldn't feel save, if just doubling the demand shows problems like that.
I'd rather like passing memtest at double the target demand.
 
Indeed. Beginning of 2019 sounds feasible. If everything goes well then somewhere in the third or fourth quarter of 2018.
I highly doubt it will be any quicker than that. But as always I would love to be proven wrong on this one.
  • Oktober ~ mid November: localize RAM error
  • November ~ December: design fix
  • January/February: produce new boards and test
  • March/April: rework for the final to be discovered issues
  • May: Final version, CE testing and final pre-production tests
  • June/July/Aug: Start Production and delay due to holidays.
  • Sep: Assembly, testing and shipping of first units
  • Oct/Nov/Dec: Final units in queue arrive
Here I assume that work on the case will be done in parallel. I am sure I have forgotten something.
There will probably be a couple of forest fires, volcanic eruptions and localized nuclear war that will add a bit of delay.

Well, this is just my opinion and guesses, I am sure many others have more optimistic estimates and a clearer picture of what needs to be done. I do think that there is no reason whatsoever to assume it will be the optimistic case.
...Why do people think it's going to take them 2+ months to design a fix and then a whopping 5 or so months to produce, rework, and test the boards? Like tweaking things on a board that's already mostly been tested and passed almost all of those tests (except for this RAM issue and a few others that already got fixed) is going to force an entire from-the-bottom-up round of testing or something...

The boards have already been tested in those senses and passed all tests, and obviously it's in good sense for them to keep whatever changes are done as minimal as possible (so that they don't have to completely retest the boards).

Am I seriously the only one who thinks that the time from final assessment to fixing to producing new boards shouldn't take much more than three months here?
 
...Why do people think it's going to take them 2+ months to design a fix and then a whopping 5 or so months to produce, rework, and test the boards? Like tweaking things on a board that's already mostly been tested and passed almost all of those tests (except for this RAM issue and a few others that already got fixed) is going to force an entire from-the-bottom-up round of testing or something...

The boards have already been tested in those senses and passed all tests, and obviously it's in good sense for them to keep whatever changes are done as minimal as possible (so that they don't have to completely retest the boards).

Am I seriously the only one who thinks that the time from final assessment to fixing to producing new boards shouldn't take much more than three months here?
No, I also don't think it will take that long. But I'm glad some people are presenting other possible scenarios.
 
I hope it doesn't go too far into 2019. Go beyond April and it's sounding like we'll likely be off into the wild yonder of WTC trade rules here, so getting stuff from the dark empire might be tricky and kind of pricey.
 
...Why do people think it's going to take them 2+ months to design a fix and then a whopping 5 or so months to produce, rework, and test the boards?
Because we have been hearing for months about "the last fixes and then they are ready". There is always something more. I round everything up, way up, because of all the problems that I cannot predict. I have seen stuff be delayed, with the Pandora, at my own work. It makes you pessimistic in estimates, though I prefer to think of it as realistic.
Am I seriously the only one who thinks that the time from final assessment to fixing to producing new boards shouldn't take much more than three months here?
Yes, most people think it will only take two™.
 
Back
Top