Wolfgang Spraul wrote:
hadez,
Hi Wolfgang,
first of all after thanking you to buy a board, I want to thank you
for
reporting the freezing problem and being persistent about it. Thanks!
I am very carefully following the thread...
No problem at all. Thanks a lot for actually caring about the issue.
Just this morning I received an email from a potential customer in
the
touring live music industry in the US, asking whether Milkymist One
is
ready for professional stage use. Well, I think it's not (yet), but
if
we continue a bit more on stability and robustness, we definitely
want
to get there. So...
Xiangfu will try to run a little stability/freezing test program, for
example let his board run every night, write down whether it froze or
not. I will do the same with my board. Let's collect more data.
Can you do me a favor and let your board sit in GUI mode for an
extended period of time? Like 10 hours or so? I just want to make
sure that's it's really something you only see in render mode.
So far whenever I've had it run in GUI mode, nothing happened.
But I've really not had it running for 10+h. More like 3-ish.
I've got this set up and will let you know the results.
Then we wait until Sebastien is back and see what he has to say about
this thread. If your board allows us to reproduce a freezing problem
(whatever the root cause) easily, Sebastien may want to take a look
at
your board. Where are you located? Did you buy only the board, or
also
a case?
We can swap your board with one that doesn't exhibit freezing
problems,
if we determine that the problem is specific to your board...
(maybe we can also give you a free camera to thank you for your help
in tracking this down)
I've bought only the board.
Right now I'm located in Stuttgart, Germany.
If anyone wants to take a closer look at the board, I can pack it up
and ship it to whoever would like to do that.
What MAC address does your board have? There should be a hand-written
small label somewhere (check the VGA connector). The test results for
every board are archived in the wiki, maybe if we know your MAC
address
we can find a clue there...
http://en.qi-hardware.com/wiki/Milkymist_One_RC2_Test_Plan#Test_Program_Procedure
I've got #16 which does in fact show the slightly suspicious note
"entire side of dram chip lifted up, had to fix that first".
What I'm seeing could very well be a memory issue.
If you think it's a good idea, I do have access to a hot-air solder
rework station to reflow the chip myself.
However, if there's an issue with the chip itself that will definitely
not help.
It would be neat to have a working board on June 18th, but that's not a
hard requirement ;)
Lets just stick to whatever debugging step you think is most suitable.
For the upcoming RC3 run, I am thinking about extending the test plan
to include a durability test, where we leave each unit in rendering
mode for some time, maybe 2 hours or so. We need to see how we can
pull it off effectively on a run of 80 units, but it could provide
us with interesting data, in case we can catch offenders.
In hardware, every board is unique, and has slight differences from
other boards. That's what makes it hardware, not software. So in the
end we may find a way to tune the software (including the soft-IC
design)
in such a way to fix the freezing, or we identify a hardware
(electric)
improvement or test that will allow us to increase the hardware
quality
coming from the production line.
Yeah, good old hardware. Debugging is definitely less straight forward
than with software.
It's difficult, and we depend on your patience and continued help to
track it down. Thanks a lot for your feedback, you already triggered
some good actions on our end.
You're welcome. Thanks again for the quick replies.
Best Regards,
Wolfgang
--
Best regards,
hadez
_______________________________________________
http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org
IRC: #milkymist@Freenode
Twitter: www.twitter.com/milkymistvj
Ideas? http://milkymist.uservoice.com
