<lekernel>
steve|m: this is bullshit, please do not propagate it
<lekernel>
see opensparc and leon for open source processors of much better quality than openrisc, which were made into ASICs, and did not trigger the announced revolution
<steve|m>
lekernel: thanks for sharing your opinion
<wolfspraul>
steve|m: we need creative ideas how we market, or tell the story, of the Milkymist SoC and Milkymist One video synthesizer
<wolfspraul>
any ideas?
<wolfspraul>
unfortunately when you do really pioneering stuff like Milkymist, and the same is true for the opencores donation call, it's super hard to find the right line between wishful thinking/fairy tales, and down-to-earth stories that are not too boring on the other hand...
<wolfspraul>
I'm still struggling with this
<wolfspraul>
Milkymist is a GPL licensed SoC
<wolfspraul>
that's a cool thing. there are others but I think in terms of really making it work in a practial way, Milkymist may be 'the best'
<wolfspraul>
then there is work on free synthesis tools, llhdl/antares. That's extremely interesting too, but still work-in-progress and needs serious contributors.
<wolfspraul>
and there is Milkymist One, a video synthesizer we are selling for 499 USD. What's the point of a 'video synthesizer'?
<mth>
not exactly a lively croud, but it does show both the result and the mixing taking place
<mth>
I could ask if he's interested in giving some feedback?
<wolfspraul>
of course, that would be fantastic!
<lekernel>
wolfspraul: for the LGM PR, we can title it with something like "Sneak peek at the upcoming Milkymist video synthesizer shown at LGM"
<lekernel>
then we are free to talk about many topics in the body
<wolfspraul>
hmm, ok. I need to start this asap then fill in details with jon...
<lekernel>
wolfspraul: let's not publicize llhdl/antares too much yet. first they do not work today, second they might irritate xilinx a bit (probably not much, though).
<lekernel>
in the end I don't think much good would come from such a PR, except people asking me all the time when it's going to work or proposing tons of ideas in endless discussions
<lekernel>
plus the usual trolls on the FPGA companies
<lekernel>
later, we should definitely do it though
<wolfspraul>
understood
<wolfspraul>
mth: 17 views for that video!
<wolfspraul>
you can ask him right away about how we all gain much more visibility for this kind of work
<wolfspraul>
I'm still downloading (caching) the video, super slow vpn link...
<wpwrak>
wolfspraul: the link probably gets slower with every keyword they add to the list of forbidden words ;-)
<wolfspraul>
steve|m: just got the opencores newsletter. I pick one thing, quoting
<wolfspraul>
"Some of us who have worked for Semiconductor companies have seen the large margins and the large price differences depending if you were a large customer with many different products, or if you are a smaller company with a great new product. Since price always is important , it is today almost impossible to compete with large companies even if you have a "better" product."
<wolfspraul>
I don't understand how they came to this conclusion.
<wolfspraul>
I know a lot about IC prices, availability, moq, etc. I just don't see it that way.
<wolfspraul>
but it's very hard to discuss these things because there are so many moving targets and numbers. Maybe they should give us an example :-)
<wolfspraul>
which chip they had trouble sourcing, and they think they can make cheaper now...
<wpwrak>
wolfspraul: well, there are chips you can't get easily in a usable condition (that is, also with the documentation you need to make them work). think of telephony or even wlan. not sure if that's what they mean, though.
<wolfspraul>
but even that is wrong
<wolfspraul>
if you look at digikey, where are they weak?
<lekernel>
wolfspraul: opencores is 99% bullshit
<wpwrak>
there are no such chips at digi-key :)
<wolfspraul>
nand (well, you would never buy that anyway, right? :-)), memory, rf ics, baseband, also weak on image sensors afaik
<lekernel>
also they claim to be a 'community' while it's only 2 or 3 people at ORSoC making all the decisions and you only hear about them when all is said and done
<wolfspraul>
I think you will have trouble getting a 16 GB nand chip, Intel/Samsung/Toshiba, from digikey
<wolfspraul>
or an MTK anything
<wolfspraul>
but...
<wolfspraul>
that's only digikey
<wolfspraul>
there are myriads of smaller distributors all over Asia
<wolfspraul>
starting to work with them is itself a lot of work though, so maybe some people stop at digikey?
<wpwrak>
for things like baseband, the problem is also documentation. they won't give you that either. (well, unless you put a lot of money on the table.) so even if you find the chips on the secondary market, you'll have a hard time using them.
<wolfspraul>
fair enough, but opencores does not talk about documentation
<lekernel>
the views expressed in that newsletter are only these of those 2-3 ORSoC people
<lekernel>
nothing more
<wolfspraul>
lekernel: sure, I know
<wolfspraul>
I just try to respond to some specific thing they say.
<wpwrak>
and yes, some people probably simply don't know how to do sourcing ;-)
<wolfspraul>
I really massively do not agree with that statement.
<wolfspraul>
whatever they do will be phenomenally more expensive than if they had just bought it
<lekernel>
if you start reading more of the crap that gets published by ORSoC under the name opencores.org, you'll probably find tons of stuff to disagree with
<wolfspraul>
if they really think there are some 'evil' big corps that are scheming against the small guys
<wolfspraul>
it's not true
<lekernel>
yeah but it's romantic and it gives them something to say
<wolfspraul>
sure
<lekernel>
maybe it covers up their failures too, so they can sleep in peace
<wolfspraul>
well I'm a bit old fashioned here, we have a responsibility to accurately report our findings.
<wolfspraul>
otherwise we cannot improve
<wolfspraul>
and that paragraph I don't like. they claim to speak from experience, and say there is this 'big corps get large discounts' thing they discovered.
<wolfspraul>
but it's wrong
<wolfspraul>
it's a bad misrepresentation of reality
<lekernel>
you know, those orsoc guys live in a very bizarre world
<wolfspraul>
so they are failing on their leadership
<lekernel>
a bit like those conspiration theory advocates
<lekernel>
oh, they are failing on tons of things
<wolfspraul>
yes, fine. but either there is a conspiracy or not :-)
<lekernel>
the only thing they do approximately right is attract attention
<wolfspraul>
someone may read this letter, or this channel, and really just don't know
<wolfspraul>
"is there such a thing?"
<wolfspraul>
"are large corps really conspiring against the small guys?"
<wolfspraul>
"is that what we have to fix for our break-through?"
<wolfspraul>
so my answer is: no
<wolfspraul>
someone can listen to me or not, but that's my answer :-)
<wpwrak>
well, there are exclusive deals. e.g., you probably can't get exactly the socs apple use. not sure if this is an issue to get excited about, though :)
<lekernel>
wolfspraul: you could spend days finding things to disagree with in the orsoc world
<wolfspraul>
that's not the thing that holds us down, or stops the 'revolution' :-)
<wolfspraul>
wpwrak: Apple SoCs!
<wolfspraul>
they hired 200+ IC designers
<lekernel>
if you like this kind of thing, you can also read stuff like reopen911, abovetopsecret, etc.
<wolfspraul>
of course yes, ahem, the 'large guys' (apple) is 'conspiring' against the small guys (everybody else) by not giving away the results of their hard work
<wolfspraul>
but that's not what orsoc means
<wolfspraul>
they are not advocating piracy
<wolfspraul>
lekernel: yes sure, we are on the same page, no worries :-)
<wolfspraul>
that one feedback is all I'll do to digest the newsletter :-)
<wolfspraul>
no it's important for us to realize - there is no such conspiracy
<wolfspraul>
yes, of course, if you buy a chip 10 times, the unit price will be higher than if you buy the same chip 100 times, or 1000 times
<wolfspraul>
but the increases for lower order sizes are very reasonable
<wolfspraul>
and the amount of investment you (!) need to make when scaling up your run (getting yield under control) will be far higher anyway
<wolfspraul>
anyway...
<wolfspraul>
:-)
<wolfspraul>
I see bigger price differences in mechanical parts actually
<wolfspraul>
a Molex ethernet connector may cost 8 USD, but a Taiwanese one only 1 USD
<wolfspraul>
maybe I find out more about that one day, the differences in mechanical parts still seem extreme to me, showing how little I know about mechanical parts...
<wolfspraul>
the Chinese one would probably cost 20 cents, but start to rust after 2 weeks...
<kristianpaul>
is a display considered mechanical part as well?
<wolfspraul>
I'd call that a module
<wolfspraul>
quite a few things inside
<wolfspraul>
the mechanical part I'd say is only the casing
<kristianpaul>
how fair are prices for modules?
<wolfspraul>
if you take one apart, you will see the various layers inside
<kristianpaul>
ah ok
<wolfspraul>
same
<wolfspraul>
sourcing is extremely competitive!
<wolfspraul>
if you think sourcing = digikey, then something is massively wrong with you
<wolfspraul>
:-)
<wolfspraul>
you can buy all sorts of lcm modules, at good prices even in low quantities
<wolfspraul>
but you need to find the proper distributor
<wolfspraul>
and if you order low quantities, you will have very little say over EOL, for example
<kristianpaul>
s/sourcing/hunting then ;)
<wolfspraul>
when the product is discontinued
<wolfspraul>
not really, it's just thousands and thousands of companies doing this
<wolfspraul>
for example the LCM for NanoNote
<wolfspraul>
Giantplus
<wolfspraul>
we buy from a distributor in Shenzhen
<wolfspraul>
not street market
<wolfspraul>
proper distributor with office etc.
<wolfspraul>
anybody can buy there
<wolfspraul>
they focus on the money, not on who you are
<wolfspraul>
guaranteed
<wolfspraul>
moq is 12 pieces, that's 1 tray
<kristianpaul>
nice :-)
<wolfspraul>
they rather sell you a whole box, easier for shipping. 1 box = 12 trays = 12*12 = 144 pieces
<wolfspraul>
but 1 tray also ok, they will wrap it for you
<wolfspraul>
they ship worldwide
<wolfspraul>
and so on
<wolfspraul>
what's the problem?
<wpwrak>
(digi-key) for chips, digi-key seems pretty okay. mechanical things or modules are an entirely different story, though ...
<wolfspraul>
yes
<kristianpaul>
hmm so if in 10 years i may get a lcm replacement for my nanonote? or should i order spare parts today?
<wolfspraul>
but sourcing != digikey
<wolfspraul>
there are THOUSANDS of electronic parts distributors
<kristianpaul>
thats shopping, totally different tpoic i think (digikey)
<wolfspraul>
kristianpaul: it's a fairly common model, but I don't know how long you can really get just that one
<wolfspraul>
finding that giantplus distributor was a bit hard btw
<wolfspraul>
we couldn't find them in shenzhen itself, city is too big and too chaotic
<wolfspraul>
so had to call Giantplus Taiwan and claim that we wanted to buy 3k
<kristianpaul>
hehe
<wolfspraul>
that's the minimum where their sales rep will continue to talk to you, and everybody just claims 3k
<wolfspraul>
then they give you the contact of their distributor in shenzhen
<wolfspraul>
and there you can buy 1 tray
<kristianpaul>
so find distributors is based on social interaction?
<wolfspraul>
absolutely
<wpwrak>
wolfspraul: maybe that's the problem of openrisc: they don't know when they're supposed to lie ;-)
<wolfspraul>
there are many distributors all competing with each other
<wolfspraul>
calling it a day
<wolfspraul>
sourcing is very diverse
<kristianpaul>
china distributors jungle :-)
<wolfspraul>
opencores talks about chips (semiconductors), that's actually easy because well standardized
<wolfspraul>
modules are probably hardest
<wpwrak>
particularly if you can customize them ...
<kristianpaul>
wonder if there are street market for chip fabbing
<wolfspraul>
not just CHina, it's exactly the same in Taiwan, and I would think Korea and Japan too, although I have little/no experience there
<kristianpaul>
ok
<wolfspraul>
what I did see is that sometimes the 'big corps' are settling IP fights among their customers
<wolfspraul>
but that's an entirely different subject
<wolfspraul>
and if they do that, they have to be very careful to not be dragged into some bad activity, that may even get them big cartel fines etc.
<wolfspraul>
but actually a small guy would be way under the radar for this kind of thing
<wolfspraul>
that's also not what opencores means
<kristianpaul>
hum..
<wolfspraul>
like Sebastien said - they are just a bit romantic :-)
<wolfspraul>
we should leave that to them
<wolfspraul>
hail the revolution!
<wpwrak>
opencores - a cleverly arranged diversion ;-)
<lekernel>
wpwrak: actually I don't think opencores has always been like that
<kristianpaul>
how old is opecores btw
<lekernel>
was founded by http://www.lampret.com/ who then abandoned it, and the domain name and website were transferred to ORSoC AB, some small Swedish consultancy (not sure under what terms) who are now super anal retentive about everything and say and do a lot of crap
<lekernel>
btw all the donation money goes directly to that company (not to the "opencores community") without any control
<kristianpaul>
bad thing
<wpwrak>
now there's an interesting concept. how about openclubbing to lekernel's account ? :)
<kristianpaul>
lol
<lekernel>
openclubbing?
<wpwrak>
something nice and broad that's a superset of vjing
<kristianpaul>
ah, i tought you mean transfer opencores domain to lekernel ;)
<wpwrak>
that's an even weirder idea ;-)
<wpwrak>
well, when he finishes llhdl, maybe they'll offer him the domain as a form of tribute ;-)
<lekernel>
hahaha
<lekernel>
in your dreams
<lekernel>
they're control freaks. they hold back every little they have, that is, some public visibility which includes that domain name.
<lekernel>
and they give little shit about stuff like llhdl
<lekernel>
(I'm speaking only about the ORSoC leaders here, not the wider Opencores community nor all ORSoC employees)