<lekernel>
btw, as I understand it, the ramdisk uses the heap
<xiangfu>
that mean now. the ramdisk is 92M?
<xiangfu>
I can try to copy a 90MB file to milkymist for test :D
<lekernel>
should be 92M minus other allocated memory on heap
<saardrimer>
hello
<saardrimer>
tweet finally made me join this channel ;)
<lekernel>
hi
<lekernel>
welcome
<lekernel>
I'm Sebastien
<saardrimer>
hi Sebastien (sorry I was AFK attending to my plants)
<saardrimer>
I contacted you a while ago... this is me saardrimer.com
<saardrimer>
anyway, back then (March) I wasn't quite ready to talk about boldport in a public forum, so I didn't follow up. Now it's open... http://www.boldport.com
<kristianpaul>
yeah clicking right now.. still clicking ;)
<saardrimer>
load an example project (project_zero)
<lekernel>
it's SaaS, basically.
<kristianpaul>
;-)
<kristianpaul>
can you switch to agplv3 license btw?
<kristianpaul>
what you do with our stats?
<kristianpaul>
:-)
<saardrimer>
go to "Make" and hit "Generate Makefile"
<kristianpaul>
k
<lekernel>
kristianpaul: I'm not questioning the MIT license, just the SaaS approach.
<kristianpaul>
sure
<kristianpaul>
i'm questioning data retention :-)
<kristianpaul>
if there is one of course
<saardrimer>
you can save projects and recipes if you register and are logged in
<kristianpaul>
i like the idea, just wondering all the implication before try
<lekernel>
saardrimer: nevertheless SaaS tools are not acceptable for Milkymist when there is a working alternative.
<lekernel>
also, I hope we'll switch the synthesis to LLHDL some day :-) (which I will continue developing when I have sold a couple thousand M1s)
<saardrimer>
lekernel: sure. It's not quite 100% SaaS because I'm not actually running the tools for you... but just know that it exists, and like I said, I'm happy to help if there's a need.
<saardrimer>
I am an experienced Verilog FPGA designer too ;)
<lekernel>
saardrimer: btw, if you dislike the EDA tools you are welcome to contribute to LLHDL/Antares, the free toolchain I have started writing
<lekernel>
contrary to tools like Boldport, there is most probably zero money to be made, but it's fun to hack on.
<saardrimer>
one of my goals is to help people use EDA tools that they may not have heard of, but are as good as the alternative.
<lekernel>
hopefully, if I sell tons of M1s, I can fund those things.
<lekernel>
but I'm not quite getting there today, unfortunately
<saardrimer>
so I could add them to the software stack support at boldport
<lekernel>
oh, well, it's way too soon to promote the llhdl tools this way
<lekernel>
they just don't work yet
<saardrimer>
maybe one day ;)
<saardrimer>
lekernel: how many M1's have you sold, if it's not a secret?
<lekernel>
about 40
<saardrimer>
and you're aiming for a 1000?
<lekernel>
no, i'm aiming for 40000
<saardrimer>
why aren't people buying?
<lekernel>
because they don't know about it
<saardrimer>
so only 40 potential users turned buyers know about Milkymist?
<Fallenou>
well the price is not like an arduino :)
<Fallenou>
that's also one of the reason
<lekernel>
yeah, but I would phrase that one less diplomatically as "most makers are incompetent freetards"
<Fallenou>
but compared to other fpga boards, it's a normal price
<Fallenou>
even a low price compared to other SP6 boards I may add
<wolfspraul>
saardrimer: chronologically, there was an rc1, the very first manufacturing of this design
<wolfspraul>
we made 6
<wolfspraul>
all 6 booted, but all 6 had various major problems
<wolfspraul>
those were all unsellable and we never planned to sell them
<wolfspraul>
then there was an rc2, including everything we learnt from rc1
<wolfspraul>
we made 40, 33 turned out to be sellable (luckily, all rc1 bugs got fixed with no regressions)
<wolfspraul>
those 33 were all sold
<wolfspraul>
we are a little behind on rc3, ideally we would have rc3 ready now
<wolfspraul>
but I think it's another 4-6 weeks
<saardrimer>
hmmm... I'm not sure that the software-borrowed "RC" versioning works well for hardware you want to sell
<wolfspraul>
rc3 is a huge step forward from rc2, in all areas not just hardware design fixes, but also massive software progress, accessories, box, etc.
<wolfspraul>
it works very well, I think you understood me :-)
<wolfspraul>
from a marketing perspective, there has not been a 'launch' yet
<wolfspraul>
rc2 was sold in a chaotic way, some as boards only without case, some with unassembled case kits, with varying accessories, etc.
<wolfspraul>
not chaotic because we are chaotic, but because rc2 allowed people to join this project at a very early stage
<wolfspraul>
with rc3 we must lift this thing to the next level, get some of those products in front of buyers from retail stores, in front of product managers of companies who see business potential, in whatever way
<wolfspraul>
I'm just explaining the 'productization' because you asked about numbers, and Sebastien said we need to sell 'thousands' which of course we do.
<wolfspraul>
what's holding the success down right now, imho, is:
<wolfspraul>
successful launch
<wolfspraul>
quickly find a strong marketing partner or strong retail chain that is willing to share some growth risks
<wolfspraul>
'quickly' meaning within 6 months or so, that would be ideal
<wolfspraul>
for price, we launch at 499 USD
<saardrimer>
wolfspraul: thanks for the info.
<wolfspraul>
then we need to bring down the costs on the manufacturing side, first of all to create a retail margin
<wolfspraul>
if someone would come now and want to buy 10, 50, 100 or more, we could probably work out a nice margin quickly
<saardrimer>
what do you mean by "need"... is that to recover the investment some project members have put in?
<saardrimer>
(as in "need to sell")
<lekernel>
fwiw, the somehow comparable edirol v4 is 3 times the price of the M1 and they sold 40000
<wolfspraul>
the retail price needs to be driven down too, let's say 399 USD as the next step, but that should be 12 months out or so since it's more important to increase value for the user first, and build the distribution network
<lekernel>
but edirol doesn't suck at publicity as much as we do
<wolfspraul>
'need' as in support our activities moving forward, we have a todo list that is so long I won't even begin to list it here
<saardrimer>
yeah, success isn't about technological superiority (unfortunately)
<wolfspraul>
that's fine by me
<wolfspraul>
success is about creating value for the buyer, and getting that message across efficiently
<wolfspraul>
that's a bit magic too, I like it
<wolfspraul>
some people are successful at selling nothing, why not
<wolfspraul>
I like magicians :-)
<saardrimer>
so how do you plan to stop sucking at publicity?
<lekernel>
get marketing people on board
<lekernel>
write tons of press releases and articles ourselves
<lekernel>
find large buyers who would invest in the product and resell
<saardrimer>
sounds like a startup
<saardrimer>
are you treating it as such?
<lekernel>
that could describe it in many ways, yeah
<wolfspraul>
I'm a bit more careful about 'publicity' than Sebastien probably. If more publicity means that expectations are going up, we may shoot ourselves in the foot.
<wolfspraul>
so I'm more worried about expectation management
<wolfspraul>
and finding buyers early on that love the product and tell their friends about it
<wolfspraul>
we haven't been that crazy successful at that, imo
<wolfspraul>
I'm sold out of products right now
<saardrimer>
I too like to use "expectation management" ;)
<wolfspraul>
but from the ones I sold, I'd say 80% have disappeared into a drawer somewhere
<wolfspraul>
that's bad
<lekernel>
wolfspraul: probably, having the case and the new "out of the box" software will help.
<wolfspraul>
sure, it will
<saardrimer>
right. Maybe you should be selling is for *more* money?
<wolfspraul>
I'm very optimistic :-)
<wolfspraul>
Sebastien sinks his time and passion into this project, I sink my savings into it :-)
<wolfspraul>
499 USD is a good price
<saardrimer>
wolfspraul: :)
<wolfspraul>
I'm a manufacturer, technology is about driving prices down
<wolfspraul>
I don't like luxury technology, actually I hate it
<wolfspraul>
luxury is for other goods, not technology
<saardrimer>
that really depends who you're selling against
<lekernel>
mounting the case and reflashing isn't hard, but a lot of people have very little motivation to start with ...
<wolfspraul>
so we start at 499, need to build out the product, in terms of yield, accessories, etc.
<saardrimer>
and what's the total market for the product
<wolfspraul>
lots of numbers moving in the background
<wolfspraul>
then we need to have a retail margin
<wolfspraul>
I sell for 499 USD now, shipping with fedex to end users
<wolfspraul>
but if I sell to retail, I cannot charge 499 USD of course
<wolfspraul>
499 is the base that lets me work on those things - accessories, yield, retail margin
<wolfspraul>
I need to stabilize all that first
<wolfspraul>
then bring the price down
<wolfspraul>
in parallel of course we work on increasing value (real or perceived) with our buyers
<wolfspraul>
that's separate
<wolfspraul>
all normal stuff imo
<wolfspraul>
saardrimer: we are marketing it as a 'video synthesizer' now, which I think is genius
<wolfspraul>
before it was an 'interactive VJ station' but that was hard to chew on, too much to digest
<wolfspraul>
reaction was "oh, I'm not a VJ, so it's not for me"
<saardrimer>
wolfspraul: there's an explicit joke in there somewhere ;)
<wolfspraul>
as a video synthesizer, potential market size is easily tens of thousands of units
<saardrimer>
sounds like a right move
<wolfspraul>
musicians (hobby or professional), night clubs, bars
<wolfspraul>
can be used at conferences to entertain people in breaks
<wolfspraul>
as always - theoretically a lot of things are imaginable, now it depends on execution
<wolfspraul>
time is limited, the world is big
<saardrimer>
wolfspraul: I know!
<wolfspraul>
that's why we need a strong marketing partner asap, as soon as we have a complete product up and running, with software, accessories, box, etc.
<wolfspraul>
so...
<wolfspraul>
in which way do you want to help?
<wolfspraul>
:-)
<saardrimer>
If only I was good at marketing, boldport.com would be in a better place... ;)
<wolfspraul>
I share lekernel's feedback about that
<saardrimer>
not that it's bad now, just "launched" it a few weeks ago
<saardrimer>
wolfspraul: which part?
<wolfspraul>
which may still make it a good business, but from our 'independence' standpoint, nah - doesn't fit :-)
<saardrimer>
wolfspraul: I don't quite agree with your view, to be honest.
<wolfspraul>
looks like a convoluted way to get boldport into the game
<saardrimer>
for Milkymist, maybe
<lekernel>
saardrimer: what view? total open source?
<saardrimer>
I knew it would hard to get existing projects to convert
<saardrimer>
lekernel: no ;)
<wolfspraul>
you worked for Xilinx before, right? so you must have a 'typical' Xilinx customer in mind who would think that boldport is great
<wolfspraul>
I have no idea about the typical xilinx or even fpga customer
<saardrimer>
the view that using boldport makes you dependant on it
<saardrimer>
wolfspraul: yeah, I've worked for Xilinx for a few years
<wolfspraul>
saardrimer: from a business/investor standpoint, I can tell you that many people would agree with the line that goes "you cannot make money with tools"
<saardrimer>
wolfspraul: yes, that might be true. Let's see... I'm giving it a shot.
<wolfspraul>
it was proven wrong a few times, but overall I can see the logic why tools is extremely hard
<wolfspraul>
there was this horrible Windows company many years ago - InstallShield
<wolfspraul>
everything about it was crap
<wolfspraul>
but they made money like crazy
<wolfspraul>
it was just a small niche in tools that many people didn't want to touch, so InstallShield gained a reputation for taking this nasty problem from developers' plates
<wolfspraul>
maybe boldport has a chance along those lines - maybe lots of people hate this problem you are trying to solve and will happily outsource it to you? don't know
<saardrimer>
there are parallels to boldport there, except that I don't plan on it sucking.
<wolfspraul>
well, InstallShield worked, if you remember or even know it :-)
<wolfspraul>
that was a very successful 'tools' business, at least
<saardrimer>
yeah, I know it
<wolfspraul>
they sold the company for a few hundred million USD, I think
<wolfspraul>
good timing too :-)
<saardrimer>
timing is important.
<lekernel>
reminds me of this thawte crap which gave us ubuntu ... lol
<saardrimer>
boldport is something I wanted to build based on my view of a problem people have that needs solving.
<wolfspraul>
wrong, 76 million USD, but still
<wolfspraul>
man, Macrovision bought InstallShield for 76 million USD in 2004
<wolfspraul>
how stupid are those Macrovision managers?
<saardrimer>
so it's build, now I want to figure out how to make it successful.
<wolfspraul>
in 2008, they handed it over to some private equity firm to get rid of it
<wolfspraul>
argh
<wolfspraul>
:-)
<wolfspraul>
anyway sorry I didn't mean to trashtalk boldport
<wolfspraul>
InstallShield was just the memory that popped into my brain
<wolfspraul>
in a good way, since they made 76 million USD in the end (and probably more before since it was very profitable)
<wolfspraul>
you are probably solving some problem, although I wouldn't know which one :-)
<wolfspraul>
tell Werner (one of our guys here) that he has to click through a web interface (!) to generate a Makefile
<wolfspraul>
the good guy may finally get his heart attack
<wolfspraul>
that's blasphemy!
<wolfspraul>
:-)
<saardrimer>
trashtalk away... getting NEGATIVE feedback is that hardest thing for me right now!
<saardrimer>
well, trashtalk to me, not to others ;)
<lekernel>
on the other hand:
<lekernel>
1) the ISE project navigator is one big piece of crapware you are replacing with something most probably better
<lekernel>
2) normal FPGA people do not do free software and would be less reluctant to use that web interface/SaaS thingy
<lekernel>
so you could still make it successful
<saardrimer>
:) I hope so
<saardrimer>
like I said, the web-interface will be optional at some point. you'll be able to have a json file for the project state and that will give you a "build environment"
<saardrimer>
through an API, or...
<lekernel>
yeah but it's still SaaS
<saardrimer>
if you pay for an offline version.
<saardrimer>
yes, it is, not not how it's used most often.
<saardrimer>
but not
<saardrimer>
once you've generated a Makefile you don't need boldport again
<saardrimer>
if you don't want to
<saardrimer>
wolfspraul: who are you IRL?
<wolfspraul>
IRL?
<saardrimer>
in real life
<wolfspraul>
don't understand your question
<wolfspraul>
I'm real, no bot :-)
<wolfspraul>
I'm manufacturing Milkymist One
<saardrimer>
ok, what should I put in the google textbox in order to find out abut you?
<saardrimer>
more about you
<wolfspraul>
let me see what comes up there
<wolfspraul>
I never care :-)
<wolfspraul>
he, all sorts of things. well, my name is quite unique, when you see 'wolfgang spraul', it's probably me
<wolfspraul>
I am manufacturing hardware the last few years
<wolfspraul>
so I work on things like sourcing, production testing, hardware quality (iqc/oqc), logistics, sales
<saardrimer>
where are you based?
<Fallenou>
you want to kill him or what ? :p
<wolfspraul>
Beijing
<Fallenou>
name and adderss !
<saardrimer>
:)
<Fallenou>
gps location
<Fallenou>
launching missile
<saardrimer>
nah, that's as far as I take it
<saardrimer>
we had a decent chat, and it would be nice to know who I was chatting with
<saardrimer>
if they are willing to say
<Fallenou>
sure
<Fallenou>
was just joking around
<wolfspraul>
you can google me, after some years in technology nowadays we all leave so many traces, it's foolish to try to erase/hide them, and why even
<wolfspraul>
plus I'm in Beijing China, you learn the importance of your middle finger here
<wolfspraul>
send whatever you want my way, I deal with it. I do it anyway every day.
<saardrimer>
I'm mostly asking to avoid embarrassment, to be fair, so to avoid having conversations with two entities and then finding out that they are the same person.
<wolfspraul>
that'd be funny, but rather unlikely, no?
<wolfspraul>
7 billion people out there...
<saardrimer>
I'm more talking about twitter/blog/IRC intersection
<saardrimer>
and meeting people in real life
<wolfspraul>
that's why I use (almost) my real name everywhere now, wolfgang_spraul is too long so I use 'wolfspraul' as my nick, for example
<saardrimer>
me too
<wolfspraul>
so I don't have to tell people I'm abc here, xyz there, etc.
<wolfspraul>
yep, we seem to have the same system :-)
<saardrimer>
back to boldport-ish stuff. If you can set aside what you think of the "boldport flow"... I'm interested in feedback on my FPGA project structure proposal here: http://www.boldport.com/docs/fpgaproj
<saardrimer>
as developers of a decently sized FPGA project, your feedback is quite valuable!
<wolfspraul>
not mine, as I have zero experience with this
<wolfspraul>
I'm just the manufacturing guy
<wolfspraul>
and before that I did software development, kernel and above, never touched HDLs
<kristianpaul>
i give up no more clicks on web interfaces..
<kristianpaul>
sorry saardrimer
<kristianpaul>
may be 3 years i could done it :-)
<saardrimer>
kristianpaul: not sure I fully parse that, but I get the gist of it ;)
<wpwrak>
bah, my motto is "one Makefile to rule them all". which evil overlord would accept a web site as his regent, because he's too feeble of mind to write his own Makefile ? :)
<kristianpaul>
also you can be xyz there and not link it to abc :)
<lekernel_>
lol, opencores is spamming again for this openrisc asic crap-project
<saardrimer>
lekernel_: why do you think it's crap?
<lekernel_>
it was already discussed here and there are lots of reasons...
<lekernel_>
first they won't pull it off technically
<lekernel_>
then even from people who did
<lekernel_>
then even people who did pull it off (e.g. OpenSPARC and LEON, who _did_ manufacture free ASICs, regardless the lies you can find on opencores) did not lead the announced revolution
<lekernel_>
the process is also super obscure, the money goes directly to orsoc ab without any control
<lekernel_>
they claim they are representing the "opencores community" when in fact it's only 2 or 3 people from orsoc who decided this thing, without any prior discussion with any opencores members
<lekernel_>
how did they compute the total amount of money they needed?
<saardrimer>
the fact that they are trying shuold be held against them, though
<lekernel>
ah, they added new Q/A entries since last time I checked
<lekernel>
anyway
<lekernel>
this thing isn't even worth being looked at
<saardrimer>
I'm not a huge opencores fan... not trying to defend them, but the treatment should be fair
<saardrimer>
oh, I mean *shouldn't* be held against them
<saardrimer>
they would have had better luck on kickstarter but maybe that didn't quite fit for their plans
<lekernel>
why are they lying? why ignore opensparc and leon, who got further than them?
<lekernel>
"THIS IS UNIQUE AND HAS NEVER BEEN DONE BEFORE IN HISTORY, SO PLEASE JOIN US." hahahah
<lekernel>
also, if you have ever tried to use opencores ip, you could probably agree that claiming they will ship http://cdn.opencores.org/img/asic_block.png in any reasonable time is like claiming we are going to ship a competitor to the Boeing 747 by the next year
<saardrimer>
"Design reuse, therefore, can be difficult and frustrating, diminishing its appeal2. (For the experience, I suggest that the reader integrate a few random cores from OpenCores.org and a couple from an FPGA vendor.) "
<lekernel>
the DRAM controller they are planning to use is already crippled with bugs when targeting SDR SDRAM on a FPGA
<lekernel>
how can they possibly think about DDR2 in ASIC?
<saardrimer>
"shipped to the OpenCores donors by Q1 2012" made me chuckle.
<lekernel>
lol
<lekernel>
yeah well
<lekernel>
"crap-project" sums it up pretty well
<saardrimer>
Well, objectively, I think that main issue for me is that the target is too high, and plan B mostly benefits them (even though they can spin it that it benefits "the community")
<saardrimer>
my main issue with opencores, though, is that there is near zero domain-specific quality control.
<lekernel>
s/domain-specific// s/control//
<lekernel>
i never used anything from there which really worked
<saardrimer>
there's absolutely no reason to use an antiquated SVN host when we have services that are orders of magnitude better.
<saardrimer>
lekernel: :)
<lekernel>
yeah, that too
<lekernel>
and the login wall
<saardrimer>
yup
<lekernel>
to their credit, the openrisc core appears to be somewhat working, though it's very slow and bloated
<saardrimer>
that's because that is what they make money from
<lekernel>
but besides that... there's plasma and aemb, which are remotely serious
<lekernel>
both being experimental and relatively small hobby projects, though.
<lekernel>
all the rest just fucking doesn't work
<lekernel>
btw you can also have a look at http://www.ohwr.org/ which has more serious stuff posted to it, and also doesn't have login wall etc.
<saardrimer>
how is this different than hosting a project on github (or equivalent)?
<saardrimer>
or, rather, how is it better?
<lekernel>
i'd say the main benefit is to be visible as an open source hardware project. github has tons of stuff posted to it, and it's damn hard to find anything on it.
<lekernel>
in fact, on github, it's mostly web-related stuff that get direct exposure