The Open Graphics Project was started in October of 2004 [story] by Timothy Miller [interview] with the intention of developing a completely open graphics card that "just works" on free and open source operating systems [story]. Since that time, the project has continued to work toward their goal [story], most recently completing the schematics [pdf] for their OGD1 board [story]. The OGD1 is an FPGA-based project board with memory and video hardware on it. The board will be utilized by the Open Graphics project to develop their graphics card, however it is also being made available commercially as a generic FPGA-based prototyping board to raise funds for the project. Timothy explains, "the OGD1 was going to be built anyway as a step towards OGC [the Open Graphics Card]. We were going to build 10 and use them ourselves. When corporate backing went away, we decided to make as many as we could sell."

At this time, no price has been set for the upcoming OGD1 board, however a recent discussion on the Open Graphics mailing list looked at the issue of determining an appropriate price. Timothy stressed that the OGD1 is not a graphics card, but instead is a prototyping board that is the first step toward creating a graphics card, "it's NOT a graphics card. It's a TOOL." He noted that similar boards on the market sell for upwards of $2,000, suggesting that the OGD1 will retail for quite a bit less, with discounts provided to people who buy the card in bulk and to active Open Graphics Project developers, "OGP developers are not customers. They are beneficiaries. :)" The pricing discussion continues, with the OGD1 expected to be completed and available for purchase within 6 months.

From: Timothy Miller  [email blocked]
Date: Mon Feb 27 21:12:03 EST 2006
Subject: OGD1 pricing

I really cannot go into details, but I'm trying to get some sense for
how to price OGD1 boards.

On the one hand, we don't want to price it too high, making it hard
for developers to get it (although we can sell it cheaper to 'signed
up' developers).

On the other hand, this is also a fund-raising product.  We are
selling it so that we can raise money to do things like work more
hours on OGA and produce the ASIC.

I was talking to someone who says that products like this are often
marked up quite dramatically, like, say, a factor of 6 over their
parts costs.

If OGD1 costs $X, what should we sell it for?


We need to take into consideration not just OGP developer discounts
but also bulk discounts.  And we still need to raise a load of money.



From: Timothy Miller [email blocked]
Date: Mon Feb 27 21:28:39 EST 2006
Subject: OGD1 pricing

On 2/27/06, Timothy Miller [email blocked] wrote:
> I really cannot go into details, but I'm trying to get some sense for
> how to price OGD1 boards.
>
> On the one hand, we don't want to price it too high, making it hard
> for developers to get it (although we can sell it cheaper to 'signed
> up' developers).
>
> On the other hand, this is also a fund-raising product.  We are
> selling it so that we can raise money to do things like work more
> hours on OGA and produce the ASIC.
>
> I was talking to someone who says that products like this are often
> marked up quite dramatically, like, say, a factor of 6 over their
> parts costs.
>
> If OGD1 costs $X, what should we sell it for?
>
>
> We need to take into consideration not just OGP developer discounts
> but also bulk discounts.  And we still need to raise a load of money.
>


One other comment.  I went over and visited some profs at the EE
department today, looking to see if they would be interested in any
sort of synergy with the OGP.  One of them had an FPGA board in his
lab with some small Altera part on it.  When I asked him how much it
cost, he said, "Only a few thousand dollars."

Aside from some LEDs and some other frills, OGD1 kicks the crap out of
what they have.  Maybe this gives some perspective on it.

I'm thinking of a pricing schedule like this:

Individual Retail:  3x
Bulk of 10 or more:  2.7x
Bulk of 25 or more:  2.4x
Bulk of 50 or more:  2.2x
Bulk of 100 or more:  2x
Individual OGP developer:  2x, limit 5

We've made some adjustments to the parts list, and the total isn't in
yet, so I'm just working on some guesses here.



From: Patrick McNamara [email blocked]
Date: Mon Feb 27 21:32:36 EST 2006
Subject: OGD1 pricing

In some ways, I think it depends on what $X is.  If it's $50US worth of 
parts, selling it for $300 is a 6x markup and would probably still be 
considered really cheap.  On the other hand if $X is $250US, a 6x markup 
is $1500 which is probably more that you want to target for as it may 
put it out of reach folks you really want to have it.

Here is another way of looking at it.  Assume you need $2M US to do the 
ASIC run.  Assume also that if you can fund half the cost, you can 
attract investors for the other half.  That means you need to raise $1M 
US.  If you make $250 net profit per OGD1 board, you need to sell 4000 
boards to make your target.  Do you think you can make that target?  Do 
you want to add more in to help offset the developers cost (time, 
effort, etc)?  This is really a business proposition.  Do a bit of 
research if you can.  Do you know anyone in university engineering 
programs?  Find out what they would be willing to pay for a board of 
this type.  Also, compare it to boards already on the market with 
similar specifications.  If you can undercut them substantially, great.  
But, remember, this is a fund raiser.  The boards will be more expensive 
to begin with and yes, you will here some grumbling from some folks 
about the price.  Prices go down over time.  If someone who is going to 
be a really productive developer can't afford one, there will be people 
who will donate them.  Those that just want a cheap card to play with 
will either have to wait, or save up their hard earned money.

Also, remember that the higher the price, the higher the expectation for 
a "flawless" product will be.  We all know there are bug in hardware and 
software, but if you are flogging a $1000 development board, people will 
expect the thing to work (rightly or wrongly).  Be sure and set peoples 
expectations for what they are getting (i.e.  A raw development board 
with no image and very little provided tools).

Just my somewhat long $.02.

Patrick M



From: Timothy Miller [email blocked]
Date: Mon Feb 27 21:48:19 EST 2006
Subject: OGD1 pricing

On 2/27/06, Patrick McNamara  [email blocked] wrote:
> In some ways, I think it depends on what $X is.  If it's $50US worth of
> parts, selling it for $300 is a 6x markup and would probably still be
> considered really cheap.  On the other hand if $X is $250US, a 6x markup
> is $1500 which is probably more that you want to target for as it may
> put it out of reach folks you really want to have it.

I wasn't QUITE hitting for 6x, but it's good to know that that happens
for comparison.

> Here is another way of looking at it.  Assume you need $2M US to do the
> ASIC run.  Assume also that if you can fund half the cost, you can
> attract investors for the other half.  That means you need to raise $1M
> US.  If you make $250 net profit per OGD1 board, you need to sell 4000
> boards to make your target.  Do you think you can make that target?

Is it possible for us to sell 4000 boards?  Sure.  Do we have the
marketing ability to reach all 4000 of the right people?  That's
really iffy.

> Do
> you want to add more in to help offset the developers cost (time,
> effort, etc)?

Well, to Howard, Andy, and me, we're not really trying to offset
anything.  We're going to spend a few thousand of our own money and
call it lost.  Every penny of income earned from sales of these boards
is going to be packed away until we can justify spending any of it on
something productive.

Say we make $1 million.  If we don't have a partner at that time,
we're still stuck.  Even if we make $2, we're still stuck between
choosing to spend money on ourselves as employees versus being able to
make an ASIC.  If we make $3, we've got a surplus.

> This is really a business proposition.  Do a bit of
> research if you can.  Do you know anyone in university engineering
> programs?

Just talked to them.

> Find out what they would be willing to pay for a board of
> this type.

"A few thousand dollars" for a board that has much less logic area
(but a lot more other frills).

> Also, compare it to boards already on the market with
> similar specifications.

The most similar was over $2000, last I saw.

> If you can undercut them substantially, great.
> But, remember, this is a fund raiser.  The boards will be more expensive
> to begin with and yes, you will here some grumbling from some folks
> about the price.  Prices go down over time.  If someone who is going to
> be a really productive developer can't afford one, there will be people
> who will donate them.  Those that just want a cheap card to play with
> will either have to wait, or save up their hard earned money.

I'm thinking that, with limits, we can sell to "OGP developers" at
deep discounts.  In that case, it's not a fund raiser.  We need to
make token profit to be sure the person is serious.  And the warranty
won't be transferrable in that case.

> Also, remember that the higher the price, the higher the expectation for
> a "flawless" product will be.  We all know there are bug in hardware and
> software, but if you are flogging a $1000 development board, people will
> expect the thing to work (rightly or wrongly).  Be sure and set peoples
> expectations for what they are getting (i.e.  A raw development board
> with no image and very little provided tools).

Excellent point.  We do not have a working PCI core.  Making some
simple logic to test the board won't be TOO hard, really.  All we need
to do is check for certain kinds of signal integrity and make sure the
memories work.  But we may have boards to sell before that work is
done.  It seems logical to sell "alpha" boards at a lower price... or
to give them a longer warranty so that should they discover a
manufacturing flaw, they can get it replaced at a later date.

> Just my somewhat long $.02.

Excellent food for thought.  Thank you.



From: nico [email blocked]
Date: Tue Feb 28 03:34:41 EST 2006
Subject: OGD1 pricing

>> Also, compare it to boards already on the market with
>> similar specifications.
>
> The most similar was over $2000, last I saw.
>

It's depend what is your market target what the number of sell did you
want to do.

600$ for a graphic card is very expensive.

600$ for a (quite big) FPGA dev board is cheap.

If you market the card like a graphic card or a FPGA dev board you will
not hit the same media and you will not get the same advertising.

A part from that, 1000\u20ac is quite expensive for an individual(that the cost
of a new computer....). I think there a lot of talented developper in the
east that can't afford this.

Nicolas Boulay



From: Timothy Miller [email blocked]
Date: Tue Feb 28 07:26:39 EST 2006
Subject: OGD1 pricing

On 2/28/06,  [email blocked] wrote:
> >> Also, compare it to boards already on the market with
> >> similar specifications.
> >
> > The most similar was over $2000, last I saw.
> >
>
> It's depend what is your market target what the number of sell did you
> want to do.
>
> 600$ for a graphic card is very expensive.
>
> 600$ for a (quite big) FPGA dev board is cheap.
>
> If you market the card like a graphic card or a FPGA dev board you will
> not hit the same media and you will not get the same advertising.

Definitely don't want to market it like a graphics card.  No way. 
It's NOT a graphics card.  It's TOOL.

Although, to an extent, I'm banking on the fact that it's associated
with the OGP, which gets it more publicity, but that publicity won't
necessarily reach the FPGA customers.

>
> A part from that, 1000\u20ac is quite expensive for an individual(that the cost
> of a new computer....). I think there a lot of talented developper in the
> east that can't afford this.

Don't get caught up in the price impact for OGP developers.  They'll
get what they need.  Think of yourself as being in my position, trying
to fund a project, and you all are my employees, so you benefit from
higher sales of a higher priced product.

OGP developers are not customers.  They are beneficiaries.  :)



From: Jeff Garzik jgarzik [email blocked]
Date: Mon Feb 27 22:18:50 EST 2006
Subject: OGD1 pricing

Patrick McNamara wrote:
> In some ways, I think it depends on what $X is.  If it's $50US worth of 
> parts, selling it for $300 is a 6x markup and would probably still be 
> considered really cheap.  On the other hand if $X is $250US, a 6x markup 
> is $1500 which is probably more that you want to target for as it may 
> put it out of reach folks you really want to have it.

Actually, the cost is only vaguely relevant.  The board should cost what 
the market will bear.  If each board costs $600 to produce, but the 
market can only support $300, then $300 (or no sale) it is.  If each 
board costs $50 and the market will bear $300, then whoopee.

	Jeff



From: Timothy Miller [email blocked]
Date: Mon Feb 27 22:24:24 EST 2006
Subject: OGD1 pricing

On 2/27/06, Jeff Garzik  [email blocked] wrote:
> Patrick McNamara wrote:
> > In some ways, I think it depends on what $X is.  If it's $50US worth of
> > parts, selling it for $300 is a 6x markup and would probably still be
> > considered really cheap.  On the other hand if $X is $250US, a 6x markup
> > is $1500 which is probably more that you want to target for as it may
> > put it out of reach folks you really want to have it.
>
> Actually, the cost is only vaguely relevant.  The board should cost what
> the market will bear.  If each board costs $600 to produce, but the
> market can only support $300, then $300 (or no sale) it is.  If each
> board costs $50 and the market will bear $300, then whoopee.

The market for FPGA-based dev boards bears costs in the thousands. 
(But they probably do more testing, etc.)

But OGP makes it a different animal.



From: Jeff Garzik [email blocked]
Date: Mon Feb 27 22:30:15 EST 2006
Subject: OGD1 pricing

Timothy Miller wrote:
> On 2/27/06, Jeff Garzik  [email blocked] wrote:
> 
>>Patrick McNamara wrote:
>>
>>>In some ways, I think it depends on what $X is.  If it's $50US worth of
>>>parts, selling it for $300 is a 6x markup and would probably still be
>>>considered really cheap.  On the other hand if $X is $250US, a 6x markup
>>>is $1500 which is probably more that you want to target for as it may
>>>put it out of reach folks you really want to have it.
>>
>>Actually, the cost is only vaguely relevant.  The board should cost what
>>the market will bear.  If each board costs $600 to produce, but the
>>market can only support $300, then $300 (or no sale) it is.  If each
>>board costs $50 and the market will bear $300, then whoopee.
> 
> 
> The market for FPGA-based dev boards bears costs in the thousands. 
> (But they probably do more testing, etc.)
> 
> But OGP makes it a different animal.

Right... so what will the OGP dev board market bear?

My guess would be no more than $399...

	Jeff



From: Timothy Miller [email blocked]
Date: Mon Feb 27 22:37:24 EST 2006
Subject: OGD1 pricing

On 2/27/06, Jeff Garzik [email blocked] wrote:

> > But OGP makes it a different animal.
>
> Right... so what will the OGP dev board market bear?
>
> My guess would be no more than $399...
>

Heh.

Raw parts costs are more than that, not even including assembly.

Keep in mind that many OGP developers won't be expected to buy their
own boards.  This project will only survive based on the generosity of
those who value it.  Plus, if we profit enough from retail, we should
eventually be able to start selling them at a loss to OGP developers.

OGP developers will number, optimistically, in the hundreds.  We need
to sell thousands of these boards.  We can swallow some loss for the
OGP.

Putting aside OGP developers, who will get boards one way or another,
what should we price it at for the sake of being able to fund future
development?



From: Attila Kinali [email blocked]
Date: Tue Feb 28 05:31:50 EST 2006
Subject: OGD1 pricing

On Mon, 27 Feb 2006 22:24:24 -0500
"Timothy Miller" [email blocked] wrote:

> The market for FPGA-based dev boards bears costs in the thousands. 
> (But they probably do more testing, etc.)

Not so true. In the company i currently write my master thesis,
they have two of these FPGA boards. The newer one never worked
at all, and the older one works only in a hand full of computers
(because they missused some PCI spec openings).

And yes, both cost in the thousands.

Now to the question what you should price.. I have no idea :(
I can only say, that i would pay up to 500CHF (about 380USD)
for one board and would buy two of them, if it gets under
300CHF (about 230USD)[1]. I know that the 230USD is totaly
unrealistic as it does not even cover the parts costs, but
then again, as Jeff already said, the costs of the product
do not matter but what the market would pay.

To sumarize that, you have to be very carefull on what
market you target. If you choose the wrong one, you'll
not be able to break even. 

I would, most probably, target the FPGA development board
market. Why? There you have prices where you can cover
most of the costs. The drawback is, that you will not be
able to sell more than a few hundreds per year, as this
market is quite small. And that if and only if, you can
place OGD right, ie small costs, good support for the
OS your customers use.

After OGD became a viable (ie a self supporting) product,
then we can think about selling it cheaper to developers.

BTW: have you tried to estimate how big the various
markets are?
 

> But OGP makes it a different animal.

How do you mean that ?

				Attila Kinali

[1] Side note: i'm still a student and thus have to be
rather conservative on how i spend my money until i get
a decent job.
-- 
wer soviel schoggi isst, kann sowieso nicht dumm sein ;-)
		-- Sandra



From: Timothy Miller
Date: Tue Feb 28 07:39:19 EST 2006
Subject: OGD1 pricing

On 2/28/06, Attila Kinali [email blocked] wrote:

>
> I would, most probably, target the FPGA development board
> market. Why? There you have prices where you can cover
> most of the costs. The drawback is, that you will not be
> able to sell more than a few hundreds per year, as this
> market is quite small. And that if and only if, you can
> place OGD right, ie small costs, good support for the
> OS your customers use.

I don't NEED to sell more than a few thousand/year.  Selling a few
thousand would be huge and prove our value to a partner.

But I do want to maximize profit.  If I can reduce the price a bit and
sell way more, then that's a win.

Actually, volume pricing is what would do it.  If one sold for $1000,
but 50 sold for $650, someone would buy a batch and resell them.

Also, I have to look intelligent to a partner.  That's another factor
in determining the price.

>
> After OGD became a viable (ie a self supporting) product,
> then we can think about selling it cheaper to developers.

I'm already thinking of selling it cheaper to developers.  If it
retails for $1000, but you can get one for $600, how would you feel?

> BTW: have you tried to estimate how big the various
> markets are?

No.

>
> > But OGP makes it a different animal.
>
> How do you mean that ?

Its association with the OGP makes it harder to think of it purely in
terms of being an FPGA board.


>                                 Attila Kinali
>
> [1] Side note: i'm still a student and thus have to be
> rather conservative on how i spend my money until i get
> a decent job.

Don't worry about yourself.  You'll get one you can afford.  That is
not the concern here.

> --
> wer soviel schoggi isst, kann sowieso nicht dumm sein ;-)

Who somuch ?? eats, can ?? not dumb be?

Mein Deutsch ist sehr rusty.  :)



From: Attila Kinali [email blocked]
Date: Tue Feb 28 07:58:54 EST 2006
Subject: OGD1 pricing

On Tue, 28 Feb 2006 07:39:19 -0500
"Timothy Miller" [email blocked] wrote:


> I don't NEED to sell more than a few thousand/year.  Selling a few
> thousand would be huge and prove our value to a partner.

Then we'll need some good marketing. I don't know how many
those FPGA board manufactorers sell, but i doubt it's a lot.
On the other hand, i know at least 3 of them and i did not
even search for them. So, given that we can do a good marketing,
a good support from our side (ie fast turaround cycles for problems,
good support "hotlines" etc), then we should be able to reach
thousands a year. But that's a big if. And noone can say anything
about the volume unless we have some idea of the market size.
 
> But I do want to maximize profit.  If I can reduce the price a bit and
> sell way more, then that's a win.

Yeah, that's always the big question with pricing.
 
> Actually, volume pricing is what would do it.  If one sold for $1000,
> but 50 sold for $650, someone would buy a batch and resell them.

Good idea. Distribute the work of selling the card to retailers.
If the margin is high enough, resellers will beat each other
to be able to sell OGD.

On the other hand. Is it possible to build a distribution channel
ourselves? Saving those 20-30% of margin?

> > After OGD became a viable (ie a self supporting) product,
> > then we can think about selling it cheaper to developers.
> 
> I'm already thinking of selling it cheaper to developers.  If it
> retails for $1000, but you can get one for $600, how would you feel?

Quite good. Knowing that those 600 are just barely
over the production cost (if at all)

> > BTW: have you tried to estimate how big the various
> > markets are?
> 
> No.

First thing to do then. I'd ask some universities and
R&D labs on how many they would buy if the price is
600, 700, 800, 900, 1000USD...
 
> >
> > > But OGP makes it a different animal.
> >
> > How do you mean that ?
> 
> Its association with the OGP makes it harder to think of it purely in
> terms of being an FPGA board.

Then we have to work against that. If we want to sell
it as an FPGA development board, then those people have
to think of it as such, ignoring the analog part completely
(unless of course they want to do some analog output for
machine control or whatever)

 
> > [1] Side note: i'm still a student and thus have to be
> > rather conservative on how i spend my money until i get
> > a decent job.
> 
> Don't worry about yourself.  You'll get one you can afford.  That is
> not the concern here.

Don't worry about me. I will buy one if i have the money.
Though i'm not sure whether i should apply for a developer
price, as i cannot say whether i can really spend some
time on developing.

> > wer soviel schoggi isst, kann sowieso nicht dumm sein ;-)
> 
> Who somuch ?? eats, can ?? not dumb be?
> 
> Mein Deutsch ist sehr rusty.  :)

"Who eats so much chocolate cannot be dumb anyways."

This came up in an IRC discussion early in the morning,
so dont take it to seriously :)

			Attila Kinali

-- 
wer soviel schoggi isst, kann sowieso nicht dumm sein ;-)
		-- Sandra



From: Timothy Miller [email blocked]
Date: Tue Feb 28 08:12:51 EST 2006
Subject: OGD1 pricing

On 2/28/06, Attila Kinali [email blocked] wrote:
> On Tue, 28 Feb 2006 07:39:19 -0500
> "Timothy Miller" [email blocked] wrote:
>
>
> > I don't NEED to sell more than a few thousand/year.  Selling a few
> > thousand would be huge and prove our value to a partner.
>
> Then we'll need some good marketing. I don't know how many
> those FPGA board manufactorers sell, but i doubt it's a lot.
> On the other hand, i know at least 3 of them and i did not
> even search for them. So, given that we can do a good marketing,
> a good support from our side (ie fast turaround cycles for problems,
> good support "hotlines" etc), then we should be able to reach
> thousands a year. But that's a big if. And noone can say anything
> about the volume unless we have some idea of the market size.

So, we need to see how many people intereted in the OGP would be
willing to do pro bono sales work for Traversal.  :)

> > Actually, volume pricing is what would do it.  If one sold for $1000,
> > but 50 sold for $650, someone would buy a batch and resell them.
>
> Good idea. Distribute the work of selling the card to retailers.
> If the margin is high enough, resellers will beat each other
> to be able to sell OGD.
>
> On the other hand. Is it possible to build a distribution channel
> ourselves? Saving those 20-30% of margin?

Now, you're thinking.  We'd still do volume pricing.  But we'd find
ways to discount less often.

> > > After OGD became a viable (ie a self supporting) product,
> > > then we can think about selling it cheaper to developers.
> >
> > I'm already thinking of selling it cheaper to developers.  If it
> > retails for $1000, but you can get one for $600, how would you feel?
>
> Quite good. Knowing that those 600 are just barely
> over the production cost (if at all)

The price isn't totally determined yet.  We know the cost of the BOM,
but the assembly cost isn't settled.  And there are other things that
are hard to anticipate.

> > > BTW: have you tried to estimate how big the various
> > > markets are?
> >
> > No.
>
> First thing to do then. I'd ask some universities and
> R&D labs on how many they would buy if the price is
> 600, 700, 800, 900, 1000USD...

I need to go back over to the EE department and ask for datasheets and
price lists on the boards they use.

> > >
> > > > But OGP makes it a different animal.
> > >
> > > How do you mean that ?
> >
> > Its association with the OGP makes it harder to think of it purely in
> > terms of being an FPGA board.
>
> Then we have to work against that. If we want to sell
> it as an FPGA development board, then those people have
> to think of it as such, ignoring the analog part completely
> (unless of course they want to do some analog output for
> machine control or whatever)

You may have noticed that we are already observing the confusion that
arises because of this distinction.

Somehow, it needs to be made clear that OGD1 was going to be built
anyway as a step towards OGC.  We were going to build 10 and use them
ourselves.  When corporate backing went away, we decided to make as
many as we could sell.

>
> > > [1] Side note: i'm still a student and thus have to be
> > > rather conservative on how i spend my money until i get
> > > a decent job.
> >
> > Don't worry about yourself.  You'll get one you can afford.  That is
> > not the concern here.
>
> Don't worry about me. I will buy one if i have the money.
> Though i'm not sure whether i should apply for a developer
> price, as i cannot say whether i can really spend some
> time on developing.

Before I let anyone have developer pricing, I'm going to search the
list for participation.  If they're participated much, they qualify. 
Therefore, you qualify.  If they're new, they've got to prove their
interest in some other way.

I have to come up with some reasonable way to judge if someone's only
pretending to be interested just to get a lower price.  It won't be an
exact science, and some people will fool us into thinking they're
seriously interested in the project.  We'll survive that.  :)

• Archive of above thread