Why did the SuperGrafx fail so miserably?

[Necromancer]

Please explain for me what exactly this "buck or two more" covers, and in what time period.

It covers the cost per unit vs. an equivalent SNES/Genny cart at the time they were produced, including all the associated overhead costs and any savings from being able to do everything in house.

1) Does that cover the cost of a larger ROM?

Yes and no.  Additional rom increased costs for everyone, so the only cost disadvantage to the HuCARD would be the cost differential from using differently sized roms (i.e. - using two 4Mb roms instead of a single 8Mb rom).

2) Does that include a mapper and/or other chips?

For larger games, it'd obviously have to include the cost of the mapper.  I'm not sure what you mean by "other chips" other than additional rom, which is covered in the first question; if you mean helper chips like those in many SNES games, they'd obviously increase cost but not necessarily any more than they did for the SNES carts that utilized them.

3) Does that include re-tooling costs to produce a new circuit board design?

Probably not.  SNES and Genny games had those same costs, as they did not use the same pcb for every game.

4) How about the costs for manufacturing line changes?

No.  See above.

And in what dollars are you measuring. Things were monetarily less expensive in the 1990's. It's not fair to quote prices without adjusting for the time period. I sure can't buy a gallon of gas for $1.50 now, so your 1$ then would be $2-3 now...

Inflation does not apply equally to all things; accounting for inflation, consoles and games today cost much less than they did back then.  Not that it matters.  The point is that HuCARDs weren't substantially more expensive to produce.

Yes they do. So even $0.10 would matter to the bottom line profit.
No company will make a more expensive product unless they can recoup the difference in sales volume.

Yet that's exactly what you're saying NEC did, though, unless you think they expected the 'coolness factor' of HuCARDs to heavily influence sales.  That's a possibility, but only if they had little faith in their system being able to sell based on capabilities and had forgotten that people didn't exactly fall all over themselves for MyCards and Bee Cards.

[Mathius]

Couldn't it be possible that NEC/Hudson had a specialized/proprietary solution to manufacture HuCards on the cheap? How much do we really know about the process they used? <---that's an actual question. Does anybody here know?

[SamIAm]

Couldn't it be possible that NEC/Hudson had a specialized/proprietary solution to manufacture HuCards on the cheap? How much do we really know about the process they used? <---that's an actual question. Does anybody here know?

The actual differences in manufacturing costs between Hucards and typical cartridges, particularly on a per-kilobyte basis and during the same time periods, is something I hope I can find in the Japanese PCE magazines.

My own speculation is that everything up to and including 512k may have been cheaper per kilobyte with Hucards because of their simplicity, but beyond that, it may have gotten more expensive because of their physical size constraints.

[TheOldMan]

Couldn't it be possible that NEC/Hudson had a specialized/proprietary solution to manufacture HuCards on the cheap?

Don't know about cheap. But from the way HuCards are constructed, it sems safe to assume that they did have a specialized manufacturing facility to produce them.

How much do we really know about the process they used?

We know they used 'glop-top' (ie, unpackaged circuits) chips. We know they used pressure welding for the chip connections; I believe they used actual gold wire, but am not positive.
This indicates to me that they used industrial robots to mount the chips and connect them to the black carrier board - whatever it may be.

I imagine it's very similar to the way IC's are manufactured today, only instead of being sealed in a plastic case, they were mounted on the board itself, and sealed.
The carrier boards were then glued onto a plastic case, which was (probably) screen printed and packaged.

My gut feeling is that a 'simple' HuCard was not expensive to produce; probably not much more in price than the actual chip would have cost to make, once the set-up costs are paid. However, it would not be cost-effective to do for only a few hundred cards; you would have to manufature 1000's to make it worth-while.
Keep in mind, however, that those are -not- eproms on the card. Those are masked roms, which had to be fabricated just like any other complex circuit. It only makes economical sense if you are producing loads of cards - and, iirc, NEC already had a chip manufaturing plant...

The cost starts to rise when you change the design of the card. Going from 256Mbit (32K) to 4Mbit (512 K) probably wasn't too bad. In fact, I would bet that the larger cards (including some dual-chip cards) were planned for at the start; that's why we have a 20 line address bus. It doesn't take much to have a robot solder 2 chips in, as opposed to only 1.

Things get much more expensive, though, when you start adding other chips in. First, you have to re-design the carrier board - and have them manufactured. It's not like a regular circuit board where you draw in an extra trace, route it, and send it off to be etched. I'm not sure what was involved there, but I am pretty sure it was a big investment to re-design the carrier board.
Then, you have to re-program the robots to handle a third chip (and a second direction of motion). Again, more cost that has to be paid somehow. Not to mention you aren't making cards (or profit) while it is being done

My belief is that NEC saw that making larger (and larger) HuCards was going to require a large investment getting their manufacturing quipment to handle it. At the time, other consoles were being produced that were designed for larger address spaces, and NEC probably couldn't justify the costs in the face of dwindling sales and/or more competition.

[I do think they should have designed the SGX with a two-sided card, though. Put it in a PCE, and it would play; put it in a SGX, and the second board kicks in, giving more space and other options. It seems to me it would have been easier to do.]

[fragmare]

I do think they should have designed the SGX with a two-sided card, though. Put it in a PCE, and it would play; put it in a SGX, and the second board kicks in, giving more space and other options. It seems to me it would have been easier to do.

Hah!  That would have been neat.

[Black Tiger]

Couldn't it be possible that NEC/Hudson had a specialized/proprietary solution to manufacture HuCards on the cheap?

Don't know about cheap. But from the way HuCards are constructed, it sems safe to assume that they did have a specialized manufacturing facility to produce them.

How much do we really know about the process they used?

We know they used 'glop-top' (ie, unpackaged circuits) chips. We know they used pressure welding for the chip connections; I believe they used actual gold wire, but am not positive.
This indicates to me that they used industrial robots to mount the chips and connect them to the black carrier board - whatever it may be.

I imagine it's very similar to the way IC's are manufactured today, only instead of being sealed in a plastic case, they were mounted on the board itself, and sealed.
The carrier boards were then glued onto a plastic case, which was (probably) screen printed and packaged.

My gut feeling is that a 'simple' HuCard was not expensive to produce; probably not much more in price than the actual chip would have cost to make, once the set-up costs are paid. However, it would not be cost-effective to do for only a few hundred cards; you would have to manufature 1000's to make it worth-while.
Keep in mind, however, that those are -not- eproms on the card. Those are masked roms, which had to be fabricated just like any other complex circuit. It only makes economical sense if you are producing loads of cards - and, iirc, NEC already had a chip manufaturing plant...

The cost starts to rise when you change the design of the card. Going from 256Mbit (32K) to 4Mbit (512 K) probably wasn't too bad. In fact, I would bet that the larger cards (including some dual-chip cards) were planned for at the start; that's why we have a 20 line address bus. It doesn't take much to have a robot solder 2 chips in, as opposed to only 1.

Things get much more expensive, though, when you start adding other chips in. First, you have to re-design the carrier board - and have them manufactured. It's not like a regular circuit board where you draw in an extra trace, route it, and send it off to be etched. I'm not sure what was involved there, but I am pretty sure it was a big investment to re-design the carrier board.
Then, you have to re-program the robots to handle a third chip (and a second direction of motion). Again, more cost that has to be paid somehow. Not to mention you aren't making cards (or profit) while it is being done

My belief is that NEC saw that making larger (and larger) HuCards was going to require a large investment getting their manufacturing quipment to handle it. At the time, other consoles were being produced that were designed for larger address spaces, and NEC probably couldn't justify the costs in the face of dwindling sales and/or more competition.

[I do think they should have designed the SGX with a two-sided card, though. Put it in a PCE, and it would play; put it in a SGX, and the second board kicks in, giving more space and other options. It seems to me it would have been easier to do.]

What I don't understand about all off the pessimism towards the theoretical possibility of HuCards being made with anything extra inside or larger than 8 megs, is that it isn't theoretical. It already happened.

Alternate board designs may cost more and taking the plunge may very well have been a deal breaking investment. But we already got SFII', the Arcade Cards and the RomRam Cards. They did setup their manufacturing equipment to handle it. Unless they paid someone else to make them all.

[TheOldMan]

What I don't understand about all off the pessimism towards the theoretical possibility of HuCards being made with anything extra inside or larger than 8 megs, is that it isn't theoretical. It already happened.

Correct. I was not discussing the theoretical possibility of it, though. I was discussing the economic viability of such a thing.

Alternate board designs may cost more and taking the plunge may very well have been a deal breaking investment. But we already got SFII', the Arcade Cards and the RomRam Cards. They did setup their manufacturing equipment to handle it.

Agreed - it did happen. The point is, after they made this move, they were in a position to see if it was a winning or losing proposition. I feel that the increased cost for the larger cards, coupled with slow sales and/or more competition, made HuCard manufacturing less profitable than it was originally.
So, NEC switched it's focus to the CD format, which was more profitable, even with a smaller user base (not everyone owned the CD attachment). That's probably why so few larger games were made. There wasn't enough profit in it.

[Necromancer]

That might be true for the Tennokkoe Bank (and Populous), but not for the Arcade Cards or SFII'; HuCARDs had been abandoned long before they came out.