The Story Behind Developing The Sifteo From An Engineer’s Perspective


The video game industry must be one of the most secretive sectors when it comes to developing the electronic hardware used in the gaming consoles. The big guys don’t want to give anything away — to the competition or to the hackers who will try to get around their security measures. But it seems Sifteo doesn’t share those secretive values. We had a great time reading about the bumpy ride for the developers bringing the gaming system from concept to market. [Micah Elizabeth Scott] wrote the guest post for Adafruit Industries. She was brought on as an engineer for the Sifteo project just after the first version of the interactive gaming cube was released. From her narrative it seems like this was the top of the big hill on the roller coaster ride for the company.

What’s seen above is one gaming cube. The system developed in [Beth’s] story puts together multiple cubes for each game. The issue at hand when she joined the company was how to put more power in the hardware and rely less heavily on a computer to which it was tethered. She discusses cost of components versus features offered, how to deliver the games to the system, and all that the team learned from studying successful consoles that came before them like the long line of Nintendo hardware. It’s a fascinating read if you’re interesting in how the sausage is made.

12 thoughts on “The Story Behind Developing The Sifteo From An Engineer’s Perspective

  1. It was an interesting read, though I find some of their choices baffling.

    At $129.95 to get started with a basic system, or $399.90 fully decked out with 12 cubes, this isn’t competing with any conventional gaming systems. I see it selling almost exclusively to people who fall in love with this particular idea of a tangible interface, have large disposable incomes, and could care less if it was slightly more expensive.

    So why did they downgrade the CPU so much? “These are significant costs, especially batteries and RAM… Every dollar you spend on the CPU turns into at least three dollars of cost to the end user.”

    I don’t buy that. First, as far as I can see, they’re using MCUs in both the old and new designs; which neither incur any separate costs for RAM (it’s integrated), nor require a larger (or addition of) external RAM.

    Second, going with a better CPU often means it’s using a smaller and more power efficient process. Plus, it can perform its tasks more quickly and then sleep most of the time, which can mostly offset any higher current draw. Unless you actually need the extra processing power, in which case you’re darn glad to have it, and won’t care if it shortens the battery life slightly. Also, the CPU is not the only big current draw – let’s not forget the LCD backlight. So doubling peak CPU capability for example won’t require a much larger or more expensive battery, if at all.

    They obviously spent a lot of time (and therefore money) making this new, minimal design work, which of course gets passed along to buyers; further eating into any advantage.

    Finally, it appears they have (or will have) an online store where you can purchase games, including from third party developers. The success of any gaming system depends on a large selection of games, and at least a few killer games; any of which may come from third parties. While the new Sifteo design is clever, clever systems are often difficult to develop on. Between that and the limited customer base, some developers may pass it over.

    1. Bunnie has stated 3x cost as well:
      “2) Margin. Everyone in the supply chain has their hand out — the distributor, the merchant, the factory; and beyond that there’s market development funds and other ‘slush money’ that has to be factored in. At the end of the day, the shelf-cost is about 3x of your parts BOM cost. This means that adding a $0.50 part on the BOM turns into a $1.50 retail price impact.”

      1. It may be a good rule of thumb, but I doubt it’s universal. Might be very close if you’re adding a new part that wasn’t there to begin with. But for example, swapping an existing $3 MCU for a $5 MCU with the same footprint, would incur neither additional PCB costs, nor assembly costs at the factory; except for losses due to defective product which should be a very low percentage anyway. If it makes development easier, that reduces costs too.

        Reading on, I thought Bunny’s next paragraph was surprisingly relevant too:

        “This is aggravated by the fact that prices are quantized into ‘magic numbers’ — i.e. $19.99, $49,99, $99.99 — that just have to be hit. You just don’t MSRP a product for $127.45 — if you’re above $99, psychologically you are binned in with the $149 or $199 products.”

        With Sifteo’s starter pack priced at $129.95, this is irony; Bunny could have been talking about Sifteo specifically. If true, Sifteo could have used that example $5 MCU instead of the $3 one, with no change in psychological impact or sales.

    2. The VM design for executing software caught my eye in particular. While clever in principle, allowing addressing of more than 64K of say graphics assets, I can’t see how it’d permit the useful application of look-up tables larger than 16K. Also the working storage of a program is severely limited by the 64K being shared not only between read-only and read-write data, but also the VM’s own requirements.

      And presumably DRM. How they’re going to get the 8-bit level developer interest on a platform that puts the developer at an arm’s length of the genuine hardware at all stages, even in production, is anyone’s guess.

      To be honest, it sounds like a design team became somehow fixated on the idea of underpowered, sub-palm size blocks that, in this way, end up with their own batteries, their own radios, and their own underpowered GPU substitutes; and then soldiered on, fixing each problem with brute effort as they came along, bending the end product’s developer interface to match. That, right there, might have worked in the eighties, back when any automatic computer capability was wondrous. However, this is 2012, the year the world’s supposed to end, and we have things such as MMUs, FPUs, sub-300nm process technology, and past 200 MHz these days.

      Heck, the GP2X worked from 2 double-A cells. It had a dual-core ARM, with one application processor (GNU/Linux capable, MMU and all) and a media core that only had a memory protection unit rather than address translation hardware. The GP2X was released in 2005, seven years ago.

  2. Big secret? The only thing stopping people from making a PS3 or x360 is the fact you have to buy bulk on embedded parts and still have to turn a profit; on top of manufacturing BGA based stuff. You don’t need a secretive silicon, anything can sign content from a bootrom too…

    The games are simple too, it just takes years to produce quality content, mostly because modeling and animation solutions(zbrush, 3ds max, blender, maya) are so poorly designed; especially with rigging and animation(it was the same workflow over two decades ago)…

    1. So…. like… all of two companies with that kind of capability?

      Let me guess… you’re a developer working on the latest/greatest/hottest game of the century for the 360 and/or PS3? Oh wait, no I was wrong, you’re a top level engineer designing the fourth console.

      You’re really missing the point of that secretive comment. The Sifteo is never going to garner the same market share as the big three, the amount of information presented is a treasure trove of technical design and engineering thought processes all in one collected place. Something you’re extremely unlikely to ever see from any major platform developer.

      1. I’m not missing the point.. There is no ‘secrets’ and the tools and resources are publicly available. I was merely pointing out the reasons, which are purely logistical…

        Let me guess, you know it all…

        BTW I’ve worked as a engineering team leader both in computer engineering and software engineering under fortune five-hundred IT companies, your assumptions are broken and desperate. Continue with your wisdom renaissance man ^^

        1. M4CGUYVER lmao no one believes that someone that has worked at a fortune 500 has a screen name like xorpunk. Team leader lmao. Dude, grow up. Only high schoolers name drop and throw out false credits like you do. Buy a life…and a new alt.

          1. I guess it’s possible he’s a kid, but it kinda reads like it’s trolling. Just hits too many of the “I have no idea what I’m doing” stereotypes.

          2. You ever seen what happens to people who criticize or are not politically correct with their real name on the net? Hint: They get fired and they have to be self-employed for the rest of their lives..

  3. I enjoyed reading the article, went to their SDK… and then was baffled.

    There is no Internet connectivity you could use in your own software, charging requires the base station. I like the small form factor and don’t mind the high price…. but the system does not work for much more than 2D games right now. Like Chris said, some developers may pass it over.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.