How do I develop, design, and manufacture a PCI video card from absolute scratch?

Sharky Forums


Results 1 to 11 of 11

Thread: How do I develop, design, and manufacture a PCI video card from absolute scratch?

  1. #1
    Expensive Sushi
    Join Date
    Mar 2013
    Posts
    13

    Lightbulb How do I develop, design, and manufacture a PCI video card from absolute scratch?

    I thought about the specs for this light GPU:

    1.Has 58,000 colors.

    2.Has firmware on the die to keep track of colors, clock speed, and RAMDAC data, and memory bandwidth.

    3.Has 14,000 transistors.

    4.Has to be accessible directly from chipset, and with minimal device driver software support and also with none from CPU address space.

    5.Will not support pixel shaders, vertex processing, 3-D graphics, floating point calculations, or DX support.

    Now, here are my questions:

    1.How much money should this cost me?

    2.How hard will this be?

    3.How long will this take me all by myself?

    4.I can't afford specialized machinery, so I'll have to microfabricate and work on the chip with few/none of any specialized machinery.

    I am serious, and serious answers only.

  2. #2

  3. #3
    LOLWUT ImaNihilist's Avatar
    Join Date
    Nov 2001
    Location
    San Francisco
    Posts
    14,034
    I'm going to say that this is not possible for a single person to do. Maybe with a lifetime of knowledge. Maybe.

    Modern computer systems are so complex it's hard to imagine that it's possible to do this from scratch. Even if you managed to figure out the hardware, you'd still have the software layer to deal with.

  4. #4
    Expensive Sushi
    Join Date
    Mar 2013
    Posts
    13
    Not really. I mean I at least don't think a life time of knowledge is essential to tinker with hardware, and build your own.

    Tech colleges, homebrewing self employers, general computer programmers, curious geniuses, engineers, and hobbyists all alike, professional or not, can and do have the capability of testing, building, configuring, and deploying their own hardware.

    Although, it's one big stretch to say from scratch entirely myself, I'd assume the necessary machinery, engineering and hardware programming languages, electronics knowledge, and programming is definitely feasible for one person to master within a life time.

    There's many Field Programmable Gate Arrays that can be tinkered with for whatever one desires. I have came across some who know how to solder semiconductor circuitry and electrical components, and even tutorials on the internet consist of much of the needed instructions.

    And to bonus all of this, Wikipedia and other articles alike give the description, information, and general how and why to a very substantial point for one to start off on.

    A massive amount of computer programmers are self taught (i.e. no college, tutoring, schooling) and are great programmers.

    I can't fathom the idea that anyone would think this is impossible. It's actually, even if blunt to say, VERY possible, given the available knowledge, dedication, time and tools.

    However, I may only have two of the latter to a great level, but the other two are out of grasp at this time due to constaints and cash.

  5. #5
    Great White Shark
    Join Date
    Nov 2000
    Location
    Alpharetta, Denial, Only certain songs.
    Posts
    9,925
    For what it's worth, I'm not sure you can make much with only 14K transistors anymore. Even with a FPGA, several million transistors are used just for cache memory alone. Designing something with only 14K that could drive a display of any reasonable size seems daunting to say the least. That being said, I believe internal design of a modern CPU/GPU/IC is what Ima is talking about. As far as manufacturing your own board using "off the shelf" components, it should be feasible. If you aren't designing the actual chips, and instead just designing the interconnects and firmware to run the solution, it simplifies things by several orders of magnitude.

    Questions to consider:
    1. What kind of display are you driving?
    2. Using what kind of connection?
    3. What kind of refresh rate are you looking for?
    4. Are you actually driving a display, or directly driving something like an LED array?


    Realistically, assuming that you have abundant free time to spend on the project, and several prototype FPGA's, and parts, this should take you less than 2-3 years to construct if you plan on a very specific application. If you keep things within reason, say targeting a display of 16x16 pixels initially, you should be well within your abilities to succeed. This is assuming you are starting with zero background knowledge in hardware design, firmware programming, FPGA programming, etc.

    *Edit: This seems as good as any for a beginners tutorial from Altera. If that document doesn't make sense, you might need to dig deeper into some of the prerequisites that they list.
    Last edited by James; 04-09-2013 at 09:55 AM.

    Crusader for the 64-bit Era.
    New Rule: 2GB per core, minimum.

    Intel i7-9700K | Asrock Z390 Phantom Gaming ITX | Samsung 970 Evo 2TB SSD
    64GB DDR4-2666 Samsung | EVGA RTX 2070 Black edition
    Fractal Arc Midi |Seasonic X650 PSU | Klipsch ProMedia 5.1 Ultra | Windows 10 Pro x64

  6. #6
    Expensive Sushi
    Join Date
    Mar 2013
    Posts
    13
    No I want to microfabricate and create the integrared circuit from absolute scratch that I plan to make in to a compatible PCSplay device. NOT BAD WITH HARDWARE, JUST NO SKILLED PERSON MUCH EITHER.'9

  7. #7
    Hammerhead Shark
    Join Date
    Oct 2000
    Posts
    1,024
    Quote Originally Posted by Big Tunda View Post
    No I want to microfabricate and create the integrared circuit from absolute scratch that I plan to make in to a compatible PCSplay device. NOT BAD WITH HARDWARE, JUST NO SKILLED PERSON MUCH EITHER.'9
    at what production volume? this would take incredibly deep pockets even if you leased time from an established foundry, or a build a low-volume starter fab with refurbished capital equipment, plus all the engineering resources it would take to bring an idea to a working prototype

  8. #8
    Expensive Sushi
    Join Date
    Mar 2013
    Posts
    13
    I just need enough silicon for one wafer for one chip. The rest can be workd around, byt I will still need it to have an electronic micro engineering, but I don't know the steps of the machinery and production.

  9. #9
    Hammerhead Shark
    Join Date
    Oct 2000
    Posts
    1,024
    Quote Originally Posted by Big Tunda View Post
    I just need enough silicon for one wafer for one chip. The rest can be workd around, byt I will still need it to have an electronic micro engineering, but I don't know the steps of the machinery and production.

    Chip design, open source, and DIY

  10. #10
    LOLWUT ImaNihilist's Avatar
    Join Date
    Nov 2001
    Location
    San Francisco
    Posts
    14,034
    Even the people who work at Intel and design chips probably couldn't complete this task end to end. Even if you could learn how to design, lease, and fabricate you'd still have to learn how to write the software, firmware, and driver for the OS. Maybe if we are talking about Pong.

    I would think that you can find tutorials for how to do this for systems that are less complex, like a pong machine or maybe even an NES. But anything beyond that is kind of inconceivable to me. The 8-bit Motorola 6800 had 11k transistors.

    I guess the real question is this: why do you want to do this? Just to learn how? I would start with finding a guide on how to create pong from scratch, and maybe work your way up to a M6800.

  11. #11
    Expensive Sushi
    Join Date
    Mar 2013
    Posts
    13
    What software, driver, firmware, and OS?

    I just want to develop to test a functional IC. I don't need any "drivers" or anything more than simple assembler/machine instructions implemented on the microarchitecture to test electrical circuitry and pathways, buses, units, the stack, registers, etc.

    I am not going to develop an "OS" for this, although it would be ideal. And assuming a memory map, I would expect direct memory access to be intercepted by devices within the same board(i.e., no "device driver" is really needed for such a trivial test, and are more accustomed to user-level environments, application software, and/or a kernel).

    And yes, it is just to know how, as it's vital to programming, and will assure more accuracy and completeness in all related studies of computers.

    We can easily implement a pong game using a simple 8-bit micrprocessor, some video mapped memory and a simple card, display device and simple video output connective interface, and some main memory. Maybe I went a bit too far thinking of a PCI IC video device. I'd rather just implement an 8-bit micrprocessor with dedicated on-die components, 16-bit address bus, 8 registers, a stack concept, simple microarchitecture, etc. Then I could develop a PCB for it, add main memory, and see what happens there. If only there were access to the right quality spinach for less money I'd be closer to my goals and desires.

    PS: CPUs usually come with microcode specifications, and ROM. That would be insane to get done by one person with almost no money virtually or income, connections or tools, but worth it just to say I did it.

    Thanks for the link, Jabber!

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •