Beiträge von Anthony

    Very interesting, just finished it. Though I must say, your card is one of those "very different" projects he speaks of, and I doubt you'd want to sell via a distribution chain given the low volume of cards you're likely going to produce. That's especially true when giving chip availability a quick thought; VSA-100's aren't going to be available at a significant number I suppose, and they probably aren't right now? Thus, you're not even remotely going to come close to a batch of a 1000 units, just my assumption of course.

    I allow myself to further assume that it just comes down to bill of materials, transport logistics (very smart of him to mention that), manufacturing, your own hourly wages (R&D, testing, administration, support) as well as failure rates & RMA. Taxes I suppose, if it's going to be an official business (at wich point I'd be a bit wary of nVidia shitting on your cake, so branding should be considered wisely).

    Naturally, if you do everything yourself, you'd need to invest a LOT more time into the whole thing (and it'd be mostly mundane and boring tasks), but given the supposedly low volume we're dealing with here, it'd probably be fine to not rely on additional parties? No?

    *Mind you, I've never been into this business, so I may be overlooking quite a few things here.

    you're right. some things cost much less for me, some much more, especially if we're talking about cost in man-hours. but a main idea in this video is obvious, even a batch with 1000pcs price should be 2.5x more. less the number more the multiplier. if i'll ever go to a sale with such an expensive card i'll have to pay taxes from every sale, overwise they will notice my transactions sooner or later and i'll get seriuos fines.

    btw. even such a huge company as apple is close to that 2.5x - 3 multiplier

    no, it is calculating in a different manner. main point is: price must be at least 2.5 times more than a production cost. i can elaborate if you interested.

    what i have to say. at this time i have only 2 assembled boards. both of them performs quite well on my opinion. at this week i'll send one board to a independed tester. maybe he'll say that board is trash, so nothing to talk about no more. let's wait a little.

    Just a thought - have you considered designing it as a PCI Express x1 card? Is that possible with another bridge chip? Then you eliminate the bandwith bottleneck legacy PCI and there are plenty Platforms out there you could use the card with the latest XP32 drivers (socket 754/939/AM2/AM3, 775 11xx and so on). And some of them can also be used with good old Win9x :)

    Somehow I 'd really like to see a PCIe Voodoo xD

    problem is: 6152 or hint bridge is hardcoded in drivers. only possible solution is to make uncouth construction as bridge behind bridge. or drivers patching required

    I'm not sure I'm with you on that. There are some pretty taxing Glide games out there as well, such as Descent³ or Freespace 2 (the nebula missions!). I doubt all games would pre-load the entire set of textures for a level (or area) into VRAM? It's more likely to happen while playing, no? There's also other things that need to move from system RAM to VRAM all the time, like geometry data?

    Voodoo 5 cards have been known to be quite dependent on bus bandwidth, so maybe that concern is valid. If I were to buy one of your 6000 remakes, I'd rather put it into a 66MHz slot on its own, dedicated PCI bus. :)

    But you said the final design will have the 66MHz notch anyway, right? So no problem.

    q3 with max textures is a good test imho, i don't see any significant difference on bus speed. but if you have something else, pls apply some numbers, i'll accept.

    anyway, pls consider my existing cards as a prototypes. new features could be implemented in future.

    i have a good example how original v56k card perfoms on a 33mhz bus (via famous agp to pci adapter). take a look: https://www.modlabs.net/articles/novaj…agp-glava-2/p/5

    you got it

    Maybe you can redesign this part of your card to make it possible that the card can fit in this kind of PCI-X slot.

    If not, I think it will lose a lot of possible performance.

    that's correct. i should make goldinger for these slots. but if you prefer glide only games with such card, performance will be the same on both 33 and 66mhz buses, since textures resides in onborad memory.

    only 66mhz bus is possible

    this card drains 5a from 12v line

    never saw killed bios chip but with all the rest alive. bios on bricked card always can be flashed as usual if you make a boot with other vga card

    All good man you do you, no one is forcing you on what to do or what not to do, personally I prefer original cards as do many others I know and are friended to for many odd years, but hey if you can make people happy with your work that is a nice achievement and I am sure there are some people out there that will gladly be interested to buy these cards when you think they are ready to be sold :)

    my cards have their own advantages. moreover, they are assembled from only new or nos components and i'm offering 12 month warranty.

    i'm thinking on how to produce considerable amount of these cards, but i don't want to invest much in production. so, most likely, i'll keep slowly make assembly by myself, about 1 card in week or two. until i get bored. my interest in making something new, not to do the robot job, pls accept.

    maybe i'll inqure pcb manufacturer on partially assembled cards. i'm not happy to assemble everything by myself. so final price mainly depends on organization of assembly process.

    don't understand what 100/133mhz pci slot you mean. it able to operate on 100 mhz pci-ex via pci to pci-ex adapter. it can work on 66mhz pci slot too.

    all address lines are routed, so 64mb per chip configuration is possible

    just wondering why you may need a socketed bios chip?

    no 320 chips available no more, someone killed all stock of these, sorry

    A digital image is not created the same way as a RAMDAC does it for an analog image.

    If the chips have no dedicated framebuffer where the entire image is stored with proper synchronization, you'll have to build something yourself and that's maybe not even possible. Thinking about the amount of effort that went into this card I'm pretty sure Anthony thought this through properly and came to the conclusion that it's not really feasible at this point.

    I'm really looking forward to see this masterpiece do its magic.

    maybe with some sort of fancy fpga it is possible to combine two streams of digital rgb signals. anyway a LOT of efforts required to make some changes to any existing 3dfx project