Next NV Infomania

Discussion in 'Pre-release GPU Speculation' started by Geo, Apr 14, 2005.

  1. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
     
  2. phenix

    Regular

    Joined:
    Feb 22, 2003
    Messages:
    620
    Likes Received:
    1
    Location:
    Cambridge, MA
    Is it impossible by nature to shrink the pad size or is it not feasable just now? What if there is a future technology and people start to fit more pads on the same die area? I am complete layman. Just asking.
     
  3. _xxx_

    Banned

    Joined:
    Aug 3, 2004
    Messages:
    5,008
    Likes Received:
    86
    Location:
    Stuttgart, Germany
  4. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,928
    Likes Received:
    230
    Location:
    Seattle, WA
    Well, considering that the wires can only come out of the chip at its edges, this limitation is a natural one. You want the wires coming out of the chip to be low-resistance (so they can't get too narrow: the smaller a wire is, the higher its resistance), and you need them to be far enough apart to prevent crosstalk, so there are physical limits to how many wires you can connect to the chip based on its size.
     
  5. Xmas

    Xmas Porous
    Veteran Subscriber

    Joined:
    Feb 6, 2002
    Messages:
    3,344
    Likes Received:
    176
    Location:
    On the path to wisdom
    phenix,
    Of course there have been (and there will be) massive improvements in packaging technology. Like going from wirebond (where pads were only at the edges of the die) to flip-chip (where the pads can cover the whole die surface). But decreasing the size of a solder bump grid is apparently much more difficult than decreasing the size of on-die elements in the nm Range :D

    Chalnoth, I think you should have a look at flip-chip packaging. It's not restricted to pads at the edges.
     
  6. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,928
    Likes Received:
    230
    Location:
    Seattle, WA
    Hrm, I thought I remembered flip chip packaging as being more of a heat transfer optimization? Well, anyway, could do some looking back...

    Edit:
    Google is nice:
    http://www.amkor.com/enablingtechnologies/FlipChip/index.cfm

    Looks like the typical size of the dots on flip chip are around 100um, so it should be relatively easy to see how that would be a limitation. Since the technology to make objects on that size is pretty darned good today, it is unlikely that the size of the dots will be shrunk much at all in the coming years.
     
  7. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
  8. _xxx_

    Banned

    Joined:
    Aug 3, 2004
    Messages:
    5,008
    Likes Received:
    86
    Location:
    Stuttgart, Germany
    Yeah, I also had a good laugh. ATI does 14x, so nV has to do more! Because more is t3h betterer!11!1!!!one1
     
  9. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
    yeah, problem is, they can't demonstrate it, they don't have the drivers etc. it's even a more paper launch than crossfire..
     
  10. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,928
    Likes Received:
    230
    Location:
    Seattle, WA
    What is? 16x AA? It's been possible to enable that via tweakers for a long while.
     
  11. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
    It's not official yet.. nV are still working out the bugs (according to them)
     
  12. HaLDoL

    Newcomer

    Joined:
    Jun 10, 2004
    Messages:
    140
    Likes Received:
    2
    Install quadro drivers and you'll see.
    So you aknowledge that crossfire was a paper launch? ;)
     
  13. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    12,058
    Likes Received:
    3,116
    Location:
    New York
    That's interesting since I can enable 16x on my GT right now yet I can't go out and buy a Crossfire motherboard or master card......
     
  14. KimB

    Legend

    Joined:
    May 28, 2002
    Messages:
    12,928
    Likes Received:
    230
    Location:
    Seattle, WA
    More likely they're waiting for the release of their next product first. There's really not a whole lot of possibility for bugs here. The only differences between 16x and the other AA modes would be related to sample position choices.
     
  15. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
    I can also buy a phase change cooling system, plug it onto an x850xt and beat both 7800GTX and 7800Ultra.

    nV does not implement it in their drivers, you need third party tools.

    I can triple buffer d3d and ogl games, can you?
     
  16. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
    Probably for the same reason as Valve is waiting with lost coast
     
  17. trinibwoy

    trinibwoy Meh
    Legend

    Joined:
    Mar 17, 2004
    Messages:
    12,058
    Likes Received:
    3,116
    Location:
    New York
    You are arguing nonsense. If it was not implemented in their drivers then how the hell am I using it. I think you mean the option is not exposed in their control panel. Point is, it's available so your analogy to crossfire does not work.

    And WTF does expensive phase change cooling, beating a 7800 or triple buffering have to do with 16xAA already being available on shipping Nvidia cards?
     
  18. Geo

    Geo Mostly Harmless
    Legend

    Joined:
    Apr 22, 2002
    Messages:
    9,116
    Likes Received:
    215
    Location:
    Uffda-land
    Re Xenos, what is the tech on the 32gb/s connection between parent and child die? Could you, for instance, design a parent/child situation where the child would be large enuf (maybe at a higher process?) to handle the 512-bit pads, with the shader cores at a smaller process maybe nestled in the middle? Or have all you accomplished is shifting the bottleneck to the communication between the parent & child (which the 32gb/s kind of suggests, otherwise it would be faster)?

    Or is that too Rube Goldberg for words? :lol:
     
  19. IbaneZ

    Regular

    Joined:
    Apr 15, 2003
    Messages:
    743
    Likes Received:
    17
    Pre-orders popping up everywhere now, and this webshop posted the product info thingy for Sparkle 7800 GTX. http://www.sbs.o.se/default.asp?artnr=7800GTXspakle

    Not sure if there's any new info though.
    I'm to tired and lazy to read it all. :lol:

    16x Aniso? Thats something i didn't know at least. :)
     
  20. _xxx_

    Banned

    Joined:
    Aug 3, 2004
    Messages:
    5,008
    Likes Received:
    86
    Location:
    Stuttgart, Germany
    I found this snippet interesting:
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...