NVIDIA shows signs ... [2008 - 2017]

Discussion in 'Graphics and Semiconductor Industry' started by Geo, Jul 2, 2008.

Tags:
Thread Status:
Not open for further replies.
  1. Squilliam

    Squilliam Beyond3d isn't defined yet
    Veteran

    Joined:
    Jan 11, 2008
    Messages:
    3,495
    Likes Received:
    113
    Location:
    New Zealand
    Since when? They don't share the same architecture. You wouldn't say that Sandy Bridge is the same generation of CPU as Donkey chip even though some of them will be produced on the same process node, would you?
     
  2. Sontin

    Banned

    Joined:
    Dec 9, 2009
    Messages:
    399
    Likes Received:
    0
    Juniper has no DP, too. So it is not the same architecture. There is no other real difference between g92 and gt200.
     
  3. Squilliam

    Squilliam Beyond3d isn't defined yet
    Veteran

    Joined:
    Jan 11, 2008
    Messages:
    3,495
    Likes Received:
    113
    Location:
    New Zealand
    No real difference? Cypress and Juniper share the same baseline architecture and aside from the DP support they are almost exactly the same chips aside from the fact that one is twice the size of the other. Juniper is to Cypress what the G94 is to the G92.

    On the other hand GT200 and the G92 are different architectures. G92 has its basis in the G80 whilst the GT200 architecture is something newer again.
     
  4. Sontin

    Banned

    Joined:
    Dec 9, 2009
    Messages:
    399
    Likes Received:
    0
    But who cares?
    Why don't we compare the other specifications like TMUs, ROPs, shaders and memory interface?
    5870 mobile vs 5870 desktop:
    40 TMUs <> 80 TMUs = 50% of Cypress
    16 ROPs <> 32 ROPs = 50% of Cypress
    128bit <> 256bit = 50% of Cypress
    800 shader <> 1600 shaders = 50% of Cypress

    GTX280m (g92) vs GTX280:
    64 TMUs <> 80 TMUs = 80% of GT200
    16 ROPs <> 32 ROPs = 50% of GT200
    256bit <> 512bit = 50% of GT200
    128 shader <> 240 shaders = 53,3% of GT200

    Nobody cares about minor changes like threads, better mul, Cuda 2.3 (or so) or three sm per cluster instead of two. The basic architecture is the same: Vec8 ALU, 8 TMUs per Cluster, D3D10, 4 ROPs per partition. I don't see a significant difference between AMD and nVidia.
     
  5. digitalwanderer

    digitalwanderer Dangerously Mirthful
    Legend

    Joined:
    Feb 19, 2002
    Messages:
    17,275
    Likes Received:
    1,788
    Location:
    Winfield, IN USA
    I used to think that, but my all my nVidia friends have been explaining to me that DP is much more important now. :yep2:




    ( ;) I keed, you really think I have friends at nVidia? :razz: )
     
  6. neliz

    neliz GIGABYTE Man
    Veteran

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

    Florin Merrily dodgy
    Veteran Subscriber

    Joined:
    Aug 27, 2003
    Messages:
    1,648
    Likes Received:
    219
    Location:
    The colonies
    Well, they made enough noise about the fact that they're not allowed to use QPI or DMI, so is anyone surprised?

    And hey Pineview is cursed with a GMA chip on board, it's not as if it doesn't need all the help it can get.
     
  8. Squilliam

    Squilliam Beyond3d isn't defined yet
    Veteran

    Joined:
    Jan 11, 2008
    Messages:
    3,495
    Likes Received:
    113
    Location:
    New Zealand
    Nobody cares about improved power management in an application where its most important? Nobody cares about programmability when its becoming extremely useful for encoding/decoding as well as photoshop etc. Nobody cares about the differences in the shader architecture of course not.

    The major difference between the two is that the AMD chips are from the same generation and share a common feature set. The Nvidia chips are not from the same generation no matter how similar some of their characteristics are. Furthermore the AMD chips follow a consistant naming scheme whereas the Nvidia parts do not. x8xx is always high end, x6xx is always enthusiast and so on.
     
  9. Squilliam

    Squilliam Beyond3d isn't defined yet
    Veteran

    Joined:
    Jan 11, 2008
    Messages:
    3,495
    Likes Received:
    113
    Location:
    New Zealand
    I have to say that doesn't put a very positive light on it there sir at least as a mobile chipset. As a light weight desktop chipset we'll have to see what AMD comes up with in competition, but I wouldn't suspect that its 100% rosy there either as they can get both the GPU boys and the CPU boys actually co-operating. My overall suspicion is that both Nvidia and Intel would have done a lot better here had they actually worked together.
     
  10. stevem

    Regular

    Joined:
    Feb 11, 2002
    Messages:
    632
    Likes Received:
    3
    Naturally, but that's the whole point. Intel just doesn't see a role for other IHV chipset/GPU logic for these parts.
     
  11. Sontin

    Banned

    Joined:
    Dec 9, 2009
    Messages:
    399
    Likes Received:
    0
    So can you tell me the difference in the feature set between g92 and gt200?

    G92 is the mainstream part of gt200. You have no poblem that amd will sell juniper as 5800 mobile but you complain about the same fact when nvidia is doing this?
     
  12. Arun

    Arun Unknown.
    Moderator Legend Veteran

    Joined:
    Aug 28, 2002
    Messages:
    5,023
    Likes Received:
    299
    Location:
    UK
    He is correct. Of course, he's AFAIK also wrong in seemingly assuming that there isn't a MCP7A replacement for the non-Atom market on the roadmap. I'll certainly be curious to see how they brand that one; I fear yet another insanity, but we'll see. The product itself should be interesting, though... (assuming it doesn't come out too late to be relevant)
     
  13. Jawed

    Legend

    Joined:
    Oct 2, 2004
    Messages:
    10,873
    Likes Received:
    767
    Location:
    London
    Why?

    Jawed
     
  14. neliz

    neliz GIGABYTE Man
    Veteran

    Joined:
    Mar 30, 2005
    Messages:
    4,904
    Likes Received:
    23
    Location:
    In the know
    Ultra-short lifespan? I'm pretty sure they would like to see Fermi parts there instead of the not-so-stellar gt21x.
     
  15. ShaidarHaran

    ShaidarHaran hardware monkey
    Veteran

    Joined:
    Mar 31, 2007
    Messages:
    3,984
    Likes Received:
    34
    G92 launched in October 2007.
    GT200 launched in June 2008.

    NV ALWAYS launches the high-end derivative of a product family first.

    You are wrong.
     
  16. Sontin

    Banned

    Joined:
    Dec 9, 2009
    Messages:
    399
    Likes Received:
    0
    And which was the performance part? Oh wait: They only selling GT200 cards since june 2008. :roll:
    g92b came in june. At the same time nVidia launched the GT200 cards.
     
    #1956 Sontin, Jan 1, 2010
    Last edited by a moderator: Jan 1, 2010
  17. Groo The Wanderer

    Regular

    Joined:
    Jan 23, 2007
    Messages:
    334
    Likes Received:
    2
    So is the rest of the industry. Yes, I know about the Google tablet, yes I know about Unreal engine, and yes, I know that both of those are also in bed with many others.

    How many design wins did NV claim for Tegra? How many materialized? How many sold more than 1K units? Tegra2 has the industry buzz, along with Qualcomm for ARM based widgets, but for real products to come out, meaning actual sales, they have to get their power use under control.

    Good luck there guys, the phone (and related widgets) guys don't believe spec sheets any more, they test. NV... umm... is.... err.... better at the paper side.

    -Charlie
     
  18. Groo The Wanderer

    Regular

    Joined:
    Jan 23, 2007
    Messages:
    334
    Likes Received:
    2
    Why did they announce the mobile line on the Tuesday after Christmas, and a week and a few days before their big press conference at the largest trade show in the world for consumer electronics? I did beat them by a day, but it was purely coincidental.

    They are in reactionary mode, or their marketing people should be taken out back and shot on grounds of improving humanity's average IQ.

    -Charlie
     
  19. Squilliam

    Squilliam Beyond3d isn't defined yet
    Veteran

    Joined:
    Jan 11, 2008
    Messages:
    3,495
    Likes Received:
    113
    Location:
    New Zealand
    G92 is the mainstream part of the G80. GT200 doesn't have much in the way of mainstream parts as it appears they were cancelled. It would be the same thing if mobile 5870 was simply the RV770 shrunk to 40nm and called the mobile 5870. Then it would be a last generation product posing as a current generation product and both would be doing the same thing.
     
  20. FrameBuffer

    Banned

    Joined:
    Aug 7, 2005
    Messages:
    499
    Likes Received:
    3

    so I guess that would mean that the GTX 285/275/265...etc are simply renamed 8800 GTX/GTS/GT (same generation.. so a GTX 295 = 9800GX2), ahh ok thanks for clearing that up

    Also interesting that the Fermi based Tesla (2050/2070) are not the same architecture as the Geforce GT300 series, being that the geforce lacks ECC. I guess that would also mean that the G80 and G92 are not the same architecture as the G80 lacks PureVideo VP2 and Compute 1.1 that the G92 has.. but wait... so they are all the same "generation" but not of the same "architecture"..

    Damn I'm soo confused now..
     
Loading...
Thread Status:
Not open for further replies.

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...