NVIDIA’s PCI-Express Graphics Family

Discussion in 'Beyond3D News' started by Dave Baumann, Feb 17, 2004.

  1. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,079
    Likes Received:
    648
    Location:
    O Canada!
    As we mentioned previously, NVIDIA have announced their initial foray into the PCI-Express bus market at IDF, with a line of boards that spans the market segments for graphics boards. The line-up consists of the GeForce PCX 5950, GeForce PCX 5750, GeForce PCX 5300 and GeForce PCX 4300.<p align="center">Click Image to Enlarge
    [​IMG] [​IMG]
    GeForce PCX 4300</p>Not entirely surprisingly, none of the boards announced today are native PCI-Express, but all utilise their previously announced PEG 16X bridge chip now labelled "HSI" for "High Speed Interconnect" - this means that although the board can operate in PCI-Express motherboards the graphics chip will still have the same inherent limitations of AGP transfer protocols, such loosing PCI-Express’s full speed bi-directional transfer, although there may be some tertiary benefits of a slightly higher sustained bandwidth to the graphics gained by new PCI-Express motherboards supporting higher FSB and memory bandwidths.

    Being these newly announced boards utilise the bridge chip they draw from the current line of AGP8X graphics chips with PCX 5950 using NV38, PCX 5750 NV36, PCX 5300 NV34 and PCX 4300 NV18. Currently its expected that the bridged solution will be utilised for the initial versions of NV40 with the refresh products heralding NVIDA’s transition to native PCI-Express.
     
  2. DerekBaker

    Newcomer

    Joined:
    Sep 19, 2003
    Messages:
    78
    Likes Received:
    0
    I was under the impression that the NV40 was going to launch just before the introduction of PCI-Express.

    If they need PCI-E versions of their existing cards, doesn't that suggest that they feel they won't have NV40 out in time for the first motherboards?

    Of course, it could be that PCI-E motherboards will be out early, but I know which idea I think is more likely. :D


    Derek
     
  3. ET

    ET
    Regular

    Joined:
    Jun 29, 2002
    Messages:
    385
    Likes Received:
    1
    Lovely naming. I expect a GeForceFX 4300 will arrive soon?
     
  4. DerekBaker

    Newcomer

    Joined:
    Sep 19, 2003
    Messages:
    78
    Likes Received:
    0
    I see that the lower derivatives of the NV40 are a little later, so that could explain most of the boards.

    Though, there must be a very big performance/price difference between the NV40 and the new PCX 5950, for the NV40 to be still on schedule.


    Derek
     
  5. keegdsb

    Newcomer

    Joined:
    Nov 10, 2002
    Messages:
    186
    Likes Received:
    0
    Is 5750 going to be upclocked from 5700U or is the 50 only because of PCI-Express?
     
  6. DerekBaker

    Newcomer

    Joined:
    Sep 19, 2003
    Messages:
    78
    Likes Received:
    0
  7. DemoCoder

    Veteran

    Joined:
    Feb 9, 2002
    Messages:
    4,733
    Likes Received:
    81
    Location:
    California
    Dave, do you really think better bidirectional transfer speed really the main value for PCIE? I would think that better bandwidth to the card would be the only immediately beneficial feature to consumers. High speed Bi-directional transfers, future scalability of the bus, etc all of these are valuable in the future, but not in the present, and certainly not within the lifetime of the bridge chip solution.
     
  8. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,079
    Likes Received:
    648
    Location:
    O Canada!
    Its not going to be an immediate benefit, but then with a bridged solution you're going to loose out on the mainstay of the bandwidth benefits.
     
  9. Sazar

    Newcomer

    Joined:
    Jun 17, 2003
    Messages:
    234
    Likes Received:
    1
    Location:
    Between Austin and Tampa
    dave do you have one of these babies in a rig @ the moment ?

    :)

    if it breaks any nda's... no need to answer either way :cool:
     
  10. Joe DeFuria

    Legend

    Joined:
    Feb 6, 2002
    Messages:
    5,994
    Likes Received:
    70
  11. Anonymous

    Veteran

    Joined:
    May 12, 1978
    Messages:
    3,263
    Likes Received:
    0
     
  12. Anonymous

    Veteran

    Joined:
    May 12, 1978
    Messages:
    3,263
    Likes Received:
    0
    Smirk! Consumer HD video editing?!? Haha.

    The the size of the *consumer* HD video editing market is ZERO. You can count the number of HD DigiCams on the market on one hand, and none of them are consumer, and barely "prosumer", unless you think a $4000 DigiCam is "prosumer". A Sony VX2000 is considered a very good, top of the line, prosumer DigiCam and it costs $2000.


    Secondly, no one uses the video card to "manipulate" video with NLE software. At best, they might use the card to do some in-RAM previewing of some special effects, but most of the plugins I use in Adobe Premiere would be very hard to translate into a pixel shader. So this is a double hypothetical: Someone using NLE software to "manipulate" non-existent consumer HD video in real time for previewing. But if you are previewing to screen, there is no need to "read back" anything to save to the hard disk. And if you are doing the "final render", you're going to do it non-realtime at full quality anyway with the CPU. Even Liquid Studio does the final render by using the CPU for everything.

    What's the hypothetical read-back issue here? That HD video editors will want to do final renders with the vidcard and not 'offline' with the CPU?


    Today, the video card is primarily used to decompress coded video in real time to reduce CPU usage, but people using NLE software work with video streamed from the harddisk and decoded by the CPU. Uploading it from harddisk to video card and back to system ram for CPU processing would be a total waste and way slower. A P4 needing to do some inmemory video munging would be way better off doing the decompression itself instead of roundtripping it to the vidcard.

    Today, the vast majority of desktop video users are using MiniDV cameras, capturing over firewire, and editing with software like Adobe Premiere, Microsoft MovieMaker, iMovie, or Final Cut Pro.

    This is such a ludicrous issue. No consumers own PCI Express motherboards, HD DigiCams are few and far between and outside the price range of consumers, and consumers are already used to real time previews done with fast CPUs using SSE or AltiVec.


    This is PR hype if I've ever seen it. ATI hype and Pinnacle hype. A dubious feature that a very slim minority of people could ever take advantage of. Perhaps in 3 or 4 years, when HD CamCorders come down in price and current consumers feel the need to edit HUGE video streams, it will be a feature worth having. Today, it is a feature that even professionals will be unlikely get too excited about. (e.g. HD content editing is not a big deal)

    I edit MiniDV streams today. They eat up 13gb of diskspace for every hour of video. An HDTV stream would eat up atleast 6 times that, or almost 80gb per hour of video. Given that a typical 1 hour production requires minimally 15 times as much video, we're talking terabytes to do "desktop" HD video editing. Chalk up another few thousand to buy a cheapass terabyte RAID array.

    (and BTW, Sun and SGI graphic unix workstations do not have the AGP read back problem, nor do unified memory systems)
     
  13. duncan36

    Newcomer

    Joined:
    Aug 14, 2002
    Messages:
    173
    Likes Received:
    0
    What exactly is your beef that ATi dares make a niche product using PCI-Express and dares to put out a press release trumping its release.
    Wow how dare they.
    No wonder you're posting as a guest you're not making a well thought out point here.
     
  14. DemoCoder

    Veteran

    Joined:
    Feb 9, 2002
    Messages:
    4,733
    Likes Received:
    81
    Location:
    California
    Well, if it isn't obvious, I posted it, just wasn't logged in by accident. I have no problem beef with ATI putting out this product. I'm just pointing out that the bridge-chip criticisms are fairly baseless. The bridge chip is a stop gap measure, and there is next to zero performance benefit to consumers of PCIE within the lifespan that the bridge chip solution is going to be used. If NVidia tried to re-spin their whole product line with native PCIE, I'd call their management stupid. It'd be a huge waste of effort. There are no consumers with PCIE motherboards, they won't be very many in the 1H 2004, and the idea of consumers editing HD Video on PCIE vidcards with PCIE motherboards is quite ludicrous for the variety of reasons I listed.

    The people who are claiming "consumer" HD video editing as the major benefit of PCIE and as a criticism against NVidia's stop-gap bridge-chip are the ones who don't have a "well thought out" argument.

    Native PCIE support will make alot more sense in Q3/Q4 or even 2005. Early support is just marketing featuritis and criticism for lack of support is baseless nitpicking.
     
  15. duncan36

    Newcomer

    Joined:
    Aug 14, 2002
    Messages:
    173
    Likes Received:
    0
    I think you will be able to make that point IF theres no difference in performance between the AGP ATi cards and the native PCI-Express cards.
    Until we see for sure I find your remarks premature.
     
  16. DemoCoder

    Veteran

    Joined:
    Feb 9, 2002
    Messages:
    4,733
    Likes Received:
    81
    Location:
    California
    Well, given that the progressive doubling of bandwidth from AGP->2x>4x->8x show rather modest, if not non-existance performance increases, I'll take my bets.

    PCIE is a long term technology. The benefits are long term. Since it will be hard for anyone to even buy a PCIE motherboard before May, it will be rather hard to test the performance of those native PCI Express parts won't it?

    This thread is about the difference between bridge-chip solutions and native PCIE cards. By the time large numbers of PCIE systems are available, there won't be any bridge chip vidcards anyway, which is precisely my point.

    I need to upgrade MB and I am desparately waiting to buy the first available PCIE MB. I don't wanna upgrade until PCIE comes out. Looks like I will have to wait until summer. That's why I find the remarks on this board with respect to PCIE rather ludicrous.
     
  17. duncan36

    Newcomer

    Joined:
    Aug 14, 2002
    Messages:
    173
    Likes Received:
    0
    Are you hoping that ATi native PCI-Express cards dont outperform their AGP counterparts so a Nvidia bridge is justified?

    We simply dont know what fruit native PC-Express is going to bring immediately. The press release ATi put out that you derided sheds some light on what that fruit might be though: "choosing a native solution over a bridge solution to enable the next wave of applications, which will need true PCI Express graphics to realize their full potential."

    Using the progression of AGP as a reason that PCI-Express wont have any performance gains is a weak argument given the major limitations of AGP that everyone is aware of.

    As far as dates are concerned I would imagine that ATi or Nvidia wont release their PCI-Express GPUs until PCI-Express motherboards are released what would be the point to release them prior to this? There is none.

    Still we'll have to wait for benchmarks to be sure of anything. I'm extremely optimistic for a nice boost from native PCI-Express however.
     
  18. DemoCoder

    Veteran

    Joined:
    Feb 9, 2002
    Messages:
    4,733
    Likes Received:
    81
    Location:
    California
    I'm not "hoping" anything, since I'll be buying the cheapest best performing PCI Express card, and I simply don't care how it's implemented.

    It is you and others who are deriding NVidia without evidence and without waiting for the facts. Given your comments in other threads, this is unsurpising.

    I'm just pointing out that a AGP bridge is a perfectly legitimate design decision to enable NVidia's lowend-to-midrange NV3x cards to work on PCI-E motherboards that don't have an AGP slot (e.g. upcoming SiS MB's)
    It is also perfectly legitimate for NVidia to ship their initial NV40 with a bridge, and release a refresh later with native PCIE if required. Criticizing them for not doing a costly refresh of their entire product line with native PCIE before PCIE even exists as a market reality is a rather weak criticism, and I suspect it has more to do with your bias than with reality.

    There are no facts available on PCIE performance, either native or bridge, so premature criticism of the bridge is the logical error. Your reaction and others is to fall back and claim some dubious "HD video editing benefit" that 99.9% of consumers won't see.
     
  19. Joe DeFuria

    Legend

    Joined:
    Feb 6, 2002
    Messages:
    5,994
    Likes Received:
    70
    No one said it isn't legitimate. It's just not as optimal as ATI's.

    It's not so much criticizing nVidia...it's praising ATI for implementing a native implementation from the get-go...which is also a perfectly legitimate design decision.

    Damn....I must fall into that 0.1% then...
     
  20. ET

    ET
    Regular

    Joined:
    Jun 29, 2002
    Messages:
    385
    Likes Received:
    1
    Thanks. Got to love:

    NVIDIA GeForce PCX 4300 - provides entry-level pricing coupled with strong performance, unbeatable visual quality, and DVD playback.
     
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...