Close
Page 2 of 2 FirstFirst 1 2
Showing results 11 to 12 of 12
  1. #11

    Default

    All your links are to ruddy benchmarks I don't care about because I'd like to think I'm talking mostly abstractions.

    You seems to think that you will get double the performance with 2 gpu as you would with 1 gpu simplly because you are 5 boxing.
    Not quite. By distributing the load between gpus you would get higher consistent performance across more clients than with one gpu. It might not be double normal performance. It might not be 25% higher. But assuming equivalent gpus you would better performance in high-intensity situations. Not having access to new cards/not multiboxing I myself can't test these things. It is my opinion.

    Looking at wikipedia the PCIe v2 x16 bus is 16GB/s. With even a dual-gpu card I don't think that's going to be an issue. Now, I'm not certain of how bandwidth is affected by having multiple cards in multiple pcie slots, but I would say each card/slot should have the same bandwidth, depending on the motherboard. With a lesser motherboard those cards would be limited depending on specs.

    If the gpu's cannot access the other gpu memory space you have to load the SAME TEXTURE into VIDEO RAM TWICE. I don't think they would specify it as 2X1G if the gpu's can access the others memeory space. With two cards we know for sure that one gpu cannot accress the other gpu's memory space. If you have to load the same texture to video ram twice you are mitigating any potential benifits of two gpus.
    Looks like you're assuming that the application in use will check for textures already in use and reuse the already used textures (i.e caching). For all we know that behavior isn't there and I would bet it isn't. So sharing the same memory space doesn't do much except clutter it up. Now, with more ram there is more space to fill up, so textures could remain in memory longer.

    In these situations when we are talking performance loss I think we are talking MINUTE losses. The bandwidth is there. The slowest part of the system is the storage drive/interface.

    And "distribution" of 3 on one vs 2 on the other means that one gpu is working 1.5X the other meaning that one gpu is doing nothing 1/3 of the time.
    So? You get better performance overall with leeway for higher load either now or in the future due to either textures/graphical needs increase or added clients.

    Just buy the cheapest 2G video card you can .... then spend the money you saved on system ram.
    That I won't argue.

    "I be frag'en while you be lag'en" - SDW '11 (copywrite Sam Ferris 2011 all rights reserved)


    LMAOROTFL

    I ain't brag'en
    but I be frag'en
    while you be lag'en
    and my epen be DRAGON!

    LOLOL
    Bugger off you.
    Hardware Lurker

  2. #12
    Member
    Join Date
    Jun 2008
    Location
    Fremantle, Western Australia
    Posts
    406

    Default

    I am taking sides with Sajuuk on this one.

    Quote Originally Posted by Sam DeathWalker View Post
    "I be frag'en while you be lag'en" - SDW '11 (copywrite Sam Ferris 2011 all rights reserved)


    LMAOROTFL

    I ain't brag'en
    but I be frag'en
    while you be lag'en
    and my epen be DRAGON!

    LOLOL
    Really?

    .... REALLY?

Tags for this Thread

Posting Rules

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