Doom 3 Performance

Under Doom 3, we see quite a performance hit. The fact that Id software makes heavy use of the stencil buffer for its shadowing means that there's more pressure on bandwidth per pixel produced than in other shader heavy games. In Half-Life 2, for instance, we see less of a performance impact when moving from the old 128-bit 6200 to the new version.

Doom 3 Performance
Doom 3 Resolution Scaling

Our resolution scaling graph shows that the performance gap between the X300 and 6200 parts closes slightly as resolution increases. Running High Quality looks great even at 640x480, but we would suggest Medium Quality at 800x600 for playing the game.

The Test Far Cry Performance
Comments Locked

43 Comments

View All Comments

  • PrinceGaz - Thursday, December 16, 2004 - link

    #28- see page 2 of the article, the text just above the diagram near the bottom of the page "Even on the 915 chipset from Intel, bandwith is limited across the PCI Express bus. Rather than a full 4GB/s up and 4GB/s down, Intel offers only 3GB/s up and 1GB/s down..."

    #25- I'd also always assumed that all PCIe x16 sockets could support 4GB/s both ways, this is the first time I've heard otherwise. And it isn't even 4/1, it's 3/1 according to the info given.

    Derek- is this limited PCIe x16 bandwidth common to all chipsets?
  • DerekWilson - Thursday, December 16, 2004 - link

    We tested the 32MB 64-bit $99 version of the card that "supports" a 128MB framebuffer.

    #31 is correct -- the maximum of 112 of 96 (or 192 for the 256 MB version) of system RAM is not staticly mapped. It's always avalable to the system under 2D operation. Under 3D, it's not likely that the entire framebuffer would be absolutely full at any given time anyway.
  • Alphafox78 - Thursday, December 16, 2004 - link

    doesnt it dynamicly allocate the extra memory it needs? so this would just affect games then if it needed more, not regular apps that done need lots of video memory.
  • rqle - Thursday, December 16, 2004 - link

    so total cost of these card is the card price + (price of 128MB worth of DDR at the time)?
  • Maverick2002 - Thursday, December 16, 2004 - link

    I'm likewise confused. At the end of the review they say:

    "There will also be a 64MB 64-bit TC part (supporting 256MB) available for $129 coming down the pipeline at some point, though we don't have that part in our labs just yet."

    Didn't they just test this card???
  • KalTorak - Thursday, December 16, 2004 - link

    #25 - huh? (I have no idea what that term means in the context of PCIe, and I know PCIe pretty well...)
  • KayKay - Thursday, December 16, 2004 - link

    I think this is a good product, i think it could be a very good part for companies like dell, if they include it into their systems. cheaper than the x300se's they currently include, but better performance, and will appeal to that type of customer
  • mczak - Wednesday, December 15, 2004 - link

    #24, from the description it sounds like for the radeon igp there is no problem with both using sideport and system memory simultaneously for directly rendering into (the interleaved mode exactly sounds like part of all buffers would be allocated in system memory, though maybe that's not what is meant).
  • IntelUser2000 - Wednesday, December 15, 2004 - link

    WTF!! I never new Intel's 915 chipsets used 4/1GB implementation of PCI Express!! Even Anandtech's own article didn't say that they said 4/4.
  • DerekWilson - Wednesday, December 15, 2004 - link

    As far as I understand Hypermemory, it is not capable of rendering directly to system memory.

    Also, when Hypermemory needs to go to allocate system RAM for anything, there is a very noticeable performance hit.

    We tested the 16MB/32-bit and the 32MB/64-bit

    The 64MB version available is only 64-bit ... NVIDIA uses four 8M x 16 memory chips.

Log in

Don't have an account? Sign up now