The GeForce GTX 280 and GeForce GTX 260 support 1,024 MB and 896 MB of
frame buffer respectively, a two-fold improvement from over prior generation
GPUs. With 1 GB of frame buffer, high-resolution antialiasing performance is
dramatically improved. For example, deferred rendered games like S.T.A.L.K.E.R.
can now be enjoyed with antialiasing.
Geometry Shading and Stream Out
Internal output buffer structures have been significantly upsized by a factor of 6× in
GeForce GTX 200 GPUs compared to the prior generation, providing much faster
geometry shading and stream out performance. Figure 8 shows the latest RightMark
3D 2.0 benchmark results, including geometry shading tests. The GeForce GTX
280 GPU is significantly faster than prior generation NVIDIA GPUs and
competitive products.
Geometry Shader Performance
Rightmark 3D 2.0 - Hyperlight Heavy
http://w w w .ixbt.com /video/itogi-video/ini/rm dx10.rar
0
100
200
300
400
500
600
Low
Med
High
8800 GTX
3870
3870X2
GTX 280
Figure 8: Geometry Shading Performance
Our own Medusa demo is highly dependent on the much faster geometry and
stream out performance.
512-bit Memory Interface
Maximum memory interface width is expanded from 384 bits in previous-generation
GPUs to 512 bits in GeForce GTX 200 GPUs, using eight 64-bit-wide frame buffer
interface units. Memory bandwidth has been significantly increased.
In terms of rebalancing the architecture versus prior generations, the texture to
frame buffer (TEX:FB) bandwidth ratio has also been modified to best support
current and future workloads. NVIDIA engineers tested many applications to arrive
May 2008 | TB-04044-001_v01
17
