Virtual Texturing versus UMA
Pros and Cons
- UMA can enable very low cost systems
- But adding graphics bandwidth load into main system memory can be a heavy burden
- Bandwidth load of graphics sub-system approaching 8 GB/s
- Vertex stream - 1GB/s
- Texture read - 4 GB/s
- Framebuffer/Z buffer - 2 GB/s
- Screen refresh - 1 GB/s (1920x1080x32x85Hz)
- A graphics card - the cost effective way of adding 8GB/s bandwidth?
- Main system memory is the most expensive place to add more bandwidth
- Absorbs framebuffer/z buffer and screen refresh bandwidth
- Virtual texturing further reduces system loading
-