PC Architecture

:workload_characterization:pc_architecture.jpg

Data Flow Form CPU to GPU

:workload_characterization:cputogpu.jpg

Fast Write

Graphics Without Fast Writes Implementation

The picture below shows the “next-generation” system, implemented using Intel® 820 chipset (AGP 4X) and Direct RDRAM. With the system front side bus operating at 133MHz (1.06 GB/sec) and system memory at 1.6 GB/sec, a bottleneck still occurs when 900MB/sec data transfers are required.

:workload_characterization:dataflow.jpg

In this case, the graphics controller does not implement Fast Writes. This causes the CPU to stall while it waits for the data to be transferred over the system. Even with 1.6 GB/sec, the system memory bus is unable to handle the constant shuffling of 900MB/sec data in and out of system memory. This thrashing on the system memory bus will also cause overall system performance to degrade.


GeForce 256 With Fast Writes Implementation

Fast Writes capability enables the CPU to send data directly to the graphics bus without going through the system memory. This benefits overall system performance in two ways:

  1. Removes the bottleneck previously that exists in systems without Fast Writes.
  2. Frees up system memory bus to perform other functions and increase system efficiency.

The picture below shows how the GeForce 256 graphics processor work with Fast Writes implemented:

:workload_characterization:dataflow2.jpg

 
workload_characterization/workload_characterization/data_flow_from_cpu_to_gpu.txt · Last modified: 2010/05/22 09:20 (external edit)
 
Recent changes RSS feed Creative Commons License Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki