Overclock.net banner
1 - 2 of 2 Posts

4,968 Posts
Originally Posted by hks85 View Post

I have no experience using a GPU for video transcoding. $ for $, what could i expect with CPU cluster vs GPUs?
Graphics cards can offer up to 40 times greater cores than modern CPUs, even though they are conceptually different and cannot handle many different types of tasks. The Radeon 5xxx series have up to 1600 stream processors. These cores are able to take on one thread each, but are packed into clusters that can make use of only one processing operation on the threads they handle. A GPU is therefore unsuitable for complex tasks which require multiple types of processing at a time. However, for instances such as video transcoding, where only one thread is processed at a time, GPUs take the load off the CPU and process at a MUCH faster rate, owing to the high number of processing cores.

CPU advantage:
If the requirements and tasks are different (multitasking), a quad core CPU can handle 2 instances per core at a time

GPU disadvantage: The GPU cannot work parallely in case of complex problems and processes the instances individually at a much slower speed.

CPU disadvantage: In case of identical tasks (such as video transcoding), the quad core still processes 2 instances per core at a time

GPU advantage:.
If the GPU has 8 shaders with 48 clusters in each, the instances will be processed one core at a time = 384 instances at a time, as opposed to 8 on the CPU.

So to answer your question, GPUs would be much faster at transcoding
1 - 2 of 2 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.