Boost Computer Performance Using CPU-GPU Collaboration
Researchers from North Carolina University have designed a way to boost computer performance by over 20 per cent. This process involves the collaboration of GPUs and CPUs on a single chip which facilitates better performance from each.
Chip manufacturers have now drifted towards the new technique of "fusion chips" which enables GPU and CPU to rest on a single chip. This approach is helpful in reducing manufacturing costs and makes the computers more energy efficient. Though on the same chip, the GPU and CPU have entirely different functions and carry on their tasks almost exclusively. The rare collaboration between CPU and GPU is exactly the issue Dr. Huiyang Zhou and his colleagues are trying to resolve.
#-Link-Snipped-#
GPUs were originally designed with a purpose of executing sophisticated graphics programs and have the ability of executing individual functions rather quickly. The CPU on the other hand is the core of the computer, responsible for complex tasks, though operating with a lesser computational power when compared to GPUs. The aim is to integrate each of their services such that an efficient mechanism is reached.
Researchers at the University have planned an approach which allows the GPU cores to execute computational functions while CPU cores pre-fetch the data the GPUs will need from off-chip main memory. This way, both cores do what they are best at. A CPU is assigned the task of determining the data GPU will need while the GPU only focuses on executing them quickly. This technique wouldn't have been possible earlier as both the cores rested on different chips. Preliminary testing of the new approach resulted in a performance increase by an average of 21.4 percent.
The paper,<em> "CPU-Assisted GPGPU on Fused CPU-GPU Architectures,"</em> is scheduled to be presented on Feb. 27 at the 18th International Symposium on High Performance Computer Architecture, in New Orleans.
Source: #-Link-Snipped-#Â Image Credit:#-Link-Snipped-#
Chip manufacturers have now drifted towards the new technique of "fusion chips" which enables GPU and CPU to rest on a single chip. This approach is helpful in reducing manufacturing costs and makes the computers more energy efficient. Though on the same chip, the GPU and CPU have entirely different functions and carry on their tasks almost exclusively. The rare collaboration between CPU and GPU is exactly the issue Dr. Huiyang Zhou and his colleagues are trying to resolve.
#-Link-Snipped-#
GPUs were originally designed with a purpose of executing sophisticated graphics programs and have the ability of executing individual functions rather quickly. The CPU on the other hand is the core of the computer, responsible for complex tasks, though operating with a lesser computational power when compared to GPUs. The aim is to integrate each of their services such that an efficient mechanism is reached.
Researchers at the University have planned an approach which allows the GPU cores to execute computational functions while CPU cores pre-fetch the data the GPUs will need from off-chip main memory. This way, both cores do what they are best at. A CPU is assigned the task of determining the data GPU will need while the GPU only focuses on executing them quickly. This technique wouldn't have been possible earlier as both the cores rested on different chips. Preliminary testing of the new approach resulted in a performance increase by an average of 21.4 percent.
The paper,<em> "CPU-Assisted GPGPU on Fused CPU-GPU Architectures,"</em> is scheduled to be presented on Feb. 27 at the 18th International Symposium on High Performance Computer Architecture, in New Orleans.
Source: #-Link-Snipped-#Â Image Credit:#-Link-Snipped-#
0