Even though ATI Radeon HD 4000-series graphics processing units (GPUs) have become very popular on the market due to their high performance in video games, the family may have limited performance in OpenCL application programming interface-based software, Advanced Micro Devices said.
“There are known performance issues for ATI Radeon HD 4000-series of cards on OpenCL and there is currently no plan to focus exclusively on improving performance for that family. The ATI Radeon HD 4000-series was not designed for OpenCL, whereas the ATI Radeon HD 5000-series series was. There will be performance improvements on this series because of improvements in the ATI Radeon HD 5000-series, so it will get better, but it is not our focus,” said Micah Villmow, a senior compiler engineer in AMD's OpenCL group, in AMD’s official software developer support forums.
Considering the fact that ATI Radeon HD 4000-series (code-named ATI R700) were designed back in 2006, it is not a surprise that they show limited performance in cases of OpenCL 1.0, which was finalized only late in 2008.
“ATI Radeon HD 4000 just has to be programmed differently than the ATI Radeon HD 5000-series to get performance because of the lack of proper hardware local support. It is possible to get good performance, just not with a direct port from CUDA. […] For example, if you are using local memory, they are all currently emulated in global memory. […] This can cause a fairly large performance hit if the application is memory bound. On the ATI Radeon HD 5000-series series, local memory is mapped to hardware local and thus is many times faster than the ATI Radeon HD 4000-series,” explained Mr. Villmow.
AMD indirectly confirms that OpenCL performance of its popular ATI Radeon HD 4000-series graphics boards will be lower compared to that of ATI Radeon HD 5000-series as well as Nvidia GeForce 200-series. The only question is how much lower and whether this will impact the end-user experience any significantly amid the lack of mass OpenCL-based software at present.