AMD SDK teases dynamic GPU clocking technology
February 6, 2013 | 11:21
Companies: #amd #nvidia #techpowerup
AMD is thought to be planning a dynamic graphics processor overclocking technology as part of its OverDrive functionality, in an attempt to better compete with rival Nvidia's GPU Boost system.
According to details uncovered in AMD's Display Library software development kit documentation by techPowerUp, the next generation version of AMD's OverDrive technology - OverDrive 6 - will bring with it the ability for applications to manipulate the performance characteristics of a graphics processor on a very low level.
Under OverDrive 6, the site claims, graphics processors can be tweaked for running speed with adjustments to both the engine clock and the memory clock, while the application programming interface (API) itself provides feedback on current GPU load levels. Combined with existing API calls for reading the temperature of the GPU, OverDrive 6 will theoretically allow for true dynamic overclocking. That's a distinct contrast to OverDrive 5, which uses a feature dubbed PowerTune with Boost to run the graphics card at a higher-than-advertised clock speed as standard and only clocks the card down should the system detect that the GPU is working too hard. The result is a card which runs in its 'overclocked' state most of the time, meaning AMD and its board partners typically set the boost level low to ensure stability during long periods of use.
Rival Nvidia, of course, already has something along the lines of OverDrive 6: GPU Boost. Using a similar API to monitor GPU load, recent Kepler-based Nvidia cards dynamically overclock the GPU within a pre-set range while dialling it back if the power draw or temperature gets too much for the system to handle. Compared to the somewhat rough control offered by AMD's OverDrive 5, that's a much better system - and one from which AMD is apparently keen to take inspiration.
AMD, naturally, is unwilling to comment on techPowerUp's interpretation of the SDK documentation, but it's hard to argue with the facts: API calls for altering the engine and memory clocks of a GPU are included in the OverDrive 6 section of the ADL, as is a call to monitor system load. Why they would exist if AMD wasn't planning true dynamic clocking for its future GPUs is not clear.
If you're curious, the full ADL SDK documentation is available direct from AMD.
According to details uncovered in AMD's Display Library software development kit documentation by techPowerUp, the next generation version of AMD's OverDrive technology - OverDrive 6 - will bring with it the ability for applications to manipulate the performance characteristics of a graphics processor on a very low level.
Under OverDrive 6, the site claims, graphics processors can be tweaked for running speed with adjustments to both the engine clock and the memory clock, while the application programming interface (API) itself provides feedback on current GPU load levels. Combined with existing API calls for reading the temperature of the GPU, OverDrive 6 will theoretically allow for true dynamic overclocking. That's a distinct contrast to OverDrive 5, which uses a feature dubbed PowerTune with Boost to run the graphics card at a higher-than-advertised clock speed as standard and only clocks the card down should the system detect that the GPU is working too hard. The result is a card which runs in its 'overclocked' state most of the time, meaning AMD and its board partners typically set the boost level low to ensure stability during long periods of use.
Rival Nvidia, of course, already has something along the lines of OverDrive 6: GPU Boost. Using a similar API to monitor GPU load, recent Kepler-based Nvidia cards dynamically overclock the GPU within a pre-set range while dialling it back if the power draw or temperature gets too much for the system to handle. Compared to the somewhat rough control offered by AMD's OverDrive 5, that's a much better system - and one from which AMD is apparently keen to take inspiration.
AMD, naturally, is unwilling to comment on techPowerUp's interpretation of the SDK documentation, but it's hard to argue with the facts: API calls for altering the engine and memory clocks of a GPU are included in the OverDrive 6 section of the ADL, as is a call to monitor system load. Why they would exist if AMD wasn't planning true dynamic clocking for its future GPUs is not clear.
If you're curious, the full ADL SDK documentation is available direct from AMD.
MSI MPG Velox 100R Chassis Review
October 14 2021 | 15:04
Want to comment? Please log in.