I recently bought a Sapphire 3850 Ultimate graphics card, and there is similar situation with power management as the one I've described for 1950Pro at http://www.phoronix.com/forums/showthread.php?t=6522. It seems that the 3850 driver uses a different method to keep the GPU cooler and use less electricity: dynamic clock management.
When running Linux and booting to Windows to measure GPU temperature, card reports 50 deg C (after a relatively short period of running linux). In Windows it goes down to 40 degrees.
Since the rovclock utility doesn't work, and aticonfig --list-powerstates lists only gibberish, there is no way for me to keep the card cool and friendlier to the environment under linux currently.
Can we have a list of registers to monitor GPU usage%, and to set and get memory and GPU clock, and possibly to read the temperature of the GPU? What about the mentioned lmsensors support for hw monitoring for the r300-r500 series?
When running Linux and booting to Windows to measure GPU temperature, card reports 50 deg C (after a relatively short period of running linux). In Windows it goes down to 40 degrees.
Since the rovclock utility doesn't work, and aticonfig --list-powerstates lists only gibberish, there is no way for me to keep the card cool and friendlier to the environment under linux currently.
Can we have a list of registers to monitor GPU usage%, and to set and get memory and GPU clock, and possibly to read the temperature of the GPU? What about the mentioned lmsensors support for hw monitoring for the r300-r500 series?
Comment