42 lines (32 with data), 1.7 kB
Here is a small plan regarding the RE the powermanagement of NVIDIA videocards
PM mode - Adaptive
1. Method of monitoring the increasing the clocks, sending a message on any
itermediate steps (if any) and in the end, i.e. when reached the maximum frequencies
- Use a simple GL application, and modify it
- Change the PM mode to "Maximum performance" - the actual procedure of "reclocking"
2. Method of monitoring the "down-clock" - DONE
- After running a GL application run another program that does the "monitoring"
and dumps messages
For both types of dumps, the following information have to be dumped
1. The performance mode and actual clocks (as reported by the driver) - DONE
2. Card specific "registers" - see note below
3. Number of fans and their speed (rpm/percentage) - would have to be implemented by
a person with a fan on the videocard
4. Temperature related readings - current, threshold, critical etc. - DONE
Look at the following and find any connection between them and the PM
1. Kernel log with enabled drm driver debugging ("drm.debug=0x2" appended when loading)
followed by the load of nouveau. The log should contain the initialisation scripts
"Checking if 0xFF equals 0xFA"
"Cond: 0x1A, Reg: 0x100231, Mask: 0x000000FF"
2. Amount and type of memory on the card
3. Look at the unused tables in the BIT_P section of the vbios
1. Implement monitor_upclocking
2. Add relevant fanspeed bits to monitor_downclocking, so that the relevant
information can be extracted easily
3. Modify monitor_downclocking so that it reads the relevant data/address
4. Setup an email account of the users' contributions
5. Working solution for SLI and multicard setups