09-02-2019 11:11 AM
Personal website: ThinkScopes
09-03-2019 03:02 AM
To get to a solution faster. Post the following information if you are affected when hybrid mode is enabled:
- BIOS version
- Integrated graphics driver version
- Discrete graphics driver version
I have the P72 and all drivers are up to date. directly from the website.
I had the issue in the beginning of the year and then switched to discrete mode and left and there.
This week, I tried the hybrid mode again after many driver updates from lenovo to check if this issue was solved.
I would say that the issue is not completely gone but much better than before. It might have something to do with the fact, that I have a 4K screen in my P72 and clone that 4K screen to a 1080p beamer via HDMI. Intel is generating the 4K desktop and Nvidia is scaling this picture to 1080p for the hdmi output. This causes high CPU usage for Desktop Window Manager (dwm.exe) which causes lags in hardware accelerated task like youtube. Running two youtubes with 60fps is not possible.
If I change the desktop resolution to 1080p and therefore remove the need of scaling to the beamer because it is 1:1 already, the problem is more or less gone.
My suggestion is, that the commnication between Intel HD Graphics 630 and Nvidia P2000 is not good.
Intel HD Graphics 630 alone should be able to run 4K without issues, even run 4K desktop +2x external extended 1080p screens.
But with the hybrid mode and Nvidia Quadros, all external monitors are always forced through Nvidia and internal laptop screen is forced through Intel, makes it inevitable that these two drivers need to communicate and synchronise when cloning desktop.
09-03-2019 08:08 AM
I finally tried yesterday on my P52 with FHD screen and P1000 connected to a FHD external screen and noticed no lag. However, I was not doing anything intense.
All was up to date on Windows 10 1809, and it was running in hybrid graphics. Intel was managing the built-in FHD screen, Quadro P1000 managed the external HDMI FHD monitor. I think I had set the GPU to auto-select.
One more data point.