I experimented a little bit further with VERY interesting results:
Initial config as reference:
Display attached to discrete AMD HD 7750 GPU
PD11 uses only the discrete 7750 GPU. 7750 GPU 80% load, HD4000 zero load, CPU load app. 15%
The acceleration offered by PD11 says 'Hardware Acceleration'.
Sample video (MPEG4.ts, 20min stream length) renders in 7min
And now the scenarios I went through
(1) I attached my display to the on-board HD4000 GPU -
Result:
PD11 uses only the on-board GPU (HD4000). HD4000 GPU 80 load %, HD7750 zero load, CPU load app. 10%
The acceleration offered by PD11 says 'Hardware Acceleration'.
Sample video renders in 5min (!)
(2) Now I added PD11 to the list of apps in Lucidlogix Virtu MVP (the Graphics virtualization layer) - display still attached to Intel HD4000 onboard graphics. Restart of PD11.
Result: The acceleration offered by PD11 suddenly changed to 'Intel Quick Sync'.
Both GPUs under load. HD4000 - 80%, HD7750 - 2-7% dependent whether preview is used or not. CPU load app. 8%.
Sample video renders in 2:30min
(3) Finally I attached my display back to my AMD HD7750 GPU as the qualityof screen-display is visibly better. PD11 still defined as an app to be accelerated in Lucidlogix Virtu MVP. restarted PD11.
Result: Get the acceleration option 'Intel Quick Sync' presented.
Both GPUs under load. HD4000 - 80%, HD7750 - 1-7% dependent whether preview is used or not. CPU load app. 8%
Sample video renders in 2:30min
NOW THIS IS WHAT I WAS LOOKING FOR - both GPUs employed for superfast rendering.
Question to Cyberlink remains why PD11 is unable to leverage both GPUs natively - i.e. why Lucidlogix is needed. The PD11 product specs explicitely say 'mullti-GPU enabled'. That would warrant a response from Cyberlink!
This message was edited 4 times. Last update was at Dec 17. 2012 05:46