integrated card has better score than dedicated

[mihaila]

just did a Bruce X 5K FCP rendering test (prores) with both cards, iris pro finished almost 1 minute earlier than nvidia 750m

late 2013 mbp 2,3ghz 16gb 500gb on Mojave 10.14.6 (gfxCardStatus v2.5)

wtf?

 

[casperes1996]

There is one, and just one, circumstance where the Iris GPU is genuinely faster than the Nvidia one. When the main bottleneck of an operation is CPU<->GPU communication. The dedicated GPU requires memory synchronisation over PCIe, where the iGPU has a shared L4 cache and main memory system to communicate over.
As soon as the bottleneck is shifted more to actual computations or on-die memory bandwidth the Nvidia chip should be faster.

Was this just a test of final render? Did you compare playback smoothness?

 

[mihaila]

There is one, and just one, circumstance where the Iris GPU is genuinely faster than the Nvidia one. When the main bottleneck of an operation is CPU<->GPU communication. The dedicated GPU requires memory synchronisation over PCIe, where the iGPU has a shared L4 cache and main memory system to communicate over.
As soon as the bottleneck is shifted more to actual computations or on-die memory bandwidth the Nvidia chip should be faster.

Was this just a test of final render? Did you compare playback smoothness?

it's the Bruce X benchmark https://blog.alex4d.com/2013/10/30/brucex-a-new-fcpx-benchmark/
the results should be ± like this:
brucex-final-cut-pro-x-benchmark2.png

it seems it's the other way around for me lol.
strange. i tried again without gfxCardStatus switcher