Opengl calls are done in the MT driver threads (I'm sure it is the same for DX)
You only have one GPU. GPU configuration (shaders/alpha...) will be done serialy anyway.
The issue is not to send lots of object into the GPU, the issue is to send them as fast as possible. Generally game sends data a couple of frame before on a standard PC to mitigate the enoumous latency of GPU communication. That why an APU could be faster to emulate the GS. A PS2 is only 2 (maybe 1) texture unit, even smartphone got more
You only have one GPU. GPU configuration (shaders/alpha...) will be done serialy anyway.
The issue is not to send lots of object into the GPU, the issue is to send them as fast as possible. Generally game sends data a couple of frame before on a standard PC to mitigate the enoumous latency of GPU communication. That why an APU could be faster to emulate the GS. A PS2 is only 2 (maybe 1) texture unit, even smartphone got more