overall performance is significantly affected.
On the other hand, the new transform matrix must be transferred to the GPU overloading the CPU-GPU communication bus.
Although the amount of data is quite small, the bus latencies of these small devices can have a negative impact which means a notable bottleneck.