It's not about parallelization, it's about flexibility.
AFAIK most graphics problems in N64 emulation come from differences in memory management model. The N64 (and most consoles, actually) had no distinction between RAM and VRAM and the CPU could access frame buffer data whenever it pleased. Also, there were a bunch of features that didn't map directly to traditional GPU feature sets. Also, the cost for software emulating the N64 GPU was orders of magnitude greater than the PSOne, due to things like texture filtering, floating point operation and depth buffer, which pushed emulators into HLE since the beginning.
Modern GPUs are far more flexible than those from the heyday of N64 emulation. They can use compute shaders to replicate hardware-specific features of the emulated hardware, like the N64's edge anti aliasing or the Dreamcast's per pixel translucency sorting. They are also much more flexible when it comes to memory management, compared to when emulators had to rely on painfully slow copying of frame buffer pixels to emulate off screen buffers.
I believe an emulator designed around DX12/Mantle/Vulkan, instead of merely a plugin, could do a lot more a lot faster.