There seems to be a little bit of confusion about what an FPGA can do.
A processor's hardware is fixed. You can only express your functionality in terms of its instructions.
An FPGA is simply a huge collection of logic gates, flipflops, registers.
The programmability is in the interconnection.
If there is a functionality your CPU cannot efficiently implement using the processor's instructions, you describe the functionality in AND's and OR's and gates, in fact you build a circuit inside the FPGA.
On hardware level the FPGA is not faster than the hardware of the CPU. It may be able to execute some functionality faster than a CPU because you create the optimal hardware (description) for that specific functionality.
In an FPGA you can also implement (part of) a processor, if that helps executing your functionality faster.
If the FPGA integrated with the AMD CPU is 'big enough' you could map an ARM or a RISC-V processor on it.
But in the end, an FPGA is just a bunch of gates and basic logic circuits you can interconnect