I started working on the AXI 5 but when it came to implementation i observed that AXI4 and AXI5 are giving the same latency. Whether using VALID-READY or CREDIT based system, what is the advantage of AXI5 credit-based System?? How AXI4 and AXI5 works in multi master and multi slave situation?
A quick point of note - AXI4 is VALID/READY only, but in AXI5, you're able to select either VALID/READY or credit based transport from AXI issue L onwards using the AXI_Transport property.
In a simple point-to-point system you’re correct that VALID/READY and credit-based flow control can show similar latency. The main difference is how well the mechanism scales as systems get larger.
VALID/READY works best when the feedback path is short. READY provides immediate backpressure, so it is very efficient when the sender and receiver are close in the pipeline. This is why AXI works well for local bus-style communication.
As systems scale (more cores, deeper pipelines, multiple interconnect hops), the round-trip latency of READY increases. The sender may continue transmitting for several cycles before it sees backpressure, which can reduce efficiency or require additional buffering.
The credit-based approach used in AXI5 decouples flow control from instantaneous backpressure. The receiver provides credits indicating how many transfers it can accept, and the sender can transmit up to that limit without waiting for per-cycle feedback. This makes the transport more tolerant of latency and better suited to deeper pipelines or multi-hop interconnects.
So while the single-transfer latency may look similar, credit-based flow control helps maintain efficiency as system scale and transport latency increase.