Little to no increase in speed after establishing flowtable software acceleration rules

[1yanzhaoli]

Create the following acceleration rules:
table ip filter {
flowtable f {
hook ingress priority filter
devices = { eth0.2, nas3_0 }
}

    chain forward {
            type filter hook forward priority filter; policy accept;
            ip protocol { tcp, udp } flow add @f
            counter packets 0 bytes 0
    }

}
Check out cat /proc/net/nf_conntrack:
ipv4 2 tcp 6 src=192.168.1.20 dst=10.10.10.13 sport=63522 dport=5201 src=10.10.10.13 dst=10.10.10.10 sport=5201 dport=63522 [OFFLOAD] mark=0 use=3
But iperf3 stream test (tcp) results: compared to when no acceleration rule was established, no change in upstream, and downstream increased a little bit (about 2 Mbit/s)
What are the reasons for this situation? How should it be solved?