Every time I get big numbers - (UDP) around 80Gbps for my CCR1009 and above 500Gbps for my Xeon E5-2403 server.

But what is it testing? How is it significant?How does it translate into the real world?

For the above, even though Xeon is more than 5x faster, in production (no FW, just simple cake queues per client) it can barely push 500Mbps with latency spikes (90% CPU). The CCR goes above 700Mbps just fine.

It's testing loopback device. Essentially CPU performance. And it only shows bandwidth test CPU capacity, nothing else. No real world meaning what so ever.

127.0.0.1 is loopback address. You're testing towards your own device.
Real life value = close to 0 for network (maybe bit CPU).

EDIT: damn, same answer as mkx

More specifically, it tests single-thread CPU performance. So a modern CPU with many cores, or a CCR, will appear relatively slow here, while in practical use where hopefully many cores are used, it is faster.

I must say, I never thought of running that test on that loopback itf.

I stand corrected ... it HAS real life value !
Proof on AC3:

NOT !!

But what is it testing? How is it significant?How does it translate into the real world?

It translates like this: that the admin instead of playing solitaire for the day, decided to make useless tests.
相同的结果。

This test has one important application. It shows the maximum capability of the processor to generate traffic. And if in the UDP everything is of little use in life, the TCP everything is clear. Many users, when trying to perform such tests, get bogged down by the performance of the processor and start looking for problems that don't exist.
In the photo tests AC2

PS Added AC Lite for comparison

Tss tss ... cap lite
Odd that cap lite (which is mostly same as map and map Lite) performs slightly better then hap ac lite ?
That's the difference between QCA9531 and QCA9533 ?
One other difference I see: I already put that device on 7.7

The performance of the processors is about the same 50mbit CPU 87%, 55mbit CPU 100%.

This test has one important application. It shows the maximum capability of the processor to generate traffic. And if in the UDP everything is of little use in life, the TCP everything is clear. Many users, when trying to perform such tests, get bogged down by the performance of the processor and start looking for problems that don't exist.
In the photo tests AC2

Ah yes, that's one thing. I did test on TCP and got a result (500Mbps), and when I hit my server with real traffic it did cap around 500Mbps. I guess it does have value.