On 2022-06-22 09:05, Berkay SAYGILI wrote:
Now, I see that DPDK works fine with given UHD examples. But, it
failed with Gnuradio. Can we say that Gnuradio is the bottleneck? or
am I doing something wrong? Has anyone made the graph in the second
test work? By the way, it is much worse when transmitting with
Gnuradio at full bandwidth with or without DPDK.
To compare Intel and Mellanox NICs, Intel was dropping samples with
Gnuradio with very low sample rates, Mellanox started dropping samples
after 100 MSPS with both channels used. But, I was using DPDK 18.11,
and UHD 4.0 with the Intel NIC, so that might be the case.
Best regards
Berkay SAYGILI
You are, to a certain extent comparing apples and oranges. The
"benchmark_rate" example never actually *DOES* anything with the samples.
Your Gnu Radio example writes them to a file. This is MUCH more expensive.
In *general* the cost-per-sample for your typical Gnu Radio application
will be considerably higher than running simple UHD "native" benchmark
programs.
_______________________________________________
USRP-users mailing list -- usrp-users@lists.ettus.com
To unsubscribe send an email to usrp-users-le...@lists.ettus.com
