On 15/11/2024 16:52, Eugene Grayver wrote:
Hi,
I am considering the X440 for a wideband record/playback system. What
benchmarks has Ettus done?
Is there any hardware out there that can continuously stream (one way,
TX or RX) the full 8 Gsps (i.e. 2x 4 Gsps) for the combined bandwidth
of 2x 1.6 GHz? Assuming DPDK is used, there is still an issue with
getting packets to/from different cores/threads. Are the packet
streams configurable to allow hardware-level queues that map to
different IRQs/cores? How does the TX side work (that's usually a lot
harder to maintain than RX)?
Thanks.
I don't have any direct experience with X440. But I know that newer
versions of "benchmark_rate" have an option for
handling different streams in different threads, so that might help
you with your own code.
Ingesting 3.2Gsps all by itself is a compu-difficult problem, let alone
trying to actually *DO* anything with those samples.
It has frequently been the case with USRPs that at introduction,
their ability to produce high-speed samples outstrips
the ability of extant computers to ingest those samples.
I know that a LONG time ago, UHD experimented with multi-threaded
"gather" from the network interface for even
a single stream. The problem that was immediately apparent was that
the MUTEX traffic that was required, along
with dealing with the potential re-ordering when multiple threads are
gathering buffers of samples, lead to much
poorer overall performance.
_______________________________________________
USRP-users mailing list -- usrp-users@lists.ettus.com
To unsubscribe send an email to usrp-users-le...@lists.ettus.com
