Any feedback? This is very easy to reproduce and happens on every single x86 machine, with tens of B210 devices around the world. The command line in my original message gives problems even when 25% of the USB throughput is removed by forcing "sc12" format over the default "sc16" format.
The hope is always that newer drivers don't break the existing systems, which did fail us here. Thanks, Dario On Wed, Apr 13, 2022 at 3:54 PM Dario Fertonani <[email protected]> wrote: > Hi, > > I'm here to report that the "golden standard" for benchmarking the B210 > streaming performance is giving us issues with the latest UHD versions. > > The following command > > ./benchmark_rate --rx_channels "0,1" --rx_rate 30720000 --tx_channels > "0,1" --tx_rate 30720000 --duration 120 > > gives flawless output with UHD 3.10.3 and 3.15.0, but orrible output with > UHD 4.1+, with everything else being equal. These results hold on several > x86 machines (Intel i7, i9, Xeon, and AMD Ryzen and Epyc) and with at least > five different B210 devices. In each case, for a given host machine, USB > cable, and B210 device, we compiled UHD from source and saw the results > degrade sharply simply by running the command from the newer folders. > > We typically run UHD from Ubuntu packages (3.10.3 for Ubuntu 18, 3.15.0 > for Ubuntu 20, or whatever is in the latest Ubuntu PPA) rather than > building it from source code. We went through the source code route because > we saw issues and wanted to reproduce them with the official > "benchmark_rate" utility, which unfortunately comes only with the source > code version. However, I want to point out that the poor results seen > with "benchmark_rate" are consistent with the poor results of our own code > when built with newer UHD versions. > > Thanks, > Dario > > > >
_______________________________________________ USRP-users mailing list -- [email protected] To unsubscribe send an email to [email protected]
