Hi Joe, Did you change the chdr_width in the rfnoc_image_core YAML file and rerun rfnoc_image_builder on that file?
Which version of UHD are you using to build the FPGA? Streaming over 10 GbE to 128-bit CHDR with 500 Msps radio isn't a use case we test, since 10 GbE can't keep up. Normally we would pair that with 100 GbE. That doesn't explain why you're getting worse performance though. I don't know why it wouldn't work as well as 64-bit CHDR. Wade On Tue, Aug 15, 2023 at 10:00 AM <jmalo...@umass.edu> wrote: > Hello, > > I am currently running into buffer overflows with the x410 where I am > unable to stream for data rates larger than ~0.32 Gb/s when the CHDR_W is > 128. However, if I stream using a CHDR_W of 64, I am able to push the data > rate much higher(at least 3.2 Gb/s). I compiled a custom image using X4_400 > > My current set up transmits data over 1 SFP port, which is rated for 10 > Gb/s. I guess is that since technically this port cant sustain the data > rate of 500 MS/s(equal to 16 Gb/s), there could be at least 1 buffer than > can not sustain the given data rate, even if I am not sampling continuously. > > > I tried increasing the endpoint buffer sizes (REQ_BUFF_SIZE_EP) as well as > increasing the buffer size set by the host computer, but both seem to have > no impact on solving this buffer overflow issue. > > I saw a hard coded value in the x4xx.v where BYTE_MTU was set to > clog(8192). Do I need to increase this buffer size as well? > > Thanks > > Joe > > > _______________________________________________ > USRP-users mailing list -- usrp-users@lists.ettus.com > To unsubscribe send an email to usrp-users-le...@lists.ettus.com >
_______________________________________________ USRP-users mailing list -- usrp-users@lists.ettus.com To unsubscribe send an email to usrp-users-le...@lists.ettus.com