Aang23
commented
8 months ago Did the same tests with the UC_200 image attempting to stream at 245.76MSPS on the exact same system. Unfortunately that does not work at all either with the same symptoms as using the CG_400 image.
Benchmark rate summary:
Num received samples: 0
Num dropped samples: 0
Num overruns detected: 0
Num transmitted samples: 2455190080
Num sequence errors (Tx): 0
Num sequence errors (Rx): 0
Num underruns detected: 1734
Num late commands: 0
Num timeouts (Tx): 0
Num timeouts (Rx): 0And while it does manage to sustain 122.88MSPS a bit better, underruns are also observed along the way. There are very few in the benchmark at random points during the test, but they happen more often in any other UHD transmit example.
Benchmark rate summary:
Num received samples: 0
Num dropped samples: 0
Num overruns detected: 0
Num transmitted samples: 1228853888
Num sequence errors (Tx): 0
Num sequence errors (Rx): 0
Num underruns detected: 5
Num late commands: 0
Num timeouts (Tx): 0
Num timeouts (Rx): 0The same system does manage to keep a x310 at 200MSPS happy however.
basti-schr
mbr0wn
ZangMT
Hello!
Lately I have been doing some tests and writing code with an USRP x410 I have access to for the moment (via https://twitter.com/SDR_Radio).
Getting it working after flashing the correct image was not too complicated, except some documentation was lacking (specifically, the MTU configuration is mentioned in the x3xx documentation but entirely missing on the x4xx side of things).
In receive, after flashing the CG_400 image and doing all the required network configuration, I am able to stream the full 471.52MSPS with no issues at all using SatDump. The
benchmark_rateexample also shows no drops, overflows etc in cf32 host format and sc16 wire format.However, when I got to attempting to transmit, nothing I was able to write was able to stream as the expected full rate of 471.52MSPS without major underruns. This occurs even with a simple while loop sending large buffers or similar, with UHD seemingly saturating that thread already. It's visible enough the transmit (red) LED on the USRP visibly blinks constantly. This behavior is also observed using the provided
benchmark_rateexample. (I have also attempted modifying various buffer sizes and network configuration to no avail.)RX Bench :
TX Bench :
The x410 is connected via QSFP+ (100GB) on Mellanox cards on a high-end workstation :
Mellanox Technologies MT27800 Family [ConnectX-5]Intel i9-10980XE (36) @ 4.600GHzI have tested raw streaming over the 100GB link and the machine is able to sustain far more than required to feed the x410 in this configuration. Considering the setup in use I would expect this to work.