We have indeed done this on our Ultrascale+ MPSoC hardware - we reconfigure the 100G ports as 40GbE or N x 10 GbE as necessary.
On Wed, Oct 2, 2024, 18:16 Matthew Schiller <mschi...@nrao.edu> wrote: > This should be relatively straightforward. > > > > There’s nothing special about the Multi-Gigabit transceivers used on the > QSFP28 port. Normally for 100G they use 4 lanes with each lane operating > at 25Gbs (effective, higher due to encoding and error correction, hence the > QSFP28). But the fact is that you can run these at 10Gigabit/sec instead > and make the port a QSFP 40Gbit port, which can then be subdivided to a > 10Gbit port using a breakout cable. > > > > To save power you can then turn off the unused transceivers to get a > single 10Gb port in the FPGA, and similarly with serial commands to the > Fiberoptic transceiver the extra lasers can also be turned off if desired. > [you don’t have to worry about that part if your using a direct connect > cable]. > > > > > > In the fpga: > > - Generate the Ethernet Port for 10Gb as if you were targeting a SFP+ > eg run it in 10Gbase-SR mode. > - This will have the effect of changing the bitwidth going into the > Xilinx IP block to a smaller AXI streaming size which will then trigger > rejiggering of any ethernet packetization logic you might have > implementing > UDP or VDIF or VITA49.2 or what have you > - Replace the casper 100G block with this slower version. > - You’ll need to ensure the bitrate going into the > Ethernet/Packetization logic is <= 10Gbit/sec > - Eg low bitwidth quantization and or decimating to a slow enough > sample rate to fit at 10Gbit/sec > > > > > > From a hardware perspective you’ll want to acquire one of these breakout > cables: > > - > > https://community.fs.com/article/how-to-convert-a-port-from-qsfp-to-sfp-port.html > - (you’ll then just not use 3 out of 4 of the 10G SFP+ connections) > > > > > > > > I hope you find someone. I’d love to volunteer, but I have a full time > job…. Hopefully you’ll see some 400G progress at the next Casper Conference > =c) > > > > > > > > [image: AB72FAB9] > > Matthew Schiller > > ngVLA Digital Backend Lead > > NRAO > > > > mschi...@nrao.edu > > 315-316-2032 > > > > > > > > > > > > > > > > *From:* casper@lists.berkeley.edu <casper@lists.berkeley.edu> *On Behalf > Of *John Swoboda > *Sent:* Wednesday, October 2, 2024 11:40 AM > *To:* casper@lists.berkeley.edu > *Subject:* [casper] Temp FPGA Development Project {External} > > > > Hi, > > > > MIT Haystack in collaboration with Spectrum X is looking for someone to do > temp work with the RFSoCs. The main goal of the project is to find a way > enable a way to stream data out of 100 Gbe on the RFSoC to a 10 Gbe > connection on a host computer (we would provide remote access to a test > bed). The reason for doing this is we are looking for low power solutions > to stream high bandwidth data from the RFSoC. We are specifically looking > at an application with the 4x2 with the hope that this could be extended to > other platforms in the RFSoC family. We will initially target the Pynq > framework to test this capability but are also interested in extending this > to CASPER. > > > > We believe we have the place to start but we need someone with experience > using Vivado and VHDL/Verilog. If you're interested in this task, or know > someone who might be, please contact me. > > > > John Swoboda, Ph.D. (he/him) > Geospace Research Scientist > MIT Haystack Observatory > Email: swo...@mit.edu > Voice: (617) 715-3489 > > -- > You received this message because you are subscribed to the Google Groups " > casper@lists.berkeley.edu" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to casper+unsubscr...@lists.berkeley.edu. > To view this discussion on the web visit > https://groups.google.com/a/lists.berkeley.edu/d/msgid/casper/SJ2PR01MB8029E82828CF4A4F42DE7DF8DC702%40SJ2PR01MB8029.prod.exchangelabs.com > <https://groups.google.com/a/lists.berkeley.edu/d/msgid/casper/SJ2PR01MB8029E82828CF4A4F42DE7DF8DC702%40SJ2PR01MB8029.prod.exchangelabs.com?utm_medium=email&utm_source=footer> > . > > -- > You received this message because you are subscribed to the Google Groups " > casper@lists.berkeley.edu" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to casper+unsubscr...@lists.berkeley.edu. > To view this discussion on the web visit > https://groups.google.com/a/lists.berkeley.edu/d/msgid/casper/BL0PR14MB3523B4244109774E82E2CA98AB702%40BL0PR14MB3523.namprd14.prod.outlook.com > <https://groups.google.com/a/lists.berkeley.edu/d/msgid/casper/BL0PR14MB3523B4244109774E82E2CA98AB702%40BL0PR14MB3523.namprd14.prod.outlook.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "casper@lists.berkeley.edu" group. To unsubscribe from this group and stop receiving emails from it, send an email to casper+unsubscr...@lists.berkeley.edu. To view this discussion on the web visit https://groups.google.com/a/lists.berkeley.edu/d/msgid/casper/CAB4rdA%2B_%2B6E0LR5hX1XdxsFbMUv%2BAvJsL8wgEJSnshJL30L9SQ%40mail.gmail.com.
