Matthew Wielgus

I just noticed the Innova2 8GB AD**L**T uses the _MT2**8**808A0-FCCF-EV_ while the Innova2 4GB AD**I**T uses the _MT2**7**808A0-FCCF-EV_. The `MT28808A0-FCCF-EV` and `MT27808A0-FCCF-EV` ICs may use different coupling interfaces.

By testing unused Quads I found that the 161.13MHz oscillators are connected to pins V27 and Y27, Quads 129 and 130.  ``` # MNV303212A-ADAT and MNV303212A-ADIT Only: # SFP...

I got DDR4 working on my Innova2 4GB ADIT board. @Applepi please test your ADAT board with the [innova2_4gb_adit_xdma_ddr4_demo](https://github.com/mwrnd/innova2_4gb_adit_xdma_ddr4_demo) project when you have some spare time.

Using the DIY SFP Loopback and the [xcku15p_ffve1517_GTY_IBERT](https://github.com/mwrnd/innova2_experiments/tree/main/xcku15p_ffve1517_GTY_IBERT) project I was able to determine that the second SFP Port on the MNV303212A-ADIT is connected to Quad 129 Channel X0Y10. ...

Someone must have run [mlxup](https://network.nvidia.com/support/firmware/mlxup-mft/) [4.15.2](http://www.mellanox.com/downloads/firmware/mlxup/4.15.2/SFX/linux_x64/mlxup) on the card. **16.28.2006** is the latest firmware version. Refer to [Upgrading the ConnectX-5 Firmware](https://github.com/mwrnd/innova2_flex_xcku15p_notes#upgrading-the-connectx5-firmware). 

This is a valid issue I have been exploring as well. The Innova-2 shows up under Linux as being composed of a **bridge** connecting two ethernet controllers and the FPGA...

I have asked about [ConnectX-5 Ethernet to FPGA Direct Communication](https://forums.developer.nvidia.com/t/innova-2-flex-connectx-5-ethernet-to-fpga-direct-communication/236059) on the [Nvidia SoC and SmartNIC Forum](https://forums.developer.nvidia.com/c/infrastructure/software-and-drivers/soc-and-smartnic/530). I have also created GitHub Issues on the [flexdriver-iot-auth](https://github.com/acsl-technion/flexdriver-iot-auth/issues/1) and [flexdriver-zuc](https://github.com/acsl-technion/flexdriver-zuc/issues/1) projects.

Seems unlikely Nvidia will provide any support for this but I did some research and came across some hopeful notes. [FlexDriver](https://marksilberstein.com/wp-content/uploads/2021/11/flexdriver-asplos22-slides.pdf) uses [PCIe Peer-to-Peer](https://www.ni.com/en-ca/support/documentation/supplemental/10/an-introduction-to-peer-to-peer-streaming.html):  PCIe Peer-to-Peer is [supported by...

[PCIe Peer-to-Peer DMA](https://www.kernel.org/doc/html/v5.8/driver-api/pci/p2pdma.html) should work on any motherboard with [Resizable BAR](https://xilinx.github.io/Vitis_Accel_Examples/2020.2/html/p2p_fpga2fpga.html) and [Above-4G Memory Decoding](https://github.com/Xilinx/XRT/blob/master/src/runtime_src/doc/toc/p2p.rst#bios-setup) [capabilities](https://www.nvidia.com/en-us/geforce/news/geforce-rtx-30-series-resizable-bar-support/). However, [it is supposedly buggy](https://docs.nvidia.com/cuda/gpudirect-rdma/index.html#supported-systems). [GPUDirect](https://developer.nvidia.com/gpudirect) is supported on [Keplar](https://en.wikipedia.org/wiki/Kepler_(microarchitecture)) or later [Tesla and...

I have made progress figuring out what **doesn't** work. I got a [Tesla K80](https://www.nvidia.com/en-gb/data-center/tesla-k80/) GPU ([SM_3.7](https://developer.nvidia.com/cuda-gpus)) and have been trying to demonstrate [P2PDMA](https://www.kernel.org/doc/html/latest/driver-api/pci/p2pdma.html). The K80 is [no longer under active...