I remember reading somewhere that the top slot doesn't support booting from NVMe. If it's non-boot then it shouldn't matter.

@c0rnelius Thanks for your input. The SSD booting was not the only issue on my system. For you, does booting https://sven-ola.commando.de/privat-in/Armbian-unofficial_26.05.0-trunk_Orangepirv2_trixie_edge_7.1.0-rc3_minimal.img from a MicroSD card work? With or without any hardware in any M.2 slot. When I tested, it did not work. However I think the SSD was connected when I tried. I'd need to try again with it disconnected. As there is only one NVMe SSD on my RV2, it should not matter in which M.2 slot my SSD is? From my boot log I can't see if the SSD is on "Bus0" "Bus2", is there some u-boot command to tell? Regarding making a MicroSD as boot disk, do you have any thought how do it. With hours of testing my way I could probably figure it out, any pointer to speed up the process would be much appreciated.

Patched the .dtb for v6.1, currently not yet in final format. I will probably create a commit to Armbian with a .dts for our board. I am attaching the dtb/dts just to keep track of the progress. Here are results of the RKNN test launch: rknn_api/rknnrt version: 2.3.2 (429f97ae6b@2025-04-09T09:09:27), driver version: 0.9.8 model input num: 1, output num: 1 input tensors: index=0, name=input, n_dims=4, dims=[1, 224, 224, 3], n_elems=150528, size=150528, fmt=NHWC, type=INT8, qnt_type=AFFINE, zp=0, scale=0.007812 output tensors: index=0, name=MobilenetV1/Predictions/Reshape_1, n_dims=2, dims=[1, 1001, 0, 0], n_elems=1001, size=2002, fmt=UNDEFINED, type=FP16, qnt_type=AFFINE, zp=0, scale=1.000000 custom string: Begin perf ... 0: Elapse Time = 8.06ms, FPS = 124.12 ---- Top5 ---- 0.935059 - 156 0.057037 - 155 0.003881 - 205 0.003119 - 284 0.000172 - 285 So: CPU: OK GPU: OK via Panfrost NPU: OK, RKNN inference works on RKNPU rk3566-box-X88PRO20-npu.dtb rk3566-box-X88PRO20-npu.dts

Can you give more details about the limitations of this NPU? I thought that they would be using shared system RAM. because all models are multiple megabytes, Can you comment about the ability to convert any model to your desired NPU target with "AI model conversion" https://tinycomputers.io/posts/rockchip-rk3588-npu-benchmarks.html (search for "Conversion Performance (x86_64)") Maybe its possible to run LCM-Lora in an armbian device, if it has 16GB. It is already running in a 16GB laptop without NPU or GPU. https://forums.raspberrypi.com/viewtopic.php?t=398225&sid=2b66e2cebe37506dd74215ead3dd264c I am thinking about buying an Orange PI zero 3W (allwinner NPU), but if you have troubles with a 6 TOPs NPU, I will have more troubles with a 2 TOPs NPU

Refer to the experiences here: I think someone wrote that the mainline kernel has NPU support. Is this NPU going to load only models that are in the Rockchip model repository?