Unanswered questions

I am trying to get the Axelera Metis M.2 module working on a NanoPC-T6 Plus board powered by the Rockchip RK3588 SoC running Ubuntu 22.04 (Linux kernel 5.10.0-1012-rockchip / ARM64).(I installed image file in this link https://support.axelera.ai/hc/en-us/articles/31859388491794-Bring-up-Voyager-SDK-in-NanoPC-T6?_gl=1*3zteth*_gcl_au*MTc5ODEwMTM1MC4xNzg3NzI0Mjgx because I have tested other OS images provided on the official NanoPC-T6 Plus wiki, but the issue persists, indicating a known PCIe memory allocation bug within the Ubuntu 22.04 kernel for this platform.)However, the driver fails to probe the device due to PCI resource assignment failure (ENOMEM / error -12).1. Device Identification & lspci StatusThe PCIe device is visible on the bus (0000:01:00.0), but the Linux kernel fails to assign memory resources (MMIO) to Region 0 and Region 2:$ lspci -vv -s 0000:01:00.0 | grep -iE 'Region|Memory|flags' Region 0: Memory at <ignored> (64-bit, non-prefetchable) [disabled] Region 2:

Dear Axelera Support,I am currently following the Voyager SDK quick start guide (YOLO inference demo) on an RK3588-based Compute Board (received out-of-the-box), using Voyager SDK version 1.5.3. However, I am unable to successfully run inference due to persistent device communication issues with the Metis AIPU.--------------------------------------------------System Overview- Platform: RK3588 (Compute Board, out-of-the-box setup)- Voyager SDK: 1.5.3 (and also tested with 1.4.0)- Connection: PCIe (Metis AIPU detected as 01:00.0 [1f9d:1100])--------------------------------------------------Initial IssueRunning the quick start inference:./inference.py yolov5s-v7-coco usb:/dev/video1Result:ERROR: No devices foundAt that moment:- lspci returned no devices, indicating a PCIe enumeration issue.--------------------------------------------------After Cold BootAfter performing a full power cycle:- lspci correctly shows:  01:00.0 Processing accelerators: Axelera AI Metis AIPU (rev 02) [1f9d:1100]

Are there Axelera powered development boards similar to raspberry pi or like a nvidia jetson ?.Has axelera considered making one.

Hi team,We are currently deploying the Axelera Metis PCIe 4-AIPU AI Accelerator Card (AXE-BME20P4) in an HPE ProLiant DL380 Gen11 server as part of a k3s-based AI inference cluster.The card is physically seated in a PCIe Gen 3.0 x16 slot and Ubuntu Server 24.04 LTS is installed. However, we are facing a challenge with the auxiliary power connection.The Metis card requires a 2x4 (8-pin) PCIe auxiliary connector, but the HPE DL380 Gen11 uses a proprietary internal power header on the riser side. We have identified HPE part P77896-B21 (DL380 Gen11 2U 8-pin GPU Power Cable Kit) as the likely adapter cable needed.Before we proceed with the order, we wanted to check with you directly:1. Have you or any of your customers successfully deployed the Metis PCIe 4-AIPU in an HPE ProLiant DL380 Gen11?2. Can you confirm that HPE P77896-B21 is the correct cable to connect the Metis card's 2x4 auxiliary power connector to the DL380 Gen11's internal power distribution?3. Are there any other HPE DL380 G

I recently acquired a Metis M.2 module and was so excited to get started with it. To my horror, I installed it into a standard, modern x86 mini PC (with power disconnected, all by the books), and the system was bricked upon my first attempt to boot. At first, I thought maybe the M.2 slot on the system was not providing adequate power for the device, and to the benefit of the Metis, I decided to try another host system. HUGE mistake. The same thing happened with a LattePanda x86 SBC, confirmed to have enough power to the M-key M.2 slot. These systems are not cheap, so to say I am disappointed is an incredible understatement.  Maybe I’m just not understanding where to look for host system specs, but it doesn’t seem like this issue is warned against anywhere in your docs or on the community page here. Has anyone else had this problem? I’m down two systems, so if someone could please tell me with certainty which system I CAN reliably plug this thing into, I would be very appreciative. EDIT

We want to connect multiple camera to detect object and track it. Each camera at different location. We could do that with Yolo. In Axelera, we can do it with 1 camera. When 2nd camera is started, error is thrown. Please advice 

I’ve been running through the installation of the Voyager SDK v1.7 and I seem to be getting stuck on the operators build step.axdevice returns “Device 0: metis-0:3:0 8GiB metis-std-temp m2-max flver=1.6.0 bcver=7.4 clock=800MHz mvm=100% power-limit=8.5W”However when I try to run `make operators` I get the following error:FAILED: axstreamer/CMakeFiles/axstreamer.dir/src/AxInferenceAxRuntime.cpp.o /usr/bin/c++ -DHAVE_LIBVA -DHAVE_OPENCL -Daxstreamer_EXPORTS -I/usr/include/opencv4 -I/home/ubuntu/axelera/axelera-py/lib/python3.12/site-packages/axelera/lib/pkgconfig/../../include/include -I/home/ubuntu/axelera/voyager-sdk/operators/axstreamer/include -I/home/ubuntu/axelera/axelera-py/lib/python3.12/site-packages/axelera/lib/pkgconfig/../../include -O3 -DNDEBUG -std=gnu++20 -fPIC -Wall -Werror -Wno-sign-compare -Wno-deprecated-enum-enum-conversion -Wno-deprecated-anon-enum-enum-conversion -Wno-return-type-c-linkage -Wno-pessimizing-move -Wno-maybe-uninitialized -Wno-error=class-memaccess -MD

On a x64 based system with a Metis M.2, i am trying to do a firmware update for the device but getting an error.(axelera-env) devuser@linuxdev:~/voyager-sdk$ axdevice interactive_flash_updateWARNING: AXELERA_DEVICE_DIR environment variable not set.Using AXELERA_DEVICE_DIR=/home/devuser/voyager-sdk/axelera-env/lib/python3.12/site-packages/axelera/omegaWARNING: AIPU_RUNTIME_STAGE0_OMEGA environment variable not set.Using AIPU_RUNTIME_STAGE0_OMEGA=/home/devuser/voyager-sdk/axelera-env/lib/python3.12/site-packages/axelera/omega/bin/start_axelera_runtime_stage0.bin>>> Metis Flash Update Script>>> =========================>>> Auto-detecting and updating all connected devices...>>> Detected 1 device(s) from axdevice>>> ==============================================>>> Found 1 device(s):>>> - metis-0:81:0>>> ==============================================>>> Device metis-0:81:0 needs firmware update (current: 1.2

 The Binary Solar Matrix and Chromatic Wave Nullification: A Unified Framework for Positive and Negative RealitiesAuthor: Gautam Pal*Affiliation: Independent Researcher, Nadia, West Bengal, India*Corresponding Author Email: Palgour438@gmail.comORCID: 0009-0004-3456-9972AbstractThis treatise proposes a groundbreaking cosmological paradigm that fundamentally redefines the ontological relationship between luminescence, primordial darkness, and the topological architecture of the universe. By examining the fluid-dynamic analogs of solar wave mechanics, we establish that darkness is not an absolute ontological void, but rather a highly ordered continuum of chromatic nullification governed by destructive phase interference. Extrapolating this localized optical mechanic to a macro-cosmological scale, we posit the existence of a Binary Solar Matrix, comprising the observable Positive Sun and its unmanifested inverse, the Negative Sun. This requisite duality dictates a state of symmetric equili

Hello, i just got my m.2 card and on my AMD64 desktop workstation, the Metis M.2 AIPU does not survive an S3(suspend-to-RAM) cycle. After resume, every large DMA transfer to the cardfails permanently. Only a full cold power-cycle restores function — a warmreboot is not enough. This makes the card unusable on any normal desktop,where suspend/resume is the standard daily workflow.I narrowed this down together with Claude Opus 4.8 (Anthropic) over a detaileddebugging session. Full diagnosis below.## Hardware / Software- Host: AMD Ryzen, B850 platform, RTX 3090 in primary slot- OS: CachyOS (Arch-based), kernel 7.0.12 (Clang/LLVM-built), Secure Boot off- Metis M.2 AIPU, PCIe 0000:0c:00.0 (1f9d:1100 rev 02), Gen3 x4- Driver: axelera-driver release/v1.6 (module version 1.4.17), built with  make LLVM=1, DKMS-managed- SDK: Voyager 1.6.1 (axelera-rt / devkit), inference via distrobox/Ubuntu 24.04## Reproduction1. Cold boot. Run inference (yolo26s-coco-onnx, fakevideo) → works, ~529 fps.2. system

I work in edge AI at Advantech, and I've already worked with Metis cards through that role — so I know what the silicon can actually do, and that's exactly why I want to push it further with this project.The problem: traffic cameras record everything and understand nothing. A camera watching an intersection can technically see a truck losing its tarp or a bike cutting into a lane it shouldn't, but nobody's watching live, and by the time footage gets reviewed the moment's gone.Haytham is the vision layer of a bigger physical-AI platform I'm building, and for this challenge I want to build and prove its core loop live on Metis: detection + tracking on a live feed, a small VLM reasoning over flagged events, and an agentic layer that turns what it sees into an actual citation — location, timestamp, plate read, violation description — instead of just a bounding box nobody looks at.Hardware: x86 platform + Axelera Metis M.2 module.What I want Wingman to build with me:Detection/tracking pipel

I am currently working with metis M2 card. Trying to run yolov8 model using inference.py file from the sdk. The throughput is very low. I set the AIPU core to 4 and verfied the same in the compile_config.json file. But there is another argument called multicore_mode("multicore_mode": "batch"). what exactly does the multi core mode do? and is there way to set it to other option than ‘batch’?

Project Idea: AI-Powered Multi-Camera Face Recognition & Smart Attendance SystemI want to build a real-time, edge-AI based face recognition and attendance system using Axelera AI Metis and Voyager Wingman.The main goal is to create a system that can process live RTSP streams from multiple CCTV/IP cameras, recognize registered people, track them across frames, and automatically maintain attendance records — while performing AI inference on the Axelera Metis accelerator instead of relying heavily on the host CPU.### What I want to buildThe system will have the following pipeline:RTSP CCTV Camera        ↓Video Processing        ↓Face/Person Detection        ↓Face Recognition        ↓Object/Face Tracking        ↓Identity Matching        ↓Attendance Recording        ↓Web DashboardThe system will initially support one camera and then be designed to scale to multiple RTSP cameras.### Key Features• Real-time face detection and recognition• Multiple registered people• Multiple reference ima

Hello everyone,I have been trying to compile the PRL Tracker model (see link below) for about a month using both Hailo and Vitis AI, but so far I have not been able to fully meet the requirements.I am now considering using the Metis Accelerator, but I am not sure whether I would run into the same issues.Has anyone here successfully deployed this model on the Metis Accelerator, or could you confirm if it is supported?Repository: https://github.com/vision4robotics/PRL-Track/tree/main Any guidance or shared experience would be greatly appreciated.Thanks in advance! 

Blitz chess can be extremely fast.  Each player typically only has a couple of seconds to think and then execute a move.  That speed makes it impossible for the players to log the moves.  Still, there are entire blitz chess tournaments of which you’d like to record hundreds of games. One way to avoid logging the moves is to use a “digital” chess board.  However, these are expensive and typically have a hard time keeping up with the speed at which the moves occur.  Especially, towards the end of a game, when the players are in time trouble, moves come thick and fast and players become sloppy, not placing their pieces neatly on the board (even knocking them over), the digital boards often misinterpret the moves made.I’d like to observe the chess board with a standard webcam and reconstruct the moves (and the move times) from the video feed.I believe this can be done as follows:Detect the board, the pieces and the clock using YOLO. Detect that the board has changed.  When it has changed,

I created a post, formatting is poor. I need to edit/update the formatting. Checked the site couldn’t find a way to do that. How do I edit a post?

I propose building an ultra-low-power, deterministic hardware accelerator engine for Edge AI using the Septenary Unitary Conversion (SUC) framework and Base-7 logic on the Metis PCIe platform. WHAT WE ARE BUILDING:1. SUC-RNS Core: A parallel combinational Adder-Tree architecture that converts raw data streams into 3-bit unitary vectors (Modulo-7 residues) without conventional long-form division or multipliers.2. Hardware Acceleration: Implementing an FPGA/RTL-level SUC module to reduce matrix multiplication latency, dynamic power consumption (up to 83.4% reduction), and carry-propagation delays in Edge AI inference pipelines.3. Voyager SDK & Wingman Integration: Mapping the SUC mathematical reduction layer into Metis hardware via Voyager Wingman for fast pipeline compilation and real-time validation. FEASIBILITY & DELIVERABLE:The mathematical foundation and synthesizable Verilog RTL code for 8-bit, 32-bit, and 64-bit buses are already fully designed, validated, and documented.

I see to many AI right now. What the golas of Axelera ai?

Developed for czela.net (a non-profit community ISP in the Czech Republic with ~400 active infrastructure devices and 2,000 member connections) this project bridges real-world physical network reality with NetBox SSoT.            Standard telemetry tools (SNMP, LLDP, FDB, OSPF) often fail due to incomplete data and unmanaged hardware cascades. Our hybrid agent combines standard sync code with a local quantized LLM. It generates practical          recommendations for volunteer network admins to fill documentation gaps, assists in local troubleshooting during outages, and preserves complete data privacy without cloud dependencies.Rather than replacing existing monitoring software, our project introduces a **Hybrid Auto-Discovery & Reliability Agent**. The core logic is **hardware-agnostic** and designed to run local LLMs (currently prototyped with Ollama/mistral-nemo). We are interested in evaluating **Axelera Metis** as a potential low-power edge accelerator for our network distribu