BRN Discussion Ongoing
- Thread starter TechGirl
- Start date
Bravo
If ARM was an arm, BRN would be its biceps💪!
A competitor's product with analog neurons.
Perhaps someone could make a comparison to Akida.
#analog | Neuromorphicism
Neuronova is ready to launch their first Evaluation Kit Silk 1 in the following months! Their H1 Neuromorphic Processor has programmable #analog neurons and synapses with a configurable analog front-end. Give them a follow to stay updated!www.linkedin.com
ChatGPT's Response
Here’s a concise, apples-to-apples comparison of Neuronova’s H1 vs BrainChip’s Akida (Gen-2) based on what’s publicly available.
What each one is (at a glance)
- Neuronova H1 — a fully-analog neuromorphic processor for spiking neural networks (SNNs) aimed at ultra-low-power edge sensing; the company positions it as “doing more with less,” with a roadmap toward sensor-in-package integration. Public materials emphasise analog in-memory/time-domain compute and extremely low power (trade press cites <1 µW for always-on sensing use cases). NeuromorphicCore.ai+3Neuronova+3Tech.eu+3
- BrainChip Akida (Gen-2) — a fully-digital neuromorphic processor IP (licensable block or FPGA dev platform) designed for event-based inference and on-chip/edge learning; supports CNN→SNN conversion and BrainChip’s TENN (Temporal Event-based Neural Networks) models, with model cards and documented HW limits. BrainChip+3BrainChip+3doc.brainchipinc.com+3
Side-by-side
| Dimension | Neuronova H1 | BrainChip Akida Gen-2 |
|---|---|---|
| Core compute style | Fully analog SNN compute at the sensor/edge | Fully digital event-domain neural processor IP |
| Target use | Ultra-low-power always-on sensing; roadmap to sensor-level integration | Edge AI inference and on-device learning (incremental/edge learning) for wearables, IoT, automotive, space, etc. |
| Power narrative | “Do more with less”; reporting around sub-µW always-on claims in media profiles (for tiny tasks) | “Millwatts-class” continuous AI; event-sparsity to cut energy; supports edge learning without cloud round-trips Tech.eu+2PoliHub+2 |
| Sensors | Messaging about “towards full sensor integration” (front-ends that adapt to drift/non-ideality) | Works with conventional frame sensors and event cameras; model zoo & conversion tools; not sensor-hardwired Neuronova+1 |
| Software/tooling | Public details limited (no open model zoo/docs on H1 at time of writing) | Published product brief, capability limits, examples docs, and Akida 2 model cards (incl. 8-bit / 4-bit quantisation) BrainChip+3BrainChip+3doc.brainchipinc.com+3 |
| Learning | (Not detailed publicly) | Supports on-chip/edge learning; TENNs for temporal efficiency; incremental learning flows BrainChip+1 |
| Business model | Early-stage startup; pre-seed funded, building chips | IP/licensing + dev platforms; 2nd-gen IP available (docs, model zoo) PoliHub+1 |
BrainChip – partners & collaborators
- Arm – official Arm Partner Catalog + Arm AI Partner Program (incl. Keil device packs/tools). Arm+2investor.brainchip.com+2
- Prophesee, Socionext, VVDN – joint demos/solutions at CES and other events. investor.brainchip.com
- RTX / Raytheon – teaming on an AFRL neuromorphic radar contract. BrainChip
- Frontgrade Gaisler – licensed Akida IP for space-grade SoC. Edge AI and Vision Alliance
- ARQUIMEA – water safety/search-and-rescue detection solution. BrainChip
- DigiKey – channel/distribution for Akida boards & dev kits. investor.brainchip.com
- Eastronics (Israel), SalesLink (Europe) – regional sales partners. BrainChip
- Intel (IFS) — Member of the Intel Foundry Services (IFS) Accelerator / IP Alliance. BrainChip+2EENews Europe+2
- SiFive — Joint edge-AI platform (Akida + RISC-V). BrainChip+1
- Edge Impulse — Platform integration / ecosystem partner. BrainChip+2edgeimpulse.com+2
- emotion3D — In-cabin automotive safety UX with Akida. BrainChip+2Electronicsclap+2
- Frontgrade Gaisler — Licenses Akida IP for space-grade SoCs. BrainChip+1
- Bascom Hunter — Defense/Intel integrations noted in ASX materials. ASX Announcements
- Onsor Technologies — Wearable seizure-prediction glasses on Akida. BrainChip+2FT Markets+2
- Chelpis / Mirle Group — Security & robotics collaboration (Akida AKD1000/M.2). BrainChip+2TechBullion+2
- Quantum Ventura — Akida used in CNRT cyber-threat tool (project partners include Lockheed Martin MFC & Penn State). BrainChip+1
- Valeo — Joint development agreement for automotive AD/ADAS use. Small Caps
Neuronova – partners & collaborators
- On Neuronova’s site and recent profiles/funding notes, no Arm (or similar large-ecosystem) partnership is listed publicly. Pages focus on the H1 analog neuromorphic chip and funding news; no partner roster is shown.
Practical takeaways
- If you need proofable tools, docs, and near-term deployability: Akida Gen-2 is more mature and documented today (model zoo, conversion pipelines, HW limits), with a defined IP path for partners. BrainChip+2BrainChip+2
- If your top priority is extreme standby/always-on power at the sensor itself: Neuronova’s H1 vision (analog SNN + sensor-in-package) is compelling for microwatt-class sentry tasks, but public specs are thin; you’ll likely need NDAs/eval hardware to validate claims and fit. Neuronova+2Tech.eu+2
- Architectural trade-off: analog neuromorphics can be exceptionally efficient but may face calibration/variability and tooling challenges; digital neuromorphics (Akida) are somewhat higher power but typically easier to integrate, reproduce, and scale across foundries and toolchains. (This analog-vs-digital tooling friction is a theme in SNN literature.) MDPI
Where the info is thin (so you know what to ask vendors)
- Neuronova H1: exact MAC/synapse counts, memory topology, input modalities, supported learning rules, compiler/SDK maturity, I/O (SPI/PCIe), latency/accuracy vs standard benchmarks — not detailed on the public site. Neuronova
- Akida Gen-2: plenty of docs exist, but real-world PPA depends on the mapped model and operating point; check the model cards and HW-limit page for your network shapes and quantisation plans. BrainChip+1
Bottom line
- Akida Gen-2 is the safer pick if you want a documented, digital, licensable neuromorphic platform with existing tools and examples today. BrainChip+1
- Neuronova H1 looks aimed at ultra-low-power sensing with a bold analog approach and sensor-integration roadmap, but you’ll need direct vendor engagement to validate specs and tooling readiness beyond the website/press. Neuronova+1
Is this a popular guy in here? Insulting people who are sceptical and are showing reasonable critism towards our beloved company painchip? No im not a shorter. Ive been invested here since 2020 and im tired of waiting. I allow myself to be sceptical since there is reason to it. It might not be good technical but keep showing your true colors. Clown
Frangipani
Top 20
Yet another BRN employee has left…
Our former ML Solutions Architect Dhvani Kothari is now an AI Solutions Architect with Contextual AI:
www.linkedin.com
Our former ML Solutions Architect Dhvani Kothari is now an AI Solutions Architect with Contextual AI:
#contextualai #genai #techcareer | Dhvani Kothari
🎉 Excited to share that I've joined Contextual AI as a Solutions Architect! 🚀 The innovation and vision here are inspiring, and I’m eager to contribute to what’s next in enterprise AI. I want to take a moment to express my heartfelt gratitude to everyone at BrainChip for an incredible...
That can be debated, he tries hard to be funny, often insulting people in the process.
Personally, I don't care, it's not his fault he is a clown and treat the forum like he is in the pub after work.
You have the right of being here just as everybody else, so just keep on posting and ignore the posters you don't like, you can put them on ignore and not see their posts anymore.
I think most here are tired of waiting, if they don't deliver before the next AGM, they are done for and can sell the company as far as I am concerned.
It's not acceptable they keep on printing shares for keeping the light on, hollowing out shareholders part of the company.
Sean said that he expect orders of 9 mill bucks this year, the window is closing, but I will give him time before I will get more harsh in my criticism.
Similar threads
