Ampere GPU, six-core Arm Cortex-A78AE CPU and LPDDR5 memory in a production SoM
Compute & Control · Embedded AI system-on-module
Jetson Orin Nano 8GB
An entry Orin module whose Super-mode figures require JetPack 6.2 and a compatible power and thermal configuration.
NVIDIA Jetson Orin Nano 8GB · Also: Jetson Orin Nano 8 GB
15 known suppliers in this product family.
Modules, single-board computers and rugged edge systems used for perception, autonomy or robot control workloads. The list is bounded to the current RCI manufacturer-directory universe and separates exact-model audited suppliers from confirmed product-line suppliers.
EXACT-MODEL AUDITED
9 suppliers
NVIDIA · Raspberry Pi · Qualcomm · Advantech · AAEON · Seeed Studio · Toradex · AMD · Intel
PRODUCT-LINE CONFIRMED
6 suppliers
ADLINK Technology · Kontron · DFI · Axiomtek · Vecow · OnLogic
This is a bounded discovery set, not a claim to enumerate every company worldwide. Open the full supplier landscape and official routes →
What this record can support today.
Coverage is computed from published, non-unknown fields. It measures the record—not product quality—and makes missing procurement evidence explicit.
Engineering field coverage70%
7 of 10 selected fields published
- Official product sourcepublished
- Lifecycle statepublished
- Mechanical dimensions / mass / mountingmissing
- Environmental operating limitsmissing
- Compute performance boundarypublished
- Memory and storagepublished
- Power mode / inputpublished
- I/O and expansionpublished
- Software / OS supportpublished
- Thermal design boundarymissing
Procurement field coverage62%
8 of 13 selected fields published
- Official product / sales routepublished
- Lifecycle statepublished
- Exact orderable SKU / part numberpublished
- Price with region, currency and quantitypublished
- Lead time with check datemissing
- Commercial availability / stock statepublished
- Minimum order quantitymissing
- Authorized distributor / sales channelpublished
- CAD / STEP / interface drawingmissing
- Certification with identifiermissing
- MTBF / failure / reliability datamissing
- Warranty termspublished
- PCN / EOL / replacement noticepublished
Coverage measures whether a selected public field is present with non-U evidence. It is not a product score, compatibility result, certification, availability confirmation or procurement recommendation. Download the complete readiness coverage JSON →
8 bounded physical-evidence records.
Publisher independence, identity precision and measurement scope vary by row. Results remain bounded by the disclosed specimen, setup, workload and metric; an inconclusive mapping is evidence associated with the catalog model, not verification of a manufacturer claim.
| Measured result | Conditions | Claim mapping | Source |
|---|---|---|---|
| Independent AI-benchmark energy resultindependent_energy_benchmark | 31.8 J CUDA; 28.3 J SYCL (12% lower)Jetson Orin Nano 8 GB board at 15 W with JetPack 5.1.2; tegrastats package measurement across the paper’s four HeCBench AI tests using the cited CUDA and SYCL implementations. | supportedThe catalog claim already carries a qualified external V source with its original condition and location. | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Table 1, Section 4 and Section 5.3: 8 GB board configuration and AI-benchmark energy results Checked 2026-08-10 |
| CUDA versus AdaptiveCpp Polybench performancepolybench_adaptivecpp | CUDA 1.17× faster on averageJetson Orin Nano 8 GB at 15 W; Polybench GPU subset; paper software environment and default parameters. | inconclusiveApplication runtime ratios between CUDA and AdaptiveCpp are not sparse or dense INT8 TOPS and were not measured under the cited JetPack 6.2 Super-mode claim boundary. | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Sections 4.1 and 5.1, Table 4 Checked 2026-08-24 |
| CUDA versus DPC++ Polybench performancepolybench_dpcpp | CUDA 1.22× faster on averageJetson Orin Nano 8 GB at 15 W; Polybench GPU subset; paper software environment and default parameters. | inconclusiveCUDA-versus-DPC++ workload speed is not an INT8 TOPS measurement and cannot confirm the 67 sparse / 33 dense TOPS claim. | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Sections 4.1 and 5.1, Table 4 Checked 2026-08-24 |
| Optical-flow GPU performance comparisonoptical_flow_device_gap | Overall difference 60.9% in favor of the Orin GPULK, HS and TV-L1 optical-flow algorithms across the paper’s selected datasets; compared with the UP Squared Pro 7000 Edge GPU. | inconclusiveRelative optical-flow runtime against a different edge GPU is workload evidence, not a measurement of the module’s sparse or dense INT8 TOPS. | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Section 5.2 and Table 6 Checked 2026-08-24 |
| Dense Embedding runtime and average powerdense_embedding_power | CUDA 3.11 s at 2.1 W; SYCL 3.24 s at 1.31 WHeCBench Dense Embedding workload measured with tegrastats on the 15 W board configuration. | inconclusiveA tegrastats workload observation on a board configured for 15 W does not validate the complete 15 W / 25 W / MAXN SUPER module-mode set or carrier-inclusive power boundary. | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Section 5.3 and Figure 4 Checked 2026-08-24 |
| Observed CPU architecture and executionobserved_cpu_architecture | Arm Cortex-A78AE CPU identified and used for AdaptiveCpp executionExact Jetson Orin Nano study board; the paper does not independently measure the six-core count or 1.7 GHz maximum clock. | inconclusiveThe paper identifies and executes workloads on the Cortex-A78AE CPU, but does not independently measure the six-core count or 1.7 GHz ceiling in the compound catalog claim. | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Sections 4.1 and 5.1 Checked 2026-08-24 |
| Observed 8 GB board identityobserved_memory_identity | Jetson Orin Nano 8 GB study boardExact 8 GB board identity used for the published benchmarks; memory bus width, LPDDR generation and 102 GB/s bandwidth were not independently measured. | inconclusiveThe exact 8 GB board identity is independently bound to the benchmark, but the paper does not test or fully report the 128-bit LPDDR5 and 102 GB/s memory specification. | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Section 4.1, Table 1 and experiment device identity Checked 2026-08-24 |
| Software-baseline reporting boundarysoftware_baseline_review | JetPack 6.2 Super-mode baseline not reportedThe paper reports compiler/runtime configuration for its 15 W benchmark but does not publish a JetPack 6.2 Super-mode test boundary. | inconclusiveThe external benchmark publishes compiler and 15 W configuration details but not a JetPack 6.2 Super-mode baseline, so it cannot resolve the catalog software requirement. | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Section 4.1 environment configuration Checked 2026-08-24 |
Commercial facts, with volatility attached.
These fields come from current manufacturer surfaces. They are evidence records—not a live quote, stock check or purchasing recommendation.
| Field | Published value | Condition / boundary | Volatility / review | Source |
|---|---|---|---|---|
| Production module part numberordering_code | 900-13767-0030-000 | Commercial Jetson Orin Nano 8GB production-module identifier; the Orin Nano Super Developer Kit is a separate product with regional 945-series SKUs. | medium90-day cycle Due 2026-11-19 | NVIDIA Jetson FAQ ↗Part-number and origin tables · Jetson Orin Nano 8GB Checked 2026-08-21 |
| Volume suggested priceprice | USD 399 at 1KU+ | NVIDIA volume suggested pricing at 1,000 units or more; not a single-unit quote, tax-inclusive price or distributor commitment. | high14-day cycle Due 2026-09-04 | NVIDIA Jetson FAQ ↗Jetson product pricing table Checked 2026-08-21 |
| Published availability horizonavailability | Available through January 2032 | Lifecycle commitment for the commercial module; it does not establish present distributor stock or lead time. | medium90-day cycle Due 2026-11-19 | NVIDIA Jetson Product Lifecycle ↗Jetson Modules · Commercial Module table Checked 2026-08-21 |
| Official channel routeauthorized_channels | NVIDIA distributor and e-tailer directory | NVIDIA directs module buyers to its regional distributor listing; channel inventory and authorization must be checked for the buyer region. | high14-day cycle Due 2026-09-04 | NVIDIA Jetson FAQ ↗Where can I buy Jetson products? Checked 2026-08-21 |
| Module warrantywarranty | 3 years | General NVIDIA warranty term for Jetson modules unless otherwise specified; distributor handling and exclusions remain separate. | medium90-day cycle Due 2026-11-19 | NVIDIA Jetson FAQ ↗Jetson product warranty table Checked 2026-08-21 |
| PCN / EOL policychange_control | JEDEC JESD-046 PCNs; minimum 8-month EOL notice | NVIDIA lifecycle policy for Jetson modules; no open product-specific EOL notice is implied. | low365-day cycle Due 2027-08-21 | NVIDIA Jetson Product Lifecycle ↗Product Change Notifications and Product End of Life Notifications Checked 2026-08-21 |
- No current stock, lead time, MOQ, CAD package, certificate identifier or exact-model MTBF was captured in this wave.
Reported values, conditions attached.
Evidence state describes the claim—not the prestige of the company or document.
| Claim | Reported value | Condition / boundary | State | Source |
|---|---|---|---|---|
| CPUcpu | 6-core Arm Cortex-A78AE up to 1.7 GHz | Super-mode column. | M · Manufacturer claimed | NVIDIA JetPack 6.2 Super Mode for Jetson Orin modules ↗Table 2, Jetson Orin Nano 8GB Super column Checked 2026-08-10 |
| AI performanceai_performance | 67 sparse / 33 dense INT8 TOPS | Super mode with JetPack 6.2; sparsity and precision conditions preserved. | M · Manufacturer claimed | NVIDIA JetPack 6.2 Super Mode for Jetson Orin modules ↗Table 2, Jetson Orin Nano 8GB Super column Checked 2026-08-10 |
| Memorymemory | 8 GB 128-bit LPDDR5; 102 GB/s | 8 GB module, Super-mode column. | M · Manufacturer claimed | NVIDIA JetPack 6.2 Super Mode for Jetson Orin modules ↗Table 2, Jetson Orin Nano 8GB Super column Checked 2026-08-10 |
| Module power modespower | 15 W / 25 W / MAXN SUPER | JetPack 6.2 Super-mode configuration; carrier and peripherals are additional loads. | M · Manufacturer claimed | NVIDIA JetPack 6.2 Super Mode for Jetson Orin modules ↗Table 2, Jetson Orin Nano 8GB Super column Checked 2026-08-10 |
| CSI camera inputcamera_io | Up to 4 cameras; 8 virtual channels over 8 MIPI CSI-2 lanes | Module capability; carrier routing determines exposed connectors. | M · Manufacturer claimed | Solving Entry-Level Edge AI Challenges with NVIDIA Jetson Orin Nano ↗Module architecture and camera-interface summary Checked 2026-08-10 |
| Super-mode software baselinesoftware | JetPack 6.2 | Required for the cited Super-mode performance table. | M · Manufacturer claimed | NVIDIA JetPack 6.2 Super Mode for Jetson Orin modules ↗Table 2, Jetson Orin Nano 8GB Super column Checked 2026-08-10 |
| Published availabilitylifecycle | Available through January 2032 | NVIDIA Jetson lifecycle table; not a warranty of distributor stock. | M · Manufacturer claimed | NVIDIA Jetson Product Lifecycle ↗Jetson Orin Nano 8GB availability table Checked 2026-08-10 |
| Independent AI-benchmark energy resultindependent_energy_benchmark | 31.8 J CUDA; 28.3 J SYCL (12% lower) | Jetson Orin Nano 8 GB board at 15 W with JetPack 5.1.2; tegrastats package measurement across the paper’s four HeCBench AI tests using the cited CUDA and SYCL implementations.The result is workload-, implementation- and software-version-specific; it is not a Super-mode TOPS validation or a whole-robot power measurement. | V · Independently verified | SYCL in the edge: performance and energy evaluation for heterogeneous acceleration ↗Table 1, Section 4 and Section 5.3: 8 GB board configuration and AI-benchmark energy results Checked 2026-08-10 |
Known limits
What this record does not prove.
- Sparse and dense INT8 TOPS are not workload throughput.
- Super mode requires the stated software and a carrier, power supply and thermal path that support it.
- Module camera capability is not the same as exposed carrier-board ports.