Jetson AGX Orin Industrial H.264 (AVC) decode capacity
Dataset 2026-09-07 · verified 2026-09-07 · last updated September 2026
NVIDIA rates the Jetson AGX Orin Industrial NVDEC at 1 × 4K60 · 3 × 4K30 · 6 × 1080p60 · 13 × 1080p30 H.264 (890 MP/s), in every power mode.
Verbatim datasheet figures with their source. Anything not in the table below is an EdgeAIStack estimate and is labelled as such.
Published decode points (class A)
Jetson AGX Orin Industrial: 2 NVDEC instances, H.264 (AVC) profiles Baseline, Main, High, 8-bit 4:2:0 unless the datasheet says otherwise. Stream counts are per module.
| Profile | Pixels @ fps | Streams (module) | MP/s per stream | Streams × MP/s | Evidence |
|---|---|---|---|---|---|
| 4K60 | 3840×2160 @ 60 | 1 | 497.7 | 498 | A datasheet |
| 4K30 | 3840×2160 @ 30 | 3 | 248.8 | 746 | A datasheet |
| 1080p60 | 1920×1080 @ 60 | 6 | 124.4 | 746 | A datasheet |
| 1080p30 | 1920×1080 @ 30 | 13 | 62.2 | 809 | A datasheet |
| Rated throughput (up to) | 890 MP/s | A datasheet | |||
Source
- NVIDIA · DS-10662-001 v1.8 §4.5 Multi-Standard Video Decoder, Table 4-1 (JAOi) · verified 2026-09-07 · Jetson AGX Orin Series Modules Data Sheet · class A
“H.264 Baseline, Main, High 4K60 (1) | 4K30 (3) | 1080p60 (6) | 1080p30 (13) 890 MPix/s 230”
v1.8 dated September 2025. Separate tables per SKU (32GB, 64GB, Industrial).
Between the published points (class D)
Camera profiles NVIDIA does not print are estimated by comparing the pixel rate against the rated 890 MP/s: max streams = ⌊890 ÷ (width × height × fps ÷ 10⁶)⌋. These are EdgeAIStack numbers, not NVIDIA's; they assume roughly linear NVDEC throughput inside the published range and are not asserted outside it. Rows that coincide with a published point show the datasheet count instead.
| Profile | Pixels @ fps | MP/s per stream | Max streams (est.) | Evidence | Full pipeline |
|---|---|---|---|---|---|
| 720p30 | 1280×720 @ 30 | 27.6 | 32 | D pixel-rate | |
| 1080p15 | 1920×1080 @ 15 | 31.1 | 28 | D pixel-rate | |
| 1080p20 | 1920×1080 @ 20 | 41.5 | 21 | D pixel-rate | |
| 1080p25 | 1920×1080 @ 25 | 51.8 | 17 | D pixel-rate | |
| 1440p15 | 2560×1440 @ 15 | 55.3 | 16 | D pixel-rate | |
| 1440p20 | 2560×1440 @ 20 | 73.7 | 12 | D pixel-rate | |
| 1440p30 | 2560×1440 @ 30 | 110.6 | 8 | D pixel-rate | |
| 4K15 | 3840×2160 @ 15 | 124.4 | 7 | D pixel-rate | |
| 4K20 | 3840×2160 @ 20 | 165.9 | 5 | D pixel-rate | |
| 4K25 | 3840×2160 @ 25 | 207.4 | 4 | D pixel-rate |
Power mode
NVDEC/NVENC maximum clock is the same in every nvpmodel power preset on Orin, so decode capacity does not change with the power mode.
Source
- NVIDIA · forums.developer.nvidia.com/t/380328 2026-08-17 · reply by NVIDIA staff (DaneLLL) · verified 2026-09-07 · NVIDIA Developer Forums — "Orin NX 16GB — NVDEC capacity for 4× 2560×1440 H.265 streams, and Super mode impact" · class A
“Maximum frequency of NVDEC is same in different power modes.”
H.264 encode (NVENC)
Encoder stream counts for Jetson AGX Orin Industrial at all nvpmodel presets, profile UHP. Needed only when you re-encode recordings or restream; passthrough recording uses no NVENC.
| Profile | Pixels @ fps | Streams (module) | MP/s per stream | Streams × MP/s | Evidence |
|---|---|---|---|---|---|
| 4K60 | 3840×2160 @ 60 | 1 | 497.7 | 498 | A datasheet |
| 4K30 | 3840×2160 @ 30 | 3 | 248.8 | 746 | A datasheet |
| 1080p60 | 1920×1080 @ 60 | 7 | 124.4 | 871 | A datasheet |
| 1080p30 | 1920×1080 @ 30 | 14 | 62.2 | 871 | A datasheet |
| Rated throughput (up to) | 895 MP/s | A datasheet | |||
Encode source
- NVIDIA · DS-10662-001 v1.8 §4.5 Multi-Standard Video Encoder, Table 4-4 (JAOi) · verified 2026-09-07 · Jetson AGX Orin Series Modules Data Sheet · class A
“H.264 UHP 4K60 (1) | 4K30 (3) | 1080p60 (7) | 1080p30 (14) 895 MPix/s 155”
What these numbers do and do not cover
- Hardware limit, not software. JetPack, DeepStream, GStreamer or Frigate may cap concurrent decoder sessions below the datasheet count.
- Decode in isolation. Simultaneous NVENC use shares memory bandwidth; expect real decode + encode totals below the sum of the two tables.
- Profile and chroma. Counts are for Baseline, Main, High at 8-bit 4:2:0; 4:4:4 halves throughput (datasheet note); Main10 shares the row.
- Decode is rarely the bottleneck. Inference, memory and preprocessing usually fail first on a detection pipeline; run the full pipeline check for your workload.
- Method. Formulas, evidence classes and the confidence rule are on the methodology page.
Other codecs on this module
H.264 on other modules
Is decode really the limit?
Run Jetson AGX Orin Industrial with your camera count, model and frame rate: every stage as a constraint, every number with its source, a permanent link.