NVIDIA · Local AI machineOlder
NVIDIA Jetson AGX Orin 64GB (64 GB)
NVIDIA machine: NVIDIA Jetson AGX Orin 64GB (Ampere GPU, 12-core Arm CPU), 64 GB memory, 204.8 GB/s memory speed, about 3.1 tok/s on a 70B model (estimate).
78.9Score
- Not ranked yet
Why this score
The score adds up these parts. Each part has a weight. A part counts as the average of its sources, pulled toward 80 when there are few sources.
AI power
50% weightcounts as 77.8Average of 2 sources: 76.7
Experts
30% weightcounts as 80.0No sources found yet. No data yet. This part counts as 80.
Users
20% weightcounts as 80.0No sources found yet. No data yet. This part counts as 80.
No reviews or buyer ratings yet: the score comes from the specs only.
80 is the field average. A part counts as 80 + (average − 80) × N ÷ (N + 1), where N is its number of sources. One source counts half of the gap from 80. No sources count as 80.
Price
€4748.99 at getgoods.com (via Geizhals.at) · Geizhals.at listing, not checked live
Developer kit 945-13730-0055-000 (64 GB), in stock, free shipping. Conrad.at and galaxus.at EUR 4,749.00; reichelt.at EUR 4,760.55 plus EUR 6.95 shipping.
Prices checked 8 Oct 2026.
Specs
- Chip
- NVIDIA Jetson AGX Orin 64GB (Ampere GPU, 12-core Arm CPU)
- CPU
- 12 cores / 12 threads (Arm Cortex-A78AE)
- Memory
- 64 GB
- GPU memory
- 64 GBUnified 64 GB shared by CPU and GPU on Jetson Linux, per NVIDIA's Jetson Orin page.
- Memory speed
- 204.8 GB/s
- FP16 TFLOPS
- 43 TFLOPS
- NPU
- —
- OS
- Linux
- Software
- CUDA
- Ethernet
- 10 Gbit/s
- Thunderbolt / USB4
- —
- QSFP port
- No
- Idle power
- —
- Load power
- 60 W
- Noise
- —
- Volume
- 0.9 L
- Weight
- —
- SSD
- 64 GB
- Release
- 2022-03
Measured
- Qwen3-8B, INT4 AWQ
- 32.2 tok/s writing, 819.8 tok/s prompt · TensorRT Edge-LLM v0.10.0, batch 1 (NVIDIA)
Notes
Previous-generation robotics kit with 64 GB unified memory at 204.8 GB/s and Ampere CUDA cores: it loads 30B-class quantised models but is slow (Qwen3-8B INT4 about 32 tok/s) and a 70B dense model barely fits. Still on sale at about EUR 4,750 in Austria, far above its USD 1,999 launch price.
NVIDIA's Orin page says 15-60 W in the text but 15-75 W in the 64 GB module column. The Orin page's columns were hard to map, so the 43 TFLOPS FP16 figure was checked against 16 SMs x 1.3 GHz (about 42.6). NVIDIA does not say whether the Orin kit is being phased out; Austrian shops still list it in stock. No expert or buyer score found; reviews of the kit that showed up in search (JetsonHacks, Hackster, Engadget, all 2022 launch pieces) were not opened.
More reading:jetsonhacks.comforums.developer.nvidia.com