The power consumption of the NVIDIA DGX Spark and the RTX 4090 varies significantly due to their different design purposes and architectures.
DGX Spark Power Consumption
The DGX Spark is designed as a compact AI supercomputer, featuring the Grace Blackwell GPU with 5th-gen Tensor cores and 20 Arm CPU cores. It is optimized for AI development and inference, providing up to 1,000 TOPS of AI compute performance. The DGX Spark consumes a maximum of 170W of power, making it a power-efficient solution for AI tasks such as model inference and fine-tuning[1][2].
RTX 4090 Power Consumption
In contrast, the NVIDIA RTX 4090 is a high-end graphics card primarily used for gaming and graphics-intensive applications. It has a much higher power consumption, with an official TDP of 450W. However, in real-world gaming scenarios, the RTX 4090 typically consumes between 300W to 400W, with occasional spikes above 400W during overclocking or very demanding games[3][4][7]. In extreme cases, such as stress tests or VR gaming, it can reach power consumption levels above 500W**[4][8].
Comparison
- Purpose: The DGX Spark is tailored for AI tasks, focusing on efficiency and compactness, while the RTX 4090 is designed for high-performance gaming and graphics rendering.
- Power Efficiency: The DGX Spark is significantly more power-efficient for its intended use case, consuming less than half the power of the RTX 4090 under typical conditions.
- Performance: Both devices offer high performance in their respective domains, but the RTX 4090 is optimized for raw graphics processing power, whereas the DGX Spark excels in AI compute tasks.
Overall, while both devices are powerful in their respective fields, they cater to different needs and have distinct power consumption profiles reflecting their design goals.
Citations:
[1] https://www.maginative.com/article/nvidia-unveils-dgx-spark-and-dgx-station-desktop-ai-supercomputers-for-the-developer-masses/
[2] https://beebom.com/nvidia-project-digits-rebranded-to-dgx-spark-dgx-station-announced/
[3] https://www.chillblast.com/blog/what-power-supply-do-you-need-for-the-nvidia-rtx-4090
[4] https://hwbusters.com/gpu/nvidia-rtx-4090-detailed-power-analysis-ideal-power-supply/
[5] https://www.pcmag.com/news/what-is-nvidias-dgx-station-a-new-specialized-desktop-line-for-ai-work
[6] https://www.youtube.com/watch?v=LZ25Z_un4bQ
[7] https://lambdalabs.com/blog/nvidia-rtx-4090-vs-rtx-3090-deep-learning-benchmark
[8] https://www.reddit.com/r/nvidia/comments/ysc7xy/rtx_4090_power_consumption/
[9] https://www.reddit.com/r/Amd/comments/13ljfd7/rtx_4090_vs_rx_7900_xtx_power_scaling_from_275w/