The power consumption of the NVIDIA GB10 Superchip and the NVIDIA GB300 differ significantly due to their distinct designs and applications.
NVIDIA GB10 Superchip
The GB10 Superchip is part of NVIDIA's Project DIGITS, designed to provide powerful AI performance on a desktop scale. It features an NVIDIA Blackwell GPU with the latest CUDA cores and fifth-generation Tensor Cores, connected to a high-performance NVIDIA Grace CPU via NVLink-C2C interconnect. The GB10 Superchip delivers up to 1 petaflop of AI performance at FP4 precision and includes 128GB of unified, coherent memory and up to 4TB of NVMe storage. Notably, it operates using a standard electrical outlet, indicating relatively low power consumption compared to high-performance data center solutions like the GB300[2][5][8].NVIDIA GB300
The NVIDIA GB300, on the other hand, is a high-performance AI GPU designed for data centers and hyperscale environments. It is powered by the B300 chip and features 288GB of HBM3E memory, offering a significant increase in both memory capacity and computational power compared to its predecessors. The GB300 has a thermal design power (TDP) of 1400 watts, which is a substantial increase over the GB200, reflecting its enhanced performance capabilities for AI and high-performance computing workloads[1][3][4].Comparison
In terms of power consumption, the GB10 Superchip is designed to be power-efficient, running off a standard electrical outlet, which typically provides up to 1500 watts in many regions. This suggests that the GB10's power consumption is significantly lower than that of the GB300, which requires 1400 watts alone for its GPU. The GB300's high power requirement is due to its advanced architecture and increased performance capabilities, making it suitable for large-scale AI computing environments. In contrast, the GB10 is optimized for desktop use, providing a balance between performance and power efficiency for AI development and prototyping[2][5][8].Overall, while the GB10 Superchip offers impressive AI performance for its size and power requirements, the NVIDIA GB300 is designed for much more demanding AI workloads in data centers, necessitating higher power consumption to achieve its superior performance.
Citations:
[1] https://www.tweaktown.com/news/103991/nvidia-gb300-blackwell-ultra-ai-gpu-288gb-hbm3e-1-4kw-power-50-faster-than-gb200/index.html
[2] https://www.stocktitan.net/news/NVDA/nvidia-puts-grace-blackwell-on-every-desk-and-at-every-ai-developer-jpbbvfo6wno4.html
[3] https://www.techpowerup.com/330154/nvidia-gb300-blackwell-ultra-will-feature-288-gb-hbm3e-memory-1400-w-tdp
[4] https://drrobertcastellano.substack.com/p/nvidia-gb300-redefining-ai-computing
[5] https://www.bigdatawire.com/2025/01/10/inside-nvidias-new-desktop-ai-box-project-digits/
[6] https://www.nvidia.com/en-us/data-center/dgx-b300/
[7] https://siliconangle.com/2024/12/26/leaks-reveal-beefed-specifications-nvidias-next-gen-gb300-ai-server/
[8] https://nvidianews.nvidia.com/news/nvidia-puts-grace-blackwell-on-every-desk-and-at-every-ai-developers-fingertips
[9] https://meta-quantum.today/?p=3460