The NVIDIA DGX Spark and DGX Station are both designed for AI computing, but they differ significantly in terms of memory bandwidth and overall architecture.
The DGX Spark features a memory bandwidth of 273 GB/s, which is impressive for its compact form factor and is designed to support AI workloads efficiently. It uses the NVIDIA GB10 Grace Blackwell Superchip, which includes fifth-generation Tensor Cores and FP4 support, enhancing its performance for AI tasks. The DGX Spark is equipped with 128GB of unified LPDDR5x memory and supports advanced networking capabilities like ConnectX-7 for high-speed connectivity[1][3][7].
In contrast, the DGX Station offers much higher memory bandwidth capabilities, although specific figures for the latest model are not detailed in the provided sources. However, it is noted that the DGX Station uses high-bandwidth memory (HBM3E) and can support up to 288 GB of HBM3E and 496 GB of LPDDR5x, potentially achieving memory bandwidths significantly higher than the DGX Spark, especially when considering its use of NVLink-C2C interconnect technology for enhanced CPU-GPU communication[4][5]. The DGX Station is built with the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, featuring a massive 784 GB of coherent memory space, which is designed to accelerate large-scale AI training and inferencing workloads[5].
Overall, while the DGX Spark provides a compact and efficient solution for AI development with a memory bandwidth of 273 GB/s, the DGX Station is designed for more demanding AI tasks, offering superior performance and higher memory bandwidth capabilities due to its advanced architecture and larger memory capacity.
Citations:
[1] https://www.pcmag.com/news/what-is-nvidias-dgx-station-a-new-specialized-desktop-line-for-ai-work
[2] https://www.reddit.com/r/LocalLLaMA/comments/1jedy17/nvidia_digits_specs_released_and_renamed_to_dgx/
[3] https://www.ainvest.com/news/nvidia-unveils-dgx-spark-dgx-station-revolutionizing-personal-ai-computing-2503/
[4] https://www.youtube.com/watch?v=krBh0Von-2A
[5] https://www.techpowerup.com/forums/threads/nvidia-announces-dgx-spark-and-dgx-station-personal-ai-computers.334300/
[6] https://www.reddit.com/r/LocalLLaMA/comments/1jef1dd/dgx_spark_previously_digits_has_273gbs_memory/
[7] https://www.nvidia.com/en-us/products/workstations/dgx-spark/
[8] https://xenon.com.au/product/nvidia-dgx-station/
[9] https://images.nvidia.com/content/newsletters/email/pdf/DGX-Station-WP.pdf