Winning Reason
1. NVIDIA DGX Spark brings powerful 1000 AI TOPS supercomputing capabilities to the user's desktop, allowing developers to easily handle large AI models without the need for a data center.
2. Featuring 128 GB of unified system memory, it enables developers to manage up to 200 billion parameters, fine-tune 70 billion parameters, and prototype, fine-tune, and infer AI models from DeepSeek, Meta, and Google, deploying them to data centers or the cloud.
3. Two DGX Sparks can be connected via NVIDIA's ConnectX-7 network to handle up to 405 billion parameters.
4. It can be directly connected to a monitor and mouse for use, or operated via network connection.
DGX Spark, formerly Project DIGITS, features the NVIDIA GB10 Grace Blackwell superchip combining the latest generation Blackwell GPU with a Grace CPU connected over NVIDIA's NVLink-C2C interconnect technology in a small form factor. The GB10 Blackwell GPU provides 6144 CUDA Cores and FP32 compute workloads, 5th generation Tensor cores that provide up to 1000 TOPS1 of FP4 AI performance, and 4th generation RT cores providing real time ray tracing acceleration. The GPU also includes 5th generation NVDEC and 9th generation NVENC engines, including support for 4:2:2 H.264 and H.265 video. The system features 128GB of coherent, unified LPDDR5x system memory. Accessible to both CPU and GPU, DGX Spark enables users to work with large workloads including AI models of up to 200 billion parameters.