Nvidia's New DGX Spark Costs More for Less: A Math Masterclass
Nvidia just dropped a $4,999 desktop AI computer that somehow manages to make previous versions look like a bargain. It’s a masterclass in creative pricing that would make even a Wall Street shark blush with embarrassment.
The new DGX Spark 64GB comes equipped with the Grace Blackwell GB10 chip and a ConnectX-7 network adapter, running on dedicated DGX OS. The device utilizes a unified memory architecture, allowing the processor and GPU to share a single pool of data without the constant, soul-crushing overhead of copying files back and forth.
According to Nvidia, this hardware is capable of running models with up to 100 billion parameters locally, provided you’re willing to rely on quantization to cram that knowledge into the available silicon. It plays nice with the usual suspects like PyTorch, Ollama, vLLM, and LM Studio.
For those who find 64GB insufficient, two units can be daisy-chained via ConnectX-7 to create a 128GB cluster. While Nvidia claims this setup significantly boosts performance, internal tests show it doesn't double the speed, proving that even a tech giant hasn't quite managed to legislate away the laws of physics.
The pricing strategy is where the real comedy begins. This 64GB variant hits the market at $4,999, while the previous 128GB model was listed at $4,699. It appears Nvidia has successfully pioneered a new economic model where getting half the hardware entitles you to pay a premium for the privilege of being a pioneer.
Source: NVIDIA Blog
Comments
Help shape the next version: Add context or suggest a correction. AI review can add points toward a rewrite. Reviews and updates may take time; a full meter does not guarantee a new version.