"NVIDIA L40S GPUs: Power real-time ad decisioning and streaming media across connected TV (CTV), cutting latency to one millisecond and boosting ROI for publishers and advertisers. NVIDIA Triton Inference Servers: Serve the right ad requests daily to DSPs via sophisticated traffic shaping for higher DSP spend and more efficient integrations. NVIDIA RAPIDS Accelerator for Apache Spark: Deliver lightning-fast analytics, empowering clients with quicker insight for smarter optimisation at a lower cost."
"By working with NVIDIA as an early adopter through an iterative process of testing, data sharing, and solution refinement, this infrastructure overhaul delivers AI processing up to five times faster than traditional systems, helping publishers recover ad spend lost to latency and timeouts, while unlocking AI-driven optimisation strategies that were previously possible only within walled gardens. Unlocking lost revenue: The new technology reduces inference latency from the industry standard of 5-10 milliseconds to approximately 1 millisecond, achieving 85% fewer auction timeouts."
PubMatic integrated NVIDIA L40S GPUs, Triton Inference Servers, and the RAPIDS Accelerator for Apache Spark to accelerate programmatic advertising. The new infrastructure reduces inference latency from typical 5–10 milliseconds to about 1 millisecond, producing approximately 85% fewer auction timeouts and recovering previously lost ad spend. AI processing runs up to five times faster than traditional systems, enabling microsecond decisioning across trillions of daily auctions and improving accuracy and efficiency. Triton supports traffic shaping and better DSP integrations, while RAPIDS enables faster analytics for smarter optimisation and lower operational cost. The solution boosts ROI for publishers and advertisers, especially across CTV.
Read at Exchangewire
Unable to calculate read time
Collection
[
|
...
]