Application Refactoring
Topaz Labs had to ship and maintain large sets of static inference engines across millions of PCs and many RTX GPU generations, which made releases heavy, slowed updates, and added weeks of precompilation work whenever the company changed its AI models.
Reported impact
The new deployment model cut processing time by more than 20% in benchmark tests, reduced engine storage by 3x to 4x, eliminated the need to ship dozens of static engines, and let Topaz move faster on future AI features for its large RTX install base.
Topaz Labs refactored the inference architecture behind Topaz Video around NVIDIA TensorRT for RTX so model engines can be generated on device instead of being precompiled and bundled for every target GPU. Previously, the company had to maintain and distribute hundreds of static engines across GeForce RTX 20, 30, 40, and 50 series systems, which increased installer size and made new releases cumbersome. With TensorRT for RTX integrated into the application, Topaz now ships a single package and lets each Windows PC create optimized engines locally for its specific hardware. The change preserved high-performance AI video enhancement while simplifying deployment, reducing storage requirements, and positioning the product to support future diffusion-based restoration and real-time enhancements without repeatedly rebuilding the distribution pipeline from scratch.
Topaz ships one application package, each RTX PC generates a local TensorRT engine for the installed model, and video frames then flow through that hardware-specific engine for upscale, denoise, and enhancement inference on the user's device.
By leveraging NVIDIA TensorRT for RTX, Topaz Labs shifted to on-device engine generation.
More than 20% speedup in processing time to upscale 1,620 frames of video from 1080p to 4K on NVIDIA GeForce RTX 5090 GPU.
The team had to pre-generate and ship large, static inference engines for multiple GPU types.
Topaz Labs successfully eliminated the need to ship dozens of static engines.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
The Organization schema identifies 'Topaz Labs LLC' and says the company is headquartered in Addison, Texas, with addressCountry 'US'.
The public profile states, 'Topaz Labs is a Software Development company located in 5100 Belt Line Rd, Dallas, Texas, 75254, US with 91 employees.'
Related
OpenAI Sample Oct 9, 2026 Asana Asana needed to cut the cost and latency of its StackAI browser agent so it could offer customers more capable models for browser-driven workflows without unsustainable operating expense. $0.47 average estimated model cost with the optimized GPT-6.1 Sol workflow Codex · Command · GPT-6 Astra OpenAI Sample Oct 8, 2026 Braintrust Braintrust wanted to respond to customer feature requests faster and run more engineering experiments, but slower coding tools and backlog-driven workflows made it costly to turn feedback into working prototypes in real… 50% of the Braintrust team moved to Codex in one month Codex · GPT-5.5 OpenAI Sample Oct 8, 2026 Notion When the feature request came in, Notion's mobile app already had a functional version, but the desktop and web client didn't. Even though Ryan wasn't entirely sure how the feature worked on mobile, he was able to give… "Whenever I need to research a task, fix a bug, or make a little tweak, Codex is just there, ready and willing.… Codex 
