NVIDIA’s RTX 40 launch highlight particularly focused on a multi-pronged approach. NVIDIA is laser-focused on capturing the entire gaming ecosystem and Ada Lovelace is their jumping off point in encroaching into the large part of markets endemic to gaming and multimedia creation.
The RTX 40-series are still gaming cards first and foremost and their gaming horsepower comes mainly from the new silicon. Ada Lovelace is more of an efficiency refinement for Ampere which was built on Samsung 8nm and with Ampere looking to stay with us for a while, NVIDIA’s pricing structure pretty much confirms that with the RTX 4080 breaking the $600 price point that the NVIDIA xx80 cards sat at for the last 4 generations.
That brings us back to the architecture. NVIDIA is frontloading a lot of functionality for various industries with the RTX 40 series cards primarily the RTX 4090 and RTX 4080. While the raster engine remains practically the same, NVIDIA’s addiitonal fixed function hardware adds extra processing power for additional graphics rendering, particularly ray traced graphics.
Third generation RT cores found on RTX 40 graphics cards include acceleration for math-heavy functions such as bounding-volume hierarchy which is used in dictating how rays interactive on objects and what their defined limits are. Traditional BVH is very precise and takes up a decent amount of computational power but thanks to NVIDIA’s Displaced Micro-Mesh engine, BVHs can now be represented as coarser sets of triangles versus the traditional finer geometries required by older generations.
Opacity Micro Mesh is another NVIDIA technology utilizing 3rd-gen RT cores which accelerate alpha channel object shading. OMM isolates non-alpha areas and reduces pipeline workload by focusing on alpha objects only.
DLSS3 is perhaps the most talked about feature of the RTX 40 series which introduces Frame Generation. With Frame Generation, RTX 40 cards can insert extra frames between every 2 frames on top of DLSS upscaling. This is also performed by the Tensor cores leaving the the raster cores and RT cores free to deliver the RTX 4090 and RTX 4080’s performance targets which ensures Frame Generation works at its best frame rates.
Despite being simple on paper, Frame Generation is limited to the RTX 40 series for now as it requires optical flow acceleration found on RTX 40 cards. This feature is required to allow the GPU to know what object is moving and which aren’t which lets it precisely map out its next potential frame. As inserted frames will not be interactive, NVIDIA Reflex plays a vital role to keep input latency in-sync with actual frames with Reflex markers denoting key frames in the pipeline.
All these technologies are further expanded into other applications as DLSS3 will see it put into action in OBS capture for 4K 120 capture. This is also extends to the RTX 40 series native support for AV1. AV1 is an emerging format for video which allows lower bitrates but with better quality. NVIDIA RTX 40 series cards support acceleration for both encoding and decoding of AV1 videos with support coming for platforms such as Discord and Youtube as well as encoding support for popular software like Adobe Premiere Pro and BlackMagic Design’s Davinci Resolve. Support for AV1 encoding via NVENC will also allow streamers to utilize their system more without adding workload to their GPU, removing the need for dual PC capture systems on supporting systems once support for AV1 arrives on popular streaming software like OBS.
Further utilization for RT cores can be seen on partner RTX Studio software that can take full advantage of the faster performance of NVIDIA RTX 40 series cards. This in turn can help those in the gaming modding scene with NVIDIA’s RTX Remix. RTX Remix is a modding tool that can easily take classic game titles and apply raytracing onto them. NVIDIA’s demo for this is Portal RTX which should arrive soon and we’ll talk about RTX Remix in another article.