Nvidia’s new RTX 40 graphics cards are said to offer a huge leap forward in performance. How is this noticeable in VR games?
As expected, Nvidia CEO Jensen Huang introduced the new Geforce RTX-40 series graphics cards in his September 20 video keynote. Based on the Ada-Lovelace architecture, the cards use AI-assisted DLSS-3 rendering. The third generation RTX is claimed to deliver results up to four times faster than the Ampere architecture.
According to the manufacturer, streaming multiprocessors are more than double the previous generation. An RTX 4090 runs with 76 billion transistors and 16,384 CUDA cores, its shader performance is up to 83 teraflops (RTX 3090 Ti: 40 teraflops). Fourth-generation Tensor Cores allow for a massive performance leap. They deliver up to 1.32 petaflops Tensor, five times more than the previous generation with FP8 acceleration.
Up to 83 teraflops with the RTX 4090
The purchase will not be cheap in any case. A lavish $1,550 buy it RTX-4090. The price for the RTX 4080 costs $1,119 (16 GB) or 899 € (12 GB). The latter relies on a smaller variant of the chip in addition to reduced memory. Nvidia may have moved a planned 4070 to the 4080 price segment here.
Micron’s high-speed GDDR6X memory is also installed on all models. According to Nvidia, the RTX 4090 consistently delivers over 100fps in games at 4K resolution.
High-resolution VR headsets, in particular, benefit from faster memory. Not to mention the general increase in performance, as long as it is confirmed by independent benchmarks. The Varjo Aero (test), for example, with its 2880 x 2720 pixels, quickly reached its performance limits with current graphics cards. Pimax models like the upcoming Pimax Reality 12K QLED (5620px × 2720px, 200Hz) should also “roll” better.
Nvidia also touted third-generation RT cores with up to 191 teraflops for more efficient ray tracing. The RT cores supposedly run 2.8 times more efficiently than in the previous generation. However, technology remains the exception in VR games.
Some titles are already taking advantage of dedicated ray tracing cores on current RTX cards. These feature real-time interactions and rendering while maintaining the high frame rates typical of VR. The mechanical simulator Wrench VR, for example, uses both the latest DLSS and ray tracing for high-quality reflections and shadows.
In Nvidia’s Omniverse XR app, you can already explore digital worlds that use ray tracing with the VR headset. There it also creates soft shadows and realistic highlights. Learn more about Omniverse Cloud, recently announced hardware, and new features here.
DLSS 3 in Microsoft Flight Simulator
The recently announced DLSS 3, on the other hand, is designed to take AI rendering to the next level. The new iteration computes full frames without burdening the actual graphics pipeline. Up to seven-eighths of the displayed pixels can thus be reconstructed with AI. This can quadruple the frame rate compared to an image without DLSS 3.
The principle is similar to reprojection techniques in the field of VR. The Spacewarp application, for example, also calculates intermediate frames to achieve up to 70% more performance.
Nvidia’s new neural rendering feature supports Unity game engines as well as Unreal Engine 4 and 5. The list of supported DLSS-3 titles also includes a VR game, Microsoft Flight Simulator. The highly detailed flight simulation occasionally suffers from slight frame rate “hangs” and is heavily CPU limited.
Since DLSS-3 uses “multi-frame optical generation” via a neural network to decouple the rendering of new frames from the game engine, the technology can work around CPU limitations as well. Tests should show how the technology works in virtual reality.
Nvidia’s relative performance data for the F1 22 (test) racing game, which is also playable in VR, is still more than two and a half times better with DLSS 3 (see screenshot for test system). However, Nvidia refers to gaming on a 4K (3840 x 2160 pixel) monitor.
In particular, demanding pit lane scenes, which showed issues when testing with an RTX 2080 Ti, could flow more smoothly into the racing game with possible use of DLSS-3. They were already slowing down the game on a Valve Index (1440×1600 pixels per eye) and causing unpleasant jerks. Generally, it should be easier to pull 144 hertz pleasantly smoothly out of Valve’s headset. Oversampling would also make sense given the sharpness of the background.
More details from Nvidia
Here are some other announced specs of the RTX-40 series, which Nvidia says provide the following benefits:
- Shader Execution Reordering (SER) improves execution efficiency by rescheduling shading workloads on the fly. This allows better use of GPU resources. SER is as big an innovation as out-of-order execution for CPUs, improving ray-tracing performance by three times and frame rates in games by up to 25%.
- The Ada Optical Flow Accelerator with twice the performance enables DLSS 3 to predict motion in a scene, allowing the neural network to increase frame rates while maintaining image quality.
- Architectural enhancements closely coupled with custom TSMC 4N process technology deliver up to two times improved energy efficiency.
- The RTX 4090 is up to twice as fast in today’s games while maintaining 450W power consumption.
The Geforce RTX 4080 and 4090 will be available in various variants from manufacturers, partners like Asus, Colorful, Gainward, Galaxy, Gigabyte, Innovision 3D, MSI, Palit, PNY, and Zotac. A reference design from Nvidia called “Founders Editions” is also planned again, but only for the RTX 4090 and RTX 4080 (16GB) models. Click here for the availability notification on the official website.
AMD Counters in November, Nvidia May Stumble on Pricing
The competition is already in the starting blocks. Shortly before Nvidia’s presentation, AMD announced November 3 as the launch date for the new RNDA-3 architecture cards on Twitter. Do not expect a gift for gamers, AMD will follow Nvidia’s pricing.
However, this could become a problem for Nvidia. High prices, especially for 4080 cards, are heavily criticized on YouTube, Reddit and other platforms. The RTX 3080 started at $699 – Nvidia is charging $200 to $500 more for the next-gen. Those who see the 4080 in the 12 gigabyte variant as a reimagining of the 4070 may see premiums ranging from $499 for the RTX 3070 to $899 for the RTX 4080 (12GB).
At the same time, prices for the 3000 series are slowly approaching realistic prices. The promise of double or quadruple frame rates from the 4000 series refers to cases where DLSS 3 and ray tracing can be used by looking closely at the few graphics published by Nvidia. For games that don’t support DLSS and ray tracing – which is the majority of VR games today – the benefit is much less.
The RTX 3090 Ti could drop below $1,000 in the coming months, and will then cost around half of the next-gen top-end model. Those who want DLSS 3, faster ray tracing, and probably around 30-60% more performance might be willing to pay the extra price. Everyone should wait for cheaper models, buy older models at lower cost – or hope for AMD.