AMD Unveils New Radeon RX 7900 XTX and Radeon 7900 XT GPUs with RDNA 3 Architecture

AMD GPU Radeon RX 7900 XTX and Radeon 7900 XT RDNA 3, GamersRD

On this day amd introduced its new $999 Radeon RX 7900 XTX and $899 Radeon RX 7900 XT GPUs with RDNA 3 architecture.

These are set to compete head-to-head with the best graphics cards, and it looks like AMD could have a legitimate shot at the top of the GPU benchmark hierarchy. Most of the details line up with what was already expected and covered about AMD RDNA 3 architecture and RX 7000 series GPUs.

AMD GPU Radeon RX 7900 XTX and Radeon 7900 XT RDNA 3, GamersRD

RDNA 3 will use chiplets, with a main GCD (Graphics Compute Die) and up to six MCDs (Memory Cache Dies). Additionally, there are many hidden changes to the architecture, including more compute units and many more GPU shaders compared to the previous generation. AMD continues to focus on power and energy efficiency, targeting a 50% improvement in performance per watt with RDNA3 compared to RDNA 2. We know that Nvidia’s RTX 4090 and Ada Lovelace pushed up the voltage and frequency curve, and as shown in our RTX 4090 efficiency scale, capping the RTX 4090’s power at 70% greatly boosted the Nvidia efficiency. However, AMD apparently doesn’t feel the need to dial power usage down to 11 by default.

AMD has two variants of the Navi 31 GPU coming out. The higher specification RX 7900 XTX card uses the fully enabled GCD and six MCDs, while the RX 7900 XT has 84 of the 96 compute units enabled and only uses five MCDs. The new RDNA 3 Unified Compute Unit has 64 two-issue stream processors (GPU shaders). That’s double the amount of RDNA 2 per compute unit, and AMD can send different workloads to each SIMD unit, or can make them both work with the same instruction type.

AMD says it has optimized its ray accelerators and that RDNA 3 versions can handle 1.5x more rays, with new dedicated instructions and improved BVH (ray/box) sorting and traversal. AMD talked more about its FidelityFX Super Resolution (FSR) technology. At this time, some versions of FSR are available. The original FSR (1.x) uses spatial scaling and can basically be built into drivers for things like Radeon Super Resolution (RSR). Meanwhile, the newer FSR 2.0 and 2.1 use time scaling and have inputs similar to DLSS and XeSS: motion vectors, depth buffer, and current and previous frame buffer.

AMD currently has over 216 games and applications that use FSR, but most are implementations of FSR 1.x; again, it’s easy to integrate and open source, and has been available for over a year. FSR 2.0 is much newer, having first arrived in May 2022. FSR 2.1 tweaks the algorithm to help remove ghosting and further improve image quality, and it’s only in a handful of games at this time. moment. FSR3 will arrive sometime next year. It will be looking to do some form of frame generation or interpolation, something similar to what Nvidia is doing with DLSS 3. The new AMD Radeon RX 7900 XTX and Radeon 7900 XT GPUs will launch on December 13.

Reference-gamersrd.com