While NVIDIA's 80-class GPUs dominate AMD's RX 9070 XT in gaming tests, the latter might have a lead in rendering tests, when conducted using Cooperative Vectors.
AMD's Native Vulkan Cooperative Vectors Takes Performance to New Levels, Beating NVIDIA's Top-End GPUs
We recently did an in-depth dive into NVIDIA's RTXNTC technology, and how by leveraging DXR 1.2, Team Green has managed to squeeze out massive performance and reduce VRAM consumption. Now, the X user @opinali has conducted extensive testing to see how AMD's RDNA 4 GPUs manage to run under Neural Text Compression (NTC), and the results are indeed surprising, since under a low-end rendering scenario, AMD's Radeon RX 9070 XT actually manages to beat out the GeForce RTX 5080, showing how well optimized the Team Red tech stack is with rendering applications.
Now, since AMD hasn't released drivers that support Microsoft's DXR 1.2 version, in order to test out performance under Cooperative Vectors, the user managed to integrate Vulkan's "VK_KHR_cooperative_matrix", which is a similar implementation that AMD has been supporting since RDNA 3 GPUs. But, without DXR 1.2, the RX 9070 XT only managed to run under DP4A mode, and with that, in both Vulkan and DX12 testing, the RDNA 4 GPU actually surpasses the GeForce RTX 5080, showing a notable 10% performance difference in Vulkan, and a whopping 110% lead in DX12.
While the difference does seem a bit large here, it is important to note that NVIDIA's Blackwell preview drivers supporting Cooperative Vectors are under the preview stages; hence, the performance isn't as optimized, whereas for AMD, they have done a phenomenal job with optimizations. This shows that beefing up GPU specifications isn't the only leading factor for performance; instead, driver and software optimizations play a massive role, so the Radeon RX 9070 XT outperformed NVIDIA's high-end GPUs in these benchmarks.
Neural Text Compression (NTC) is a major factor in scaling GPU performance on a software level. When you combine RTXNTC with rendering applications, you see a 90% decline in VRAM usage, while this technique could take a lot of time to enter mainstream gaming titles, it surely is an optimistic technique.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
