AMD CEOs David Wang and Rick Bergman have confirmed that we will see much more AI in the next generation graphic cards from the tech giant, which will be built on the AMD RDNA 4 architecture.
In a recent interview with Japanese gaming site 4gamer, AMD executives detailed what we can expect from RDNA 4. Of course, the front and center were confirmation that we’d see a second iteration of Team Red’s AI accelerator cores (similar to Nvidia’s Tensor cores), which were first introduced in the current generation RDNA 3 GPUs – such as excellent Radeon RX 7900 XTXCurrently AMD’s best graphics card in the shop.
Nvidia’s tech is still light years ahead of AMD when it comes to AI processes – just look at RTX4090 – but these second-generation AI cores should offer a major step forward. Aside from the accelerator cores, the two also discussed some other cool new features, most notably the new GPU independent pipeline which allows rendering and texturing processes to be generated purely on the GPU without having to communicate with the CPU.
This has great potential to increase the processing speed of RDNA 4 GPUs as it will not have to rely on the CPU and system RAM to perform certain tasks, effectively eliminating two potential system bottlenecks. According to Wang and Bergman, we can expect a massive 2.2x performance increase over current RDNA 3 cards.
Analysis: AMD fully supports the fashion for artificial intelligence
Perhaps even more interestingly, Wang was eager to discuss implementing AI in the gaming space. In an interview with 4gamer, he stated that AI hardware should not only be used to improve the graphical aspect of games, claiming that the inference cores on the GPU can actually be used to improve the gameplay itself.
This has potentially terrifying implications – imagine you’re playing a shooter where the enemies are actually driven by a deep learning AI housed in your computer that really reacts, learns, and adapts to your actions. Sounds the same the youngest somewhat dystopian to imagine the boss fight in the game ChatGPT.
Still, on a slightly less terrifying level, it’s interesting to consider the uses of AI cores for things like NPC pathfinding in games. Every player has complained at some point that the game’s characters walk too slowly or get stuck on the door frame due to wobbly pathfinding – Wang believes that AI cores can solve many such immersion-breaking problems.
It’s kind of funny that AMD is so committed to AI now, ever since Team Red he was basically mocking Nvidia last year for relying too much on AI technology in their RTX i GPUs DLSS. In its defense, AMD has proven that upscaling technology such as DLSS can be achieved without the use of AI, as seen in the Radeon FidelityFX Super Resolution (FSR) software. AMD’s seniors seem to have become AI converts, but they have different – and intriguing – ideas on how it should be used compared to Nvidia’s plans. We cannot blame them for wanting to go in a different direction; Nvidia has been having a lot of problems lately.
While we don’t know exactly when RDNA 4 will land, AMD’s 2-year tenure when it comes to GPU launches means we can most likely expect them towards the end of 2024. Bergman assured 4gamer that new cards will be released in the near future, although “close” may be a stretch here.