Inching closer to launch of next-generation GPUs from Nvidia, the rumour mill goes into RTX 40-series overdrive. The latest titbit comes by way of seasoned Twitter prognosticator kopite7kimi, who believes range-topping RTX 4090 will have a boost clock of 2,520MHz and in-game frequency of more than 2,750MHz.
It would be easy to dismiss such rumour as wild speculation, but as AMD has shown recently, leading-edge processes enable high frequencies across the range. Take Radeon RX 6950 XT as an example, boosting to over 2,400MHz with relative ease.
Nvidia is using TSMC’s 4nm process for next-gen 40-series GPUs, so it ought to be a given clock frequencies will run much higher than present generation GPUs fabbed on Samsung’s 8nm technology. Nevertheless, comparing like for like, GeForce RTX 3090 has a 1,695MHz boost speed mandated by Nvidia. If RTX 4090’s specifications hold true, it will enjoy a 49 per cent frequency hike.
But that’s only half the story. We know RTX 3090 produces 35.6TFLOPs of single-precision compute. RTX 4090’s 16,384 shaders and sky-high core speed leads on to 82.6TFLOPs compute, and likely above 90TFLOPs during real-world gameplay. That’s a step-change in performance even if improvements in architecture don’t yield huge results.
Notwithstanding benefits of using a leading-edge process, most commentators agree RTX 4090 is bit of a power beast, pulling at least 450W at full chat. We expect most add-in card partners to fine-tune cooling already used on existing RTX 3090 Ti boards that pull the same wattage presently.
RTX 4090 lands sometime in September, and it’s sensible to consider Nvidia will claim RTX 4090 is at least twice as fast as RTX 3090 at 4K. Going by specs, RTX 4080 ought to be also faster, too, and even RTX 4070 will get close to today’s RTX 3090 levels.
AMD has an uphill fight to match Nvidia’s impressive specifications; RDNA 3 better bring the performance mojo.