The performance growth of GPUs recently has shown some fascinating trends. NVIDIA, for the past three generations, has been increasing the power of its fastest graphics chips by approximately 50% each time, with an additional 10% boost achievable over the architecture’s lifecycle. However, this growth primarily occurs in the high-end segment, with budget options being conspicuously absent and a general price increase across all graphics card tiers. For instance, the GeForce RTX 3050, initially overpriced, now sells for $250.
While one could justify these price hikes due to inflation, economic challenges, and the costs of developing new architectures, the stagnation in performance improvements is concerning. For example, the GeForce RTX 2060 Super released in 2019 was matched in performance by the RTX 3060 in 2021. The RTX 4060, released in 2023, showed no significant performance improvement over its predecessor. This pattern is evident in more expensive models as well and is not exclusive to NVIDIA; AMD shows similar trends.
However, for those who can afford the flagship models, there’s exciting news. Insiders from Moore’s Law is Dead have revealed details about the GeForce RTX 5090. This upcoming flagship, expected to offer ultimate performance at a significant cost, could be 80% faster than the currently available RTX 4090. The RTX 5090, based on the Blackwell architecture, is anticipated for release by late 2024, though NVIDIA has not finalized launch dates, and market performance of the RTX 4090 could push the release to 2025. AMD’s market activities and potential releases might also influence NVIDIA’s decisions.
Rumors suggest the Radeon RX 8900 XTX won’t aim for flagship status but will instead offer performance similar to the RX 7900 XTX, focusing on a balance of cost, performance, and energy efficiency, potentially targeting the mid-high segment. This strategy could signify AMD’s return to its roots, focusing on products that resonate with gamers globally.
The specifics of how NVIDIA plans to significantly boost the performance of the RTX 5090 are not detailed. The RTX 4090 is based on a 4-nm process, and as of 2024, nothing beyond a 3-nm process is expected to be available. Nevertheless, skepticism is advisable when considering these insider reports, as past predictions about GPU performance have not always been accurate.
The upcoming release of the GeForce RTX 40 Super series is also on the horizon, promising more memory and faster performance than current models. It remains unclear whether these will replace older models or simply occupy a higher tier in NVIDIA’s lineup, potentially commanding a premium price.
