The cancellation of the Navi 4C GPU was seen as a blow to AMD’s presence in the enthusiast graphics segment. The chipmaker made no official claims regarding the rumored RDNA 4 flagship, but has acknowledged the renewed focus on the midrange segments in recent months. In a YouTube video with “Moore’s Law is Dead,” former AMD employee James Prior explains why the massive Navi 4C GPU was scrapped.
High Production Costs & Limited Market Share
AMD originally planned to make a complex chiplet-based GPU for the enthusiast gaming audience, with up to 14 disaggregated dies. Codenamed Navi 4C, this SKU would likely power the now-cancelled Radeon RX 8900 XTX. The RDNA 4 flagship would cost somewhere between $1500 and $2000, and tackle NVIDIA’s upcoming GeForce RTX 5090 Blackwell flagship.
Unfortunately, Navi 4C’s manufacturing costs were too high to sustain the targeted $1500 MSRP. Pricing it higher would merely make it another Titan rather than the price-effective flagship expected from AMD. To make matters worse, the enthusiast market hasn’t grown much in recent years, making it a riskier investment still.
NVIDIA’s marketing campaigns often undermine Radeon products, often with proprietary technologies that offer superior performance. An excellent example would be DLSS, a deep learning upscaler that continues to be the best in the gaming industry.
The Radeon RX 7900 XTX tends to be faster than the GeForce RTX 4080 in rasterized games, but falls behind in ray-tracing. This greatly diminished its standing in the high-end graphics segment, limiting sales and prompting several price cuts.
RDNA 5/UDNA 5 Creeps Up
Then there’s the RDNA 5 graphics architecture, now rebranded as UDNA to signify the unification of RDNA and CDNA into a scalable one. According to Prior, the UDNA/RDNA 5 architecture is a more mature chiplet design, demonstrating better scaling and lower latency. A UDNA GPU should be able to achieve Navi 4C’s performance targets at significantly lower production costs, cementing the cancellation of 14-chiplet behemoth.
