A roadmap for NVIDIA GPUs recently leaked online, linking NVIDIA's Feynman architecture to the RTX 80 series and listing AMD RDNA 7. However, there are several obvious contradictions in the chart, making it an unreliable basis for confirming related products and release dates.
"Moore's Law Is Dead" claims that this presentation material comes from a credible source and states that a senior Broadcom executive displayed the chart during a demonstration attended by hundreds of people. NVIDIA has confirmed that Feynman is the next-generation data center architecture scheduled to debut in 2028, but has not confirmed any association with consumer graphics cards.
According to the roadmap, NVIDIA's products jump directly from the RTX 60 series using the Rubin architecture to the RTX 80 series using the Feynman architecture, completely skipping the RTX 70 series. The AMD section jumps from RDNA 5 to RDNA 7 and incorrectly labels the Radeon RX 9000 series, which actually uses RDNA 4, as RDNA 5.
Other content is also problematic. The chart lists Intel Celestial discrete graphics cards for 2025, yet no corresponding consumer products have appeared; the canceled Falcon Shores accelerator is still included, and the material mentions the Druid architecture, which is reportedly still undergoing adjustments. Some projects span six years, leading "Moore's Law Is Dead" to suggest that this span may represent product lifecycles rather than the interval between two generations.
"Moore's Law Is Dead" acknowledged defects in the document and speculated that its original version might date back to 2024, having only been incompletely updated for a 2026 presentation. Conflicting accounts of recent plans also fail to corroborate each other: "Moore's Law Is Dead" has repeatedly stated that the RTX 6090 might launch in 2027, while Kopite7kimi and Kepler_L2 believe the RTX 60 series cards will not arrive until 2028; Kepler_L2 also claimed that RDNA 5 might be announced in the second half of 2027.
At best, the existing materials reflect what may be an outdated internal or industry presentation and cannot be used to confirm the release dates for the RTX 80, RDNA 7, and related products.
Detailed report from foreign media:
A file purported to be a GPU roadmap recently circulated online, mentioning NVIDIA's future RTX 80 series and AMD's RDNA7 architecture. Although this document seems to herald upcoming graphics hardware, the numerous contradictions within suggest its content may be outdated or erroneous. This information comes from the YouTube channel "Moore's Law Is Dead," which claims to have obtained the presentation from a reliable source. Reportedly, an executive from Broadcom presented this roadmap during a speech attended by hundreds of people.
Perhaps the most striking detail is the apparent link between NVIDIA's Feynman architecture and the GeForce RTX 80 series. NVIDIA has confirmed that the Feynman architecture is its next-generation data center architecture, planned for release in 2028. However, it has not been associated with any consumer products, nor has it been mentioned by leakers.
Additionally, the document completely skips the RTX 70 series, jumping directly from RTX 60 GPUs based on the Rubin architecture to the RTX 80 series based on the Feynman architecture. On the AMD side, the roadmap mentions RDNA7 but appears to skip RDNA6. However, this naming convention holds no special implication, as AMD executive Jack Huynh had already discussed RDNA5, RDNA6, and RDNA7 in 2024 when outlining the company's long-term graphics architecture plans.
The document incorrectly associates RDNA5 with the Radeon RX 9000 series graphics cards, which actually use the RDNA4 architecture. It also shows Intel's Celestial discrete graphics card launching in 2025, although no such consumer products have emerged yet. Worse still, the canceled Falcon Shores accelerator project still appears in the product roadmap, alongside mentions of Intel's Druid architecture, which is reportedly undergoing adjustments but has not been completely canceled.
Some entries appear to span six years, but "Moore's Law Is Dead" points out that these likely represent product lifecycles rather than intervals between GPU generations. Tom also acknowledged flaws in the document, suggesting it may have originally been completed in 2024 and subsequently underwent incomplete updates before being used for a 2026 presentation. In other words, even if the slide content is authentic, it may not reflect current development plans.
The uncertainty surrounding RDNA7 and the RTX 80 series shown in the Broadcom roadmap also extends to NVIDIA's more immediate GPU plans. The website "Moore's Law Is Dead" has repeatedly predicted that the RTX 6090 might launch in 2027, a prediction seemingly corroborated by the alleged Broadcom presentation. However, well-known leakers Kopite7kimi and Kepler_L2 hold different views, believing that NVIDIA's gaming GPUs based on the Rubin architecture will not appear until 2028.
The release of RDNA5, the direct successor to AMD's RDNA4, may be later than previously expected. Kepler_L2 recently hinted that RDNA5 might not arrive until the second half of 2027. Furthermore, it is reported that only AT2 GPUs are expected to launch in 2027, while the flagship AT0 and other models will not be released until 2028.
Adjustments to GPU development schedules could also impact the next generation of gaming consoles. While continuing to support the PS5 and PS5 Pro, Sony is collaborating with AMD on Project Amethyst, which is expected to influence future PlayStation hardware. Meanwhile, Microsoft's next-generation Xbox console, Project Helix, will utilize a custom AMD SoC, with development hardware scheduled for delivery to studios in 2027 and a commercial release reportedly planned for late 2028.
Of course, the development progress of these two gaming consoles may not align with AMD's desktop GPU roadmap. However, since the RTX 60 series and RDNA5 architecture remain shrouded in various contradictory rumors, the so-called information regarding RTX 80 and RDNA7 provides few clear clues about future directions. Currently, the alleged Broadcom presentation seems to raise more questions than it answers.