The tech community is buzzing with controversy after AMD’s highly anticipated Radeon RX 9070 XT, part of its next-gen RDNA 4 lineup, received scathing criticism on Userbenchmark. The GPU, which was recently teased by AMD as a “game-changer” for 4K gaming and AI workloads, is now under fire for allegedly underperforming in early benchmark tests posted by users.
According to a Userbenchmark report published earlier this week, the RX 9070 XT scored significantly lower than competing models like NVIDIA’s RTX 5090 in several key metrics, including rasterization performance and ray-traced workloads. The test, conducted by an anonymous user, claims the card struggled to maintain stable frame rates at 1440p and 4K resolutions in titles like Cyberpunk 2077 and Starfield.
“This isn’t just a minor dip—it’s a disaster,” wrote one commenter on the forum. “For a card marketed as ‘premium,’ it’s barely keeping up with last-gen GPUs.” Others criticized AMD’s driver optimization, citing frequent crashes and thermal throttling during stress tests.
The backlash comes just days after AMD teased the RX 9070 XT at a virtual event, where the company highlighted its 3nm architecture and 24GB of GDDR7 memory. Industry analysts had speculated the card would close the performance gap with NVIDIA’s Lovelace-based GPUs, but early adopters are now questioning whether AMD overstated its capabilities.
AMD has yet to officially respond to the criticism, but insiders suggest the company is rushing to address driver issues ahead of the card’s rumored Q2 2025 launch. Meanwhile, skeptics argue that Userbenchmark’s results should be taken with caution, as the site has historically faced accusations of bias toward Intel and NVIDIA hardware.
For now, the RX 9070 XT’s rocky debut serves as a reminder of the high stakes in the GPU arms race—and the unforgiving scrutiny of PC enthusiasts. With official reviews still weeks away, AMD’s next move could make or break its bid for next-gen supremacy.
Update: AMD has released a statement urging patience, claiming the tested unit was an “early engineering sample” and not reflective of final performance.