Leaked RTX 5080 benchmark: it’s slower than the RTX 4090

Recently leaked benchmarks have shed light on the performance of Nvidia’s upcoming RTX 5080 GPU, set to debut alongside the flagship RTX 5090 on January 30. The GPU surfaced on the Geekbench browser, undergoing OpenCL and Vulkan benchmark tests. However, the results suggest it might fall short of ranking among the best graphics cards available.

The leak, initially reported by Benchleaks, appears to have originated from a reviewer who unintentionally made the data public. The RTX 5080 tested was an MSI-branded variant with the model number MS-7E62. It was paired with AMD’s Ryzen 7 9800X3D—a processor hailed as one of the most powerful for gaming—an MSI MPG 850 Edge TI Wi-Fi motherboard, and 32GB of DDR5-6000 memory, creating a robust test setup.

In a recent performance benchmark, the RTX 5080 demonstrated impressive results with a score of 261,836 in the Vulkan test and 256,138 in OpenCL. When compared to its predecessor, the RTX 4080, the RTX 5080 delivers approximately 22% higher performance in Vulkan, showcasing a significant leap in capabilities. However, the improvement in OpenCL is more modest, with a 6.7% increase, highlighting a more incremental upgrade in that area. These results underline the RTX 5080’s focus on delivering a robust boost in Vulkan-intensive tasks, while providing steady gains in OpenCL performance.

The highly anticipated NVIDIA RTX 5080 has made its first appearance on the Blender Open Data platform, as reported by Reddit user TruthPhoenixV. According to the platform, the RTX 5080 achieved an impressive median score of 9,063.77. This performance marks a 9.4% improvement over the RTX 4080 and an 8.2% edge over the RTX 4080 Super, showcasing its potential as a significant upgrade for professionals and enthusiasts using Blender.

Leaked benchmarks indicate that the RTX 5080 may not outperform the RTX 4090 in terms of raw performance. Historically, GPUs in the 80-class series have consistently surpassed the 90-class models of the previous generation. However, this pattern appears to be shifting, as early reports suggest the RTX 5080 falls short of the RTX 4090’s capabilities.

It’s worth mentioning that these benchmark results are preliminary and have not been officially verified. Potential buyers are advised to wait for the official review embargo to lift before forming any conclusions.

To summarize, the RTX 5080 leverages Nvidia’s cutting-edge Blackwell architecture, packing an impressive 10,752 CUDA cores across 84 Streaming Multiprocessors (SMs)—a notable upgrade from the 9,728 cores in its predecessor, the RTX 4080. The GPU is equipped with 16GB of advanced GDDR7 memory operating on a 256-bit bus, delivering higher bandwidth and efficiency. Nvidia touts theoretical performance figures of 1,801 TOPS in AI tasks via Tensor Cores and 171 Teraflops of ray tracing power through its RT Cores, promising substantial improvements for gamers and creators alike.

The RTX 5080 starts at $1,000, though models from Nvidia’s board partners are expected to be priced higher.