Apple M5 Ultra Beats RTX 5080 in Single GPU Benchmark
Leaked Geekbench results show the new Apple chip outscoring Nvidia's flagship card by 26%, though specific game tasks still favor discrete GPUs.
Key points
- The Apple M5 Ultra scored 360,019 in Geekbench 7 Metal tests, beating the RTX 5080's 284,429 in OpenCL.
- The M5 Ultra is 41% faster than the previous M3 Ultra but loses to the RTX 5080 in particle physics tasks.
- The comparison is limited by different software interfaces and the high cost of the Mac Studio hardware required.
Leaked benchmark data suggests the Apple M5 Ultra GPU has surpassed the Nvidia RTX 5080 in raw graphical throughput. According to reports from Overclocking.com, the new chip achieved a score of 360,019 in the Metal section of Geekbench 7. This places it 26% ahead of the RTX 5080, which recorded 284,429 points in the OpenCL test.
The performance jump is significant compared to Apple's previous generation. The M3 Ultra scored 255,009 in the same metric, meaning the M5 Ultra represents a 41% increase in efficiency. However, these figures come from a single leaked submission on a Mac Studio configured with 256 GB of unified memory, a setup that carries a substantial price tag.
Mixed results in specialized tasks
While the overall score is higher, the M5 Ultra does not win every subtest. In horizon line detection, the RTX 5080 maintains a slight advantage. However, Apple’s chip dominates in background blur, operating three times faster than the Nvidia card. The gap is even wider in face tracking, where the M5 Ultra scored 589,323 points against 74,241 for the RTX 5080.
Conversely, the Nvidia card excels in particle physics simulations. It doubled the performance of the Apple chip, scoring 902,982 points compared to 373,097. This discrepancy highlights that raw benchmark scores do not capture the full picture of real-world utility. Different hardware architectures prioritize different types of calculations, leading to uneven performance across specific workloads.
Benchmark limitations and cost trade-offs
Direct comparisons between these scores are technically flawed because they use different software interfaces. Apple’s Metal framework is not identical to Nvidia’s OpenCL, making a perfect apples-to-apples comparison impossible. Furthermore, the data is not official; it relies on leaked submissions that may not reflect standardized testing conditions or thermal stability over time.
There is also a significant economic trade-off. A Mac Studio equipped with the M5 Ultra and high-capacity memory costs enough to purchase multiple RTX 5080 graphics cards. For users focused solely on cost-per-frame performance in gaming, the discrete GPU option remains more accessible. The ultimate verdict will require platform-agnostic benchmarks like Vulkan or native game tests, which have not yet been widely reported.
