Full Breakdown
Apple Unveils GPU Benchmarks for M5 Ultra and M6 Chips Ahead of New Macs
By Drooid · · How we work
Core Event: Benchmark Results Released
Apple’s latest high-performance silicon, the M5 Ultra and the entry-level M6, have had their first GPU benchmark data published. The figures come from Geekbench 7 Metal tests and an OpenCL comparison that places the M5 Ultra’s graphics capability near that of NVIDIA’s GeForce RTX 4080.
Performance Data: Metal Scores and Comparisons
- M5 Ultra: 366,744 Metal score, roughly 59 % higher than the M3 Ultra’s average of 230,420.
- M6: 93,217 Metal score, about 34 % higher than the M5’s average of 69,563.
- OpenCL testing shows the M5 Ultra’s performance aligning closely with the RTX 4080, while the M6’s OpenCL results were not highlighted.
For reference, other recent Apple silicon scores include the M3 Ultra (230,420), M2 Ultra (201,144), M1 Ultra (150,098) and the M1 (27,277).
Context: Upcoming Mac Studio and Mac mini Launch
The benchmark release precedes the rollout of new Mac Studio and Mac mini models that will incorporate the M5 Ultra and M6 chips. Apple has positioned these machines as the next step in its transition away from Intel graphics, targeting professionals who need high-throughput GPU workloads such as video rendering, 3D modeling, and machine-learning inference.
Implications for Users and Developers
Higher Metal scores suggest substantially faster graphics rendering and compute tasks in native Apple environments. Developers can expect reduced frame times and higher throughput for Metal-based applications, while creative professionals may see shorter export times for high-resolution video projects. The proximity of the M5 Ultra’s OpenCL performance to a high-end RTX 4080 indicates that Apple’s integrated GPU is competitive with leading discrete solutions, potentially reshaping purchase decisions for users who previously relied on external GPUs.
Upcoming Availability
Apple plans to ship the new Mac Studio and Mac mini models featuring the M5 Ultra and M6 chips in the coming days, aligning product availability with the benchmark disclosures. Real-world performance will still depend on software optimization and workload characteristics.