According to YouTuber iCaveDave, the Apple M2 'will have GPU performance...in line with a GTX 1080 Ti'. A huge claim indeed, considering that the GeForce GTX 1080 Ti is not far off the pace of high-end RTX 30 series cards. Unfortunately, there is little data behind iCaveDave's tweet and subsequent video.
iCaveDave has based his assertions around rumoured Apple A15 Bionic GPU benchmark scores, which leaked earlier this week. Supposedly, the A15 Bionic offers nearly 14% better GPU performance than the A14 Bionic in GFXBench 3.1 Manhattan. We should stress that these performance numbers are from an engineering sample, so retail chips may perform differently.
Apple is thought to have based the A15 Bionic and M2 on comparable architectures, with the latter expected to debut in next year's MacBook Air. iCaveDave has extrapolated the performance increase between the A14 Bionic and A15 Bionic, applying it to the Apple M1 GPU score in the Geekbench Metal benchmark database. From that, iCaveDave has determined that the Apple M2 would score 30,085 points in the same benchmark, putting it a smidge ahead of the Radeon Pro 5500M.
A score of around 30,000 points is also close to the GeForce GTX 1080 Ti, which scored 30,597. Relying on this score alone is unhelpful though, because NVIDIA cards do not fully support Metal. For instance, the NVIDIA TITAN XP scores less than the Radeon Pro 5600M, two cards that differ wildly in performance.
Frankly, the idea that the Apple M2 could match the GPU capabilities of the GeForce GTX 1080 Ti is risible. The latter is a four-year-old desktop GPU with a 250 W TDP and 3,584 shading units at its disposal. By contrast, the M2 GPU will sit within an SoC intended for laptops. In short, there is no chance that the Apple M2 will offer anywhere the GPU performance of a high-end desktop graphics card.