iPhone 15 Pro Max Neural Engine Sees Huge Performance Boost After iOS 18 Update, Up To 25 Percent Gains In Early Benchmarks

Apple's iOS 18 announcement gave us more than we asked for, and the company's approach toward the end-user has changed over the past few years. The company has set a limitation for its high-end Apple Intelligence features, requiring an iPhone 15 Pro that features an A17 Pro chip with a faster Neural Engine than its predecessors. Now that iOS 18 is out and about as a developer beta, it has been discovered that the Neural Engine in the iPhone 15 Pro Max is 25 percent faster than before.

Early benchmarks show a 25 percent gain in iPhone 15 Pro Max Neural Engine performance

iOS 17 was not able to fully take advantage of the A17 Pro chip in the iPhone 15 Pro models and remained untempered until the company announced iOS 18. It has been observed that the Neural Engine in the iPhone 15 Pro Max is up to 25 percent faster than before, which would prove to be a major help in terms of machine learning or Apple Intelligence tasks.

Related Story Download: Apple Releases iOS 17.6, watchOS 10.6, macOS 14.6, And More Beta Builds To Developers After iOS 18 Unveiling

Discovered by INIYSA, the Geekbench score for the Core ML Neural Engine Inference Score on the iPhone 15 Pro Max was 6,000 when running iOS 17.4.1. However, after updating the device to the first beta of iOS 18, the score saw a notable jump to 7,816. It can be seen that the score is up to 25 percent higher than the iPhone 15 Pro Max running the currently available iOS 17.4.1. The company could have kept the A17 Pro's capabilities hidden, as it would have been an overkill for features present in iOS 17.

The massive performance jump could be related to the new Core ML type, MLTensor, which allows multidimensional arrays of computations with a familiar API. Check out how the company explains the new Core ML type.

The company also anticipates an increase in the performance of the Neural Engine for the iPhone 15 Pro, catering to different machine learning models. This also includes a marginal 2 percent increase in MobibleNet