Apple unveils M2 Pro and M2 Max SoCs for Mac mini and MacBook Pro, claims M2 Max GPU prowess is unmatched even by Quadro RTX 6000 and RTX 3080 Ti-powered laptops
In a rather uneventful release, Apple today unveiled the M2 Pro and M2 Max SoCs alongside the new Mac mini and MacBook Pros. The M2 Pro offers up to a 12-core CPU, 19-core GPU, and up to 32 GB of unified memory while the M2 Max offers the same 12-core CPU but with up to a 38-core GPU and up to 96 GB of unified memory.
Apple M2 Pro
The second gen 5 nm M2 Pro packs in 40 billion transistors (about 20% more than the M1 Pro) and offers 200 GB/s of unified memory bandwidth. The M2 Pro is available in both 10-core and 12-core CPU configurations with up to eight high-performance cores and four high-efficiency cores. The high-performance cores feature 32 MB L2 cache while the high-efficiency cores come with 4 MB L2 cache.
Apple said that the M2 Pro is up to 20% faster than the 10-core M1 Pro while Xcode compilation can be up to 2.5x faster than a Core i9-based 16-inch MacBook Pro.
The GPU in the M2 Pro (19 cores) now offers three more cores than the one in the M1 Pro that should result in faster image processing and console-quality gaming, according to Apple.
Apple M2 Max
Apple is taking things a step further with the M2 Max — this flagship SoC crams in 67 billion transistors or 3x more than that of the M2. The M2 Max doubles the memory bandwidth of the M2 Pro to 400 GB/s and supports 96 GB of unified memory.
The M2 Max uses the same 12-core CPU as the M2 Pro but bumps the GPU cores to 38 with larger L2 cache. Apple said that the M2 Max can be up to 30% faster than the M1 Max in GPU-based workloads.
Apple even goes to the extent of claiming that a 16-inch MacBook Pro with M2 Max can run workloads such as OTOY OctaneRender 2022.1 using a scene that requires 40 GB GPU memory. This while competing Windows laptops powered by Core i9 processors and Nvidia Quadro RTX 6000 with 24 GB VRAM or GeForce RTX 3080 Ti GPUs cannot even run the test, according to the company.
Speaking in terms of TFLOPs numbers, the M2 Max is rated to offer 13.6 TFLOPs of FP32 throughput, which puts it between an RTX 3060 (10.94 TFLOPs) and RTX 3070 (15.97 TFLOPs) Laptop GPU, or somewhat higher than an RX 6800M (11.78 TFLOPs).
Both M2 Pro and M2 Max offer a 16-core Neural Engine capable of 15.8 trillion operations per second. The M2 Pro offers a media engine that provides hardware acceleration for H.264, HEVC, and ProRes encode/decode while the M2 Max comes with two video encode and two ProRes engines. There seems to be no AV1 encode/decode support, however.
The new chips also get an updated image signal processor and next-gen Secure Enclave.