Apple, of course, used all the words and phrases you'd expect from the new line of M3 chips: groundbreaking, dramatic performance improvements, revealing new features, biggest step forward, etc.
The company also shared some graphs indicating improved performance of the 3nm M3 chips compared to their predecessors …
Here are some edited excerpts of what Apple said:
Apple today announced M3, M3 Pro and M3 Max—three chips with revolutionary technologies that deliver dramatic performance improvements and new possibilities for Mac. These are the first personal computer chips built using industry-leading 3nm technology, enabling more transistors to fit into less space for greater speed and efficiency […]
The M3 family of chips feature a next-generation graphics processor that represents the biggest advance in graphics architecture for Apple processors. The GPU is faster and more efficient, features a new technology called dynamic caching, and brings new rendering features like hardware ray tracing and mesh shading to the Mac for the first time […]
“Apple Silicon has completely redefined the Mac. Every aspect of its architecture is designed for performance and energy efficiency,” said Johnny Srouji, Apple's senior vice president of Hardware Technologies. “With 3nm technology, next-generation GPU architecture, a more powerful processor, a faster Neural Engine and support for even more unified memory, M3, M3 Pro and M3 Max are the most advanced chips ever made for personal use.
Arstechnica dived into the hype to take an in-depth look at how the M3 line of chips stacks up against previous generations of Apple Silicon. Let's start by looking at the specs:
The M3's performance improvement over the M2 is largely consistent with the M2's performance improvement over the M1.
Apple claims peak GPU performance on the M3 series is 65 percent faster than the M1, although as with CPU performance, it will likely vary depending on which chips you're comparing. In our tests, the various M2 series GPUs are typically 25 to 30 percent faster than their M1 equivalents, so the year-over-year performance gains from M1 to M2 to M3 are fairly even. Apple says the M3 has 2.5 times the rendering performance of the M1 for workloads that take advantage of new hardware acceleration features.
The M3 Pro doesn't do that. It looks like it will get the same improvements as the base M3 and M3 Max.
Compared to the M2, the good old M3 doesn't have any extra cores, so it will rely solely on architectural improvements and clock speed increases to improve performance – and it has 25 billion transistors, 5 billion more than the M2, so… 8217; There is still quite a lot of new equipment here […]
The M3 Pro is a 37 billion transistor chip, 3 billion less than the M2 Pro. This makes sense if you look at the number of cores; The M2 Pro had eight performance cores and four efficiency cores, as well as a whopping 19 GPU cores. The M3 Pro still has 12 cores, but it's evenly split between six performance cores and six efficiency cores, and the maximum number of GPU cores goes up to 18. Maximum memory increases slightly, from 32 GB to 36 GB.
I still expect the M3 Pro to be an upgrade over the M2 Pro due to the updated architecture, but it looks like a lesser upgrade than the M3 (which keeps the core count the same) or the M3 Max (which increases it). The M2 Pro and M2 Max used the same CPU core configuration, and I think the company wanted to create an extra incentive to upgrade from Pro to Max for people who don't care about GPU performance.
Regarding M3 Max's 92 billion transistor count is a huge jump from the M2 Max's 67 billion. This is largely due to the M3 Max processor, which includes 12 performance cores and four efficiency cores, which is four more performance cores than the M2 Max. The GPU is also getting a little bigger, with the maximum number of cores increasing from 38 to 40. The maximum memory is also increasing from 96 GB to 128 GB.
Of course we'll do that. Find out more about how significantly improved the M3 chips are when we start seeing benchmarks as well as real-world usage reports.
