- The M5’s 12x neural performance leap marks Apple’s biggest architectural leap
- Dedicated neural accelerators in each GPU core redefine how Apple handles AI processing
- The M5 Pro, Max and Ultra are expected to push neural throughput dramatically
Apple’s M-series chips have seen regular performance gains over the past five years, but the leap from M4 to M5 stands out by far the most.
The latest-generation chip changes how Apple handles AI workloads, delivering an increase in neural compute far beyond anything seen before in the tech giant’s own silicon.
When Apple’s first chip, the M1, arrived in November 2020, its neural engine could handle about 11 trillion operations per second. The M2 pushed that to just under 16, and the M3 climbed to around 18. When the M4 arrived last October, the figure was double that.
Inside Apple silicon: Part two of a five-part series on the M-class processors
This article is the second in a five-part series that takes a deep dive into Apple’s M-class processors, from the early M1 to the recently announced M5 and our expected M5 Ultra. Each piece will examine how Apple’s silicon has evolved in architecture, performance and design philosophy, and what those changes could mean for the company’s future hardware.
The top of the TOPS
With the M5, the number has increased to around 133 TOPS, or around twelve times the M1’s starting point.
This upscaling is the sharpest increase in Apple’s internal processor history. Instead of relying solely on a faster Neural Engine, the M5 has a dedicated Neural Accelerator inside each GPU core.
This lets the graphics hardware take on AI workloads directly, distributing inference tasks across the chip rather than pushing them through a single engine.
The result is a system that handles model-based processes much more efficiently.
Functions such as on-device transcription, local image generation or creative tools that rely on Apple Intelligence all benefit from the new structure.
Each part of the chip now contributes to neural processing, making the overall speed increase look less like a step and more like a jump.
On paper, the rest of the chip has also been improved. The 10-core CPU delivers about 15 percent faster multithreaded performance than the M4, and unified memory bandwidth increases to 153 GB/s. This supports larger models and more efficient multitasking without increasing power consumption.
The M5 is inside the new 14-inch MacBook Pro and the new iPad Pro. The tablet version uses either a nine-core or ten-core CPU, depending on storage space, but both share the same Neural Engine and GPU layout.
Looking beyond the chips that Apple has actually released, the projected numbers for potential future versions suggest just how far this design can stretch.
Estimates from Google Gemini suggest that an M5 Ultra chip could reach between 600 and 800 TOPS, with Pro and Max variants falling between 190 and 320.
None of these chips have been announced (neither has the M4 Ultra, for that matter – the M3 Ultra was only announced earlier this year and is in Mac Studio), and the numbers are just projections, but they follow the growth pattern seen in previous generations, so there’s a solid foundation for them.
Such increases would inevitably raise familiar problems. A desktop-class M5 Ultra would need more cooling and power than Apple’s current compact cases could handle.
What the new M5 shows is that Apple’s chip roadmap is now shaped by neural performance more than raw CPU or GPU power. The company has linked its future Macs and iPads to on-device AI. The next few generations will determine how far it can scale before physics and thermals catch up.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, video unboxings, and get regular updates from us on WhatsApp also.
