Skip to content
News

Apple’s new M6 and M5 Ultra chips, explained

Apple today introduced two new chips, the M6 and the M5 Ultra, along with a new Mac Mini and a new Mac Studio to put them in. Both machines ship on 22 September.

Normally, since Apple launched it’s own silicon line: Plain M chips go in the cheaper machines. Then more expensive devices get the Pro, Max and Ultra chips.

The M6

The M6 is Apple’s first chip made on a 2 nanometre process. The processor has 12 cores, two more than the M5, in three types:

  1. 2 super cores,
  2. 4 performance cores and
  3. 6 efficiency cores.

Super cores take single-threaded work, meaning tasks that cannot be split up and shared out, and Apple claims the M6 now offers the fastest single-threaded performance of any chip. Efficiency cores handle background jobs cheaply. Apple says multithreaded performance is 1.2 times better than on the M5 and 2.4 times than on the M1.

Graphics also goes to 12 cores, each with a Neural Accelerator built in for the matrix maths that AI models run on. Apple says peak graphics compute for AI are up nearly 30 percent above the M5. The chip carries a Dual 16-core Neural Engine, two AI engines rather than one, which Apple says software can use at the same time. Memory bandwidth reaches 170GB/s, and the M6 takes up to 32GB.

// On Google

Make tech-ish your favourite source

Star us on Google — you'll see more tech-ish in Top Stories and AI results.

1Tap Open Google below
2Find tech-ish.com in the list
3Tap to favourite us
Open Google

The M5 Ultra

Apple always Ultra chips by fusing smaller ones together over a link it calls UltraFusion. It usually has fused two. The M5 Ultra fuses four, which Apple says is a first for the M series.

Underneath are two M5 Max chips, each already made of two dies. The connection between them runs at more than 4.4TB/s, quick enough that software sees a single processor rather than four parts.

That buys up to 36 processor cores and up to 80 graphics cores, which Apple calls its most powerful graphics ever. Neural Accelerators reach an Ultra chip for the first time, and Apple is claiming 4.3 times the AI compute of the M3 Ultra.

The M5 Ultra also takes up to 512GB of unified memory at 1.2TB/s. For localised AI work that number is more important than the cores. A model has to fit in memory before it will run at all, and the big freely available models need hundreds of gigabytes.

No M6 Pro, no M6 Max

This time, accoridng to explanation by Bloomberg’s Mark Gurman in July, we won’t be seeing the M6 Pro and M6 Max. Apple is said to be skipping even the M6 Ultra to get the M7 out sooner.

“Apple had been planning major neural-processing upgrades for the M7 family and ultimately decided those improvements were important enough to justify accelerating the next generation rather than completing the M6 lineup,” he wrote in his Power On newsletter. He expects the M7 in the first half of 2027, the M7 Pro and M7 Max late that year, and an M7 Ultra in 2028.

Apple has not confirmed any of it. If Gurman has it right, the M6 spends about six months as Apple’s newest chip, and the Mac mini is the only machine that will ever ship with one.

The Analyst

The Analyst delivers in-depth, data-driven insights on technology, industry trends, and digital innovation, breaking down complex topics for a clearer understanding. Reach out: Mail@Tech-ish.com

Join the discussion

0 comments
posting as Kicheche Mwepesi

Anonymous by default — no sign-up or email needed. Prefer to be recognised? Add a name or email above, your call. We don't email you about replies, so do check back.

protected, no CAPTCHAs
Back to top button