Apple's new M6 processor is offering an early look at what TSMC's 2nm class manufacturing process could mean for future PC CPUs and GPUs.
The chip is built on TSMC's N2 process and early testing points to meaningful improvements in both performance and efficiency compared with the previous M5 generation.
That matters beyond Apple's own hardware. AMD, Nvidia and potentially Intel are all expected to use advanced manufacturing nodes for future products, and the M6 provides one of the first practical examples of how TSMC's latest process performs in a shipping consumer processor.
M6 posts large single core gains
In Geekbench 6.7 testing, the Apple M6 reportedly reaches 4,698 points in single core performance.
For comparison, AMD's Ryzen AI 9 HX 470 scores around 2,872 points in the same benchmark.
The M6 is also roughly 10 percent faster than Apple's previous M5 in single core testing.
| Processor | Geekbench 6.7 single core |
|---|---|
| Apple M6 | 4,698 |
| AMD Ryzen AI 9 HX 470 | 2,872 |
| Apple M5 | About 10% behind M6 |
Similar improvements have also been seen in Cinebench testing.
These comparisons should be treated carefully because Apple and AMD use very different architectures, operating systems and platform designs. However, the results still provide evidence that Apple's move to N2 has not created obvious performance or efficiency problems.
That is important because shrinking to a new manufacturing process no longer guarantees dramatic gains.
Multi core performance improves even more
The generational improvement becomes larger in multi threaded workloads.
Apple increased the M6 CPU from 10 cores in the M5 to 12 cores, which contributes heavily to the performance increase.
Overall CPU performance can be around 50 percent higher than the previous generation in workloads that make effective use of all available cores.
That gain therefore cannot be attributed entirely to TSMC's N2 process.
The larger core count and architectural changes are also important.
Still, the fact that Apple can add more CPU resources while maintaining strong efficiency is another encouraging sign for the manufacturing technology.
Efficiency may be the most important improvement
The M6 reportedly offers nearly 20 percent better performance per watt than the M5.
Compared with the Ryzen AI 9 HX 470, testing suggests the M6 can deliver around twice the performance per watt in some workloads.
Efficiency improvements are particularly valuable for laptops because they can translate into better battery life, lower temperatures or higher sustained performance within the same power envelope.
They also matter for desktop processors and graphics cards, where rising power consumption has become an increasingly important design constraint.
TSMC's N2 process uses a new transistor architecture and is intended to improve density and efficiency compared with previous nodes.
The M6 results suggest those improvements are translating into real products.
AMD could bring N2 to Zen 6
AMD is one of the most likely PC companies to benefit from the process relatively soon.
The company has already confirmed that some next generation Zen 6 server processors are being manufactured using TSMC N2.
If desktop and laptop Zen 6 products also use N2, AMD could move directly from its current process technologies to 2nm class manufacturing without extensively using TSMC N3 for mainstream PC CPUs.
That remains dependent on the exact Zen 6 product configuration, since AMD can use different manufacturing nodes across chiplets and product families.
Still, server production establishes that AMD has already begun working with N2 silicon.
Nvidia may take longer to reach 2nm
Nvidia's timeline could be further away.
Current Blackwell graphics processors use TSMC's N4 family, while next generation Rubin products are expected to move to N3.
If Nvidia continues advancing one major node at a time, its first N2 based graphics processors may not arrive for several years.
Some current expectations place future consumer architectures using N2 closer to the end of the decade, although exact product schedules remain uncertain.
AI demand could also affect those plans.
Advanced manufacturing capacity is extremely valuable, and Nvidia may prioritize data center accelerators before allocating large volumes of cutting edge silicon to consumer GeForce products.
Intel has its own advanced process option
Intel is taking a different route because it also operates its own semiconductor manufacturing business.
Its 18A process is generally considered part of the same broad advanced generation as TSMC N2, although direct comparisons between process node names are imperfect.
Industry estimates often place TSMC N2 ahead in transistor density, but final product performance depends on far more than density alone.
Intel has also used TSMC extensively for recent processors.
Lunar Lake already uses TSMC N3 for important compute components, and some future Nova Lake variants could potentially use N2.
That means Intel may also benefit directly from TSMC's latest manufacturing technology even while developing its own 18A products.
2nm PCs may still take time
The main limitation is availability.
Advanced semiconductor capacity is under heavy pressure from AI accelerators and server processors, which generally offer chipmakers much higher margins than mainstream consumer products.
AMD, Nvidia and other companies may therefore prioritize enterprise hardware before deploying large volumes of N2 silicon in gaming PCs.
Cost is another concern.
New manufacturing nodes are increasingly expensive, which could make early 2nm PC chips premium products.
Even so, the Apple M6 provides an encouraging early indication.
Its combination of higher CPU performance and improved efficiency suggests TSMC N2 is delivering meaningful gains rather than functioning mainly as a marketing transition.
PC gamers may have to wait before 2nm CPUs and GPUs become common, but the first consumer results suggest the technology has the potential to provide one of the more significant semiconductor upgrades of the next hardware generation.
Discussion (0)
Be the first to comment.