Skip to content

Topic

Semiconductor Process Nodes

Manufacturing generations (e.g., Intel 18A, TSMC N2) defining chip transistor density, power efficiency and performance, used to benchmark fabs and products.

Current stories

investConfirmed3 publishers

CXMT's two 50 per cent claims multiply to 2.25 times the bits per wafer

Changxin Memory says its fifth-generation platform is in mass production at 11.95nm with at least 50 per cent more dies per wafer. Both of the figures that would let a buyer price it come from the company.

Perspective Coverage

3 publishers
Builder
Builder 38%
Operator
Operator 19%
Investor
Investor 43%

Reality

Evidence40
Adoption45
Hype gap+35
Incentives70
Confidence58
buildOne report1 publisher

Apple ships its whole M6 family as one monolithic N2 die

A dev.to post says Apple announced the M6 on August 25 as a single die on TSMC's N2 and put the top of its lineup on a quad-die M5 Ultra built on last year's N3P, leaving a Pro-class refresh with no part to order.

Publishers:dev.to

Reality

Evidence30
Adoption15
Hype gap+30
Incentives55
Confidence35
buildOne report1 publisher

Apple's A20 Pro reportedly hits 4,006 single-thread in early Geekbench 7 test

One early Geekbench 7 submission reported by Tom's Hardware puts the six-core chip 32% ahead of Intel's Core i9-14900KS in single-thread, while its multi-thread total works out to 2.86 times that score.

Publishers:tomshardware.com

Reality

Evidence30
Adoption18
Hype gap+45
Incentives55
Confidence55
productOne report1 publisher

Apple spreads the AI math across all six cores of its first 2nm iPhone chip

Every performance figure Apple has published for the A20 Pro is a peak number, and none of them tells a product team how long a local model takes on its fourth run, or how large a model fits in memory.

Publishers:pcmag.com

Reality

Evidence38
Adoption15
Hype gap+35
Incentives72
Confidence45
investOne report1 publisher

Apple answers the premium question with a titanium hinge priced at $1,999

John Ternus opened his first keynote eight days into the job with a claim about where computation happens rather than what the model can do. The product behind that claim is a $1,999 foldable engineered around a hinge.

Publishers:techtimes.com

Reality

Evidence52
Adoption12
Hype gap+20
Incentives60
Confidence50
productOne report1 publisher

Apple's quad-die M5 Ultra puts 512GB of inference memory under the desk

Apple says the M5 Ultra reaches 512GB of unified memory at 1.2TB/s, which settles how big a local agent can get, though the September launch arrives without prices and every AI speed figure is measured against an M1.

Publishers:pcmag.com uk.pcmag.com

Reality

Evidence32
Adoption20
Hype gap+45
Incentives60
Confidence45
productOne report1 publisher

Arm claims 1.7x on AI models by doubling the matrix units inside its CPU cluster

The uplift runs through the regular instruction pipeline, so adopting it costs a developer a compile step. That changes what an accelerator-specific inference path has to prove before a team keeps paying to maintain it.

Reality

Evidence38
Adoption15
Hype gap+34
Incentives70
Confidence44
productOne report1 publisher

Apple debuted its first 2nm chip in an $899 Mac mini instead of an iPhone

The M6 reached TSMC's N2 process before the A20 Pro did, pinning down what the iPhone 18 Pro's chip is built on. Its memory bandwidth number tells on-device AI teams how much room next year's phone actually adds.

Publishers:macrumors.com

Reality

Evidence58
Adoption30
Hype gap+12
Incentives62
Confidence55
productOne report1 publisher

Apple's M6 is a 2nm inference upgrade with a CPU refresh attached

The published deltas put the Neural Engine at 2x and multithreaded CPU at 1.2x. Apple's own workload figures widen that gap further, and the lineup stops at base tier.

Publishers:macrumors.com

Reality

Evidence44
Adoption22
Hype gap+28
Incentives68
Confidence41
buildOne report1 publisher

Intel dates Diamond Rapids: 256 P-cores, 1.28 GB of cache, and 2027 to plan around

Intel's Hot Chips disclosure gives capacity and licensing planners real numbers for 2027, but leaves out the one figure that would let anyone model per-core performance.

Publishers:tomshardware.com

Reality

Evidence54
Adoption9
Hype gap+14
Incentives63
Confidence55