Intel claims its 14A process node will deliver performance within 5% of TSMC's A14 technology, addressing competitive gaps in advanced chip manufacturing.

Intel's 14A (1.4nm-class) process technology is expected to deliver performance 'within 5%' of TSMC's A14 (1.4nm-class) production node, according to Naga Chandrasekaran, chief technology and operations officer and general manager of Intel Foundry, in comments to investment banking firm KeyBanc. While this statement may sound encouraging given TSMC's steady track record of PPA gains with each new node, it warrants closer examination.

The phrase 'within 5%' is ambiguous—it could mean 14A performs 5% faster or 5% slower than A14. More importantly, while matching a rival's performance within a narrow margin might appear competitive, Intel's process nodes have historically lagged TSMC's in transistor density, even when remaining competitive in performance and power. This gap has partly reflected Intel's traditional emphasis on CPU performance over maximum transistor density, an approach enabled by its own fabrication capacity offsetting the cost of larger dies.

Real-world processor data offers a useful reference point. Intel's 18A at maximum CPU frequency slightly leads TSMC's N2: Intel's Core Ultra 9 388H reaches 5.10 GHz (at 80W max turbo power), compared to AMD's EPYC 9586F's 5.0 GHz peak (at 500W default CPU power), Apple's A20 Pro at 4.93 GHz, and Apple's M6 at 4.78 GHz. These processors employ different architectures, voltages, standard-cell libraries, thermal envelopes, and physical implementations, so direct performance comparisons are unwarranted. Nevertheless, they establish that available N2 processors do not demonstrate a substantial frequency advantage over 18A—if anything, the highest observed CPU frequencies favor Intel's process.

Intel states that 14A should deliver 15–20% higher performance at the same power compared to 18A, or 25–35% lower power at equivalent frequency and transistor count. TSMC projects A14 to be 10–15% faster than N2 at matched power, or 25–30% lower power at the same clocks and transistor count.

Combining observed 18A and N2 CPU frequencies with Intel's stated 14A gains and TSMC's assumed A14 gains suggests a modest 14A performance advantage in even conservative scenarios. Intel's new 'within 5%' assertion is therefore notably less ambitious than its published process specifications imply, though the statement does not clarify which process it expects to lead.

Intel's potential 14A advantage over TSMC's A14 can be estimated using the highest observed 18A and N2 CPU clocks combined with each company's official iso-power performance projections:

| Scenario | 14A gain vs. 18A | A14 gain vs. N2 | 14A extrapolation from 5.10 GHz | A14 extrapolation from 5.00 GHz | Implied 14A advantage |

|----------|------------------|-----------------|----------------------------------|----------------------------------|----------------------|

| Intel worst, TSMC best | 15% | 15% | 5.865 GHz | 5.75 GHz | 2.00% |

| Both minimum gains | 15% | 10% | 5.865 GHz | 5.5 GHz | 6.60% |

| Both maximum gains | 20% | 15% | 6.12 GHz | 5.75 GHz | 6.40% |

| Intel best, TSMC worst | 20% | 10% | 6.12 GHz | 5.5 GHz | 11.30% |

*Starting points: Intel 18A = 5.10 GHz (Core Ultra 9 388H); TSMC N2 = 5.00 GHz (EPYC 9586F).*

With these devices as reference points, the companies' official iso-power performance projections imply a 2–11.3% potential performance advantage for 14A over A14, depending on process-performance assumptions.

Using Apple's M6 as an alternative N2 reference point yields similar results:

| Scenario | 14A gain vs. 18A | A14 gain vs. N2 | 14A extrapolation from 5.10 GHz | A14 extrapolation from 4.78 GHz | Implied 14A advantage |

|----------|------------------|-----------------|----------------------------------|----------------------------------|----------------------|

| Intel worst, TSMC best | 15% | 15% | 5.865 GHz | 5.506 GHz | 6.50% |

| Both minimum gains | 15% | 10% | 5.865 GHz | 5.267 GHz | 11.40% |

| Both maximum gains | 20% | 15% | 6.12 GHz | 5.506 GHz | 11.20% |

| Intel best, TSMC worst | 20% | 10% | 6.12 GHz | 5.267 GHz | 16.20% |

*Starting points: Intel 18A = 5.10 GHz (Core Ultra 9 388H); TSMC N2 = 4.78 GHz (Apple M6).*

With the M6 as reference, calculations suggest Intel 14A could hold a 6.5–16.2% advantage over TSMC A14. Even under the most conservative scenario for Intel—where 14A achieves only +15% while A14 reaches the full +15%—Intel's node substantially exceeds the 'within 5%' estimate Chandrasekaran provided.

It must be noted that these calculations do not predict actual 14A or A14 CPU frequencies; rather, they apply the companies' published process gains to current processor clocks. The exercise primarily illustrates what the published numbers imply relative to today's products.

Intel's 'within 5%' assessment raises a puzzling question: why does Intel expect such a narrow margin when its published process specifications suggest a wider gap? The discrepancy might reflect factors these calculations overlook, or it might suggest that the head of Intel Foundry, like his boss Lip-Bu Tan, prefers to underpromise.