Same Shell, 55% More GPU: Inside the iPhone 18 Pro's 2nm Jump

Externally the iPhone 18 Pro is close to a rerun of last year's phone. Underneath, the A20 Pro built on TSMC's 2 nm process is not: a 55 percent jump in a GPU benchmark, single-core scores that embarrass some desktop silicon, and a vapour chamber with three times the surface area of its predecessor's to feed it. The interesting figure is not the peak, though. It is what survived twenty minutes of load.

The Silicon Is the Product This Year

Start with the CPU result, because it is the one that reads strangely. On Geekbench 7, the A20 Pro's single-core performance lands in territory normally occupied by desktop processors, which is not a sentence anyone wrote about a phone five years ago. One caveat travels with the number: Geekbench 7 changed how it scores, so nothing here can be lined up against older Geekbench runs. Cross-version comparisons are now meaningless, and a lot of the internet will make them anyway.

The GPU side is less ambiguous. In 3DMark Steel Nomad Light, which simulates a game workload rather than measuring compute throughput, the A20 Pro came in 55 percent ahead of the A19 Pro in the iPhone 17 Pro. Against Qualcomm's Snapdragon 8 Elite Gen 5 for Galaxy, the chip inside Samsung's S26 Ultra that shipped in March, the margin is 33 percent. That is not a generational nudge. It is a different performance class in a chassis that looks almost identical to the old one.

Twenty Minutes Is Where It Matters

Benchmark peaks are easy. Sustained load is where phones fall apart, and it is where this design change pays off. Over a twenty minute loop of 3DMark Wild Life Extreme, the iPhone 18 Pro was still beating the iPhone 17 Pro's best case by the end of the run — the newer phone's throttled state sat above the older phone's peak. Both eventually settled at roughly 66 to 69 percent of their own starting performance. The S26 Ultra, for reference, drops to around 53 percent.

The hardware behind that is a vapour chamber whose surface area is triple what the 17 Pro's offered, sitting under a chassis Apple has deliberately made thicker two generations running. Thermal imaging with the phones face down on a wooden table, twenty minutes of benchmark then twenty minutes of cooldown, shows the heat spreading across the lower body rather than pooling in one corner. The comparison is not clean, and nobody should pretend otherwise: the 18 Pro is doing meaningfully more work during the same test, so it may well be drawing more power while running cooler. A 2 nm part is likely more efficient, but tuning and chassis changes moved at the same time, and separating those three variables is not possible from outside.

What the stress test does prove is the practical thing buyers care about: this phone holds a much higher steady state than the last one. Frames stay steadier in long sessions, exports finish without the fanless slide into molasses, and the delta is large enough that it survives the measurement caveats.

Battery Grew Less Than Runtime Did

Capacity barely moved on the smaller phone. The eSIM-only iPhone 18 Pro carries 4,288 mAh against 4,252 mAh last year, an increase of 0.8 percent, yet Apple claims roughly a tenth more streaming video runtime. Those two numbers only reconcile if the chip and the display are doing less work per frame, which is exactly what a 2 nm process and a more efficient GPU should deliver. The larger model got the more conventional treatment: 5,088 mAh to 5,567 mAh, a 9.4 percent gain, on top of the efficiency work.

Charging moved more visibly than capacity. The 17 Pro and 17 Pro Max could briefly pull 30 W and 38 W; the 18 Pro clears 40 W, and the 18 Pro Max holds 50 W for about ten minutes. Measured as energy delivered at the port, half the pack arrived in 15 minutes on the Pro and 16 on the Pro Max, against a claim of about 15 minutes — versus 20 minutes for the previous generation. Apple meets its own number here, which is rarer than it should be.

The Display Claim and the Measurement Gap

Brightness specifications are unchanged: 1,000 nits typical, 1,600 peak for HDR, and 3,000 nits in outdoor conditions that require auto-brightness and direct sun. Measured with a luminance meter, the panel came within 130 nits of the outdoor figure and 50 to 80 nits below what the 17 Pro managed. That gap is small enough to be explained by auto-brightness behaviour, panel temperature or a restart, and it is the sort of shortfall that later testing often closes. In ordinary indoor use the ceiling is around 800 nits, rising to 1,500 or 1,600 for HDR highlights.

There are also reports of a faint magenta cast on some units. Nothing in the specification changed, and the most plausible explanation is calibration variation between panels rather than a design fault — the same lottery that has followed every recent iPhone launch. One more thing worth knowing before trusting any early number: iOS 27 runs an "Optimizing Search and Siri" indexing pass after an update, with no way to force it to finish. Benchmarks here were taken after it completed; the battery discharge test ran while it was still active, and the results lined up with Apple's claims anyway.

What I Think

This is the least interesting iPhone in years on the outside and one of the more interesting ones underneath. The case for upgrading does not rest on the camera bump or a new colour; it rests on the fact that the phone no longer falls off a cliff after five minutes of load, and that the charge curve finally got the wattage it deserved. Those are the two things you feel every day and neither shows up in a spec sheet comparison.

If you are on a 17 Pro, the honest answer is that you are paying a premium for thermal headroom and a faster charge, and only you know whether that is worth it. If you are on anything older, the 2 nm jump is large enough that the phone will feel like a different category of device. Apple's real problem is that the best part of this release is invisible from across a table, and invisible does not sell phones.