AN INTERACTIVE MEASUREMENT · 1969 → TODAY
How far did computing come? Put the computer that flew Apollo next to the computer in your pocket — and measure it.
01 / THE GAME
Every chapter of this article plays the same game:
Here is one thing. Now use it as the
unit.
How many of those fit into the other thing?
The machines never change. The unit does. Four scales, four ratios — and every one is computed from published specifications, never invented for effect.
There is no synthetic “power score” anywhere in what follows. And where no honest comparison exists — raw computation — no number is offered at all.
Both machines, one true scale. The AGC's front is 61 × 32 cm — a piece of furniture. The phone's is 16.34 × 7.8 cm. The phone is already the small object, and we haven't even turned either one sideways.
Depth: the AGC's enclosure runs 17 cm deep. The phone is 8.75 mm — about 19× thinner. One is a box you bolt into a spacecraft; the other is nearly a plane.
Make the AGC's box transparent and fill it with phone-volumes. About 298 phones of volume fit into the AGC's bounding box (33.18 L vs 111.5 mL) — you just watched the multiplication happen.
≈298× less bounding-box volumeVolume told us how much space the machines occupy. Mass tells us how much material we're carrying. Put one of each on the scale: 31.8 kg against 233 g. The beam doesn't hesitate.
Keep adding phones — one, ten, fifty, a hundred — until the beam levels. It takes 136 phones to balance one Apollo Guidance Computer.
Not “136× lighter” as an abstract number — one AGC weighs approximately the same as 136 iPhones. The ratio appears on its own, from the unit you chose.
≈136× less massSo the box shrank and the weight vanished. Now the part the eye can't hold.
The AGC had 2,048 words of erasable memory — 15 bits per word, 3,840 byte-equivalents. One square = one word. This grid is the machine's entire working memory, and every address in it was a physical budget an engineer allocated by hand.
Collapse the whole square into one unit: a complete Apollo memory. From here on, that orange square is what we count with.
Replicate the unit until the screen is a wall of Apollo memories: 1,024 complete AGC memories. A room full of guidance computers, in writable memory alone.
The entire wall collapses into one new tile. Repeat: this wall holds 1,048,576 Apollo memories — each tile is the wall you just watched build. The unit keeps swallowing its own predecessor.
A blue marker for the phone's reported 11.42 GB. It lands at ≈2.84 walls — about 2.97 million complete Apollo memories. The unit you built twice wasn't even close to big enough.
≈2.97 million× the writable memoryUNITS → MEANING
That ≈2.97 million is not just a big number — it's a different relationship between a programmer and a machine. In 1969, memory was something engineers counted almost word by word: every one of the AGC's 2,048 erasable words had a job, and the ledger of those jobs was reviewed like cargo manifest.
Today, memory is something software treats as an abundant resource — allocated, discarded, and never mourned. The unit had to become “an entire Apollo computer” before the two worlds could even share a scale. Without that unit change, the number is just unit conversion.
Give the AGC one second: its clock ticks 1,024,000 cycles and it completes about 41,700 instructions on average (≈24 μs each). That entire second is the orange bar — drawn small on purpose. Remember the scale.
The phone's reported 4.26 GHz clock ticks 4.26 billion times in that same second — so at the orange bar's scale, the blue bar doesn't fill its lane. It leaves your screen, hundreds of screen-widths long (the count under it is computed for your window). And to keep this honest: clock cycles aren't instructions, and this bar is one core's reported peak — the phone has five more CPU cores and twenty-two specialized engines. We don't multiply peak clocks by core counts: a peak is not a promise of sustained all-core speed.
About 4,160 phone cycles elapse for every single AGC cycle — and that's one peak core, so if anything it understates the machine. You don't need to understand microarchitecture to read this picture: the temporal scale of computing has exploded.
≈4,160× the reported frequencyWe can't honestly compare their speed with one number — but we can ask how much computing infrastructure each one packs into its physical body. Two objects. Everything below is printed specification, no inference.
The AGC: 33 litres, 31.8 kilograms, one processor at 1.024 MHz. The phone: 111.5 mL, 233 g, and 6 CPU + 6 GPU + 16 Neural Engine cores at a reported 4.26 GHz. The big card holds one engine; the little one holds twenty-eight.
This is the more interesting historical fact than any FLOPS count: computing didn't just get faster — it got dense enough to vanish inside an everyday object. The computer stopped being the thing you noticed.
28 COMPUTE CORES · 111.5 mLStart with the AGC: 33.18 litres, 31.8 kilograms, 2,048 words, 1.024 MHz, ≈41,700 instructions per second. Now let each number become the object it was in its chapter — and bring in the phone.
One AGC bounding box ≈ 298 phone-volumes (33.18 L vs 111.5 mL). The blue silhouettes you counted in the transparent box, again.
One AGC ≈ 136 iPhones on the scale (31.8 kg vs 233 g). The pile that leveled the beam.
One orange square — one complete Apollo memory — against the phone's ≈2.84 walls: ≈2.97 million AGC writable memories in 11.42 GB (reported).
Sparse orange pulses against a solid blue band: ≈4,160 phone cycles for every AGC cycle — a frequency comparison, labeled as one. Four defensible scales. No magic number.
ONE GAME · FOUR RATIOS · NO MAGIC NUMBERTHE LEDGER — EVERY NUMBER IN ONE PLACE
Each row answers one question. None of the ratios should be multiplied together — they describe different quantities, and one of them (computation) has no honest ratio at all.
| Scale | Apollo Guidance Computer | Modern phone | Ratio |
|---|---|---|---|
| Volume | 33.18 L 61 × 32 × 17 cm bounding box | ≈111.5 mL 16.34 × 7.8 × 0.875 cm | ≈298× smaller box |
| Mass | 31.8 kg 70.1 lb | 233 g published specification | ≈136× lighter |
| Writable memory | 3.75 KiB 2,048 × 15-bit words | 11.42 GB reported sample | ≈2.97M× more capacity |
| Clock | 1.024 MHz AGC processor | 4.26 GHz reported sample | ≈4,160× higher frequency |
| Computation | ≈41,700 instr/s average; ≈24 μs average instruction | not published only benchmark scores exist — Geekbench 6 multi-core ~9,000+, varies by sample | — no honest single ratio |
And this is where every
“which one is more powerful?”
argument quietly falls apart.
THE TAKEAWAY
In 1969, computing was a machine: a 31.8-kilogram box, bolted into the spacecraft, its 2,048 words of memory budgeted word by word, its program woven by hand into rope. It was cargo. It had a mass, a serial number, and a crew that thought about it.
Today, computing is a layer underneath everything. The phone in your pocket carries millions of Apollo memories and twenty-eight compute cores as a matter of course — and nobody, reading this on one, gave that a second's thought until just now.
The computer that went to the Moon
became invisible.
RESEARCH & METHOD
METHOD — Every ratio on this page is computed by the page's own script from the published figures above; the stage readouts are live counts, not decorations. Volume is bounding-box volume, not usable internal space. Memory compares the AGC's 15-bit erasable words (as byte-equivalents) against reported RAM. Clock is a frequency comparison only — clock cycles are not instructions. The clock bars compare one reported peak phone core against the AGC's single clock; core counts are never multiplied into frequencies (peak is not sustained all-core). Transistor counts and energy efficiency are deliberately absent: no authoritative like-for-like figures exist for both machines, and four defensible scales beat seven questionable ones.