Many youngsters, born safely after the dawn of the new millenium, do not have proper context to appreciate the modern miracle of the smartphone, or what i call the “iPhone vs Supercomputer” comparison. The fact is, that the phone you carry in your pocket literally packs the computing punch of the world’s fastest supercomputer of y2k.
In other words, when IBM built their famous Deep Blue with the sole purpose of defeating the best chessplayer on earth, today any smartphone on the planet can beat any human chess player, without even breaking a sweat. And we’re not even talking cloud compute here. Your phone can accomplish that all on its own, no net connection necessary.
While the raw compute power of your smartphone has seen insane levels of improvement (and power efficiency) — with a 2023 flagship phone having roughly 10,000 times the computing power of a 2003 device, housed in a roughly identical form factor — the functions that humans use it for have (not yet) changed substantially — we text, we (less and less) call, we doomscroll the social feeds, we browse the “web.” With all this extreme power in our pockets, it would seem that times are ripe for a change.
Simple fact: an iPhone 14 Pro Max has more raw on-board processing power than the top 500 supercomputers in the world, combined, harnessed in 1995. (for calculation details, see below)
iPhone vs Supercomputer, 2025 update:
an iPhone 16 Pro Max has the computing power equivalent to ASCI RED, which was the fastest supercomputer in the world in 1999. But really, that only scratches the surface.
Lets see how far we’ve really come in a brief span of 25 years:
| ASCI Red Supercomputer |
iPhone 16 Pro Smartphone |
||
| speed (FLOPs) | 3.1 teraFLOPs | 2.3 teraFLOPs | |
| processors | 9,632 Pentium Pros | 1 (one) Apple A18 Pro | |
| fits in | a large building | your pocket | |
| physical size | 1,600 sq. feet (approx 4 tennis courts) |
0.2 sq. feet | |
| power draw | 850,000 watts (850 kilowatts) |
1 watt (yes, ONE watt) |
|
| total transistors | 55 billion | 22 billion | |
| unit cost | $130 million | $1,300 | |
| manufacturer | Intel | Apple | |
| launch | 1999 | 2024 |
Put another way:
the iPhone, compared to the world’s fastest supercomputer in 1999, has:
- 1/10,000 the number of processors,
(yet still half the transistors!!!) - is about 1/100,000th the physical size,
- costs 1/100,000th as much to build
- consumes 1/1,000,000th (one-one-millionth)
the electrical power,
and yet… - has almost identical computational ability.
Excerpting from AI Circuit Design Software: Designing the Designers:
“I stared at my 9-pound state-of-the-art Sony laptop. Considered the chip controlling all of its awesome functions. Some 10 million transistors on that Pentium III, running about 900 million instructions per second, (MIPS)…
for comparison: my 1981 IBM PC (Intel 8088) had 30,000 some transistors, topped out at 5 mHz, and did about 750,000 instructions per second (1/1000th the speed).
My 2022 iPhone 14 Pro’s A16 Bionic chip has some 16 billion transistors, runs at a clock speed of 3.5 GHz, and performs, at peak, around 17 trillion operations per second (that’s the equivalent of about 20 million times the power of my ’81 PC, or 20,000x as fast as my sony laptop.. yet, fits in my pocket).”
Supercomputers are generally defined as the highest performance monolithic computing platforms in the world, which are contained within a single physical location and do not rely on externally networked machines for raw compute power (though they are often tasked and managed by network connected clients). Since Moore’s Law (the doubling of compute power every 18 months) is in effect, the definition of supercomputer, in terms of performance, is a continually moving bar:
(yes, this chart is 2011… but it has the best data labels. The trend has continued unabated ever since.)
These are the rough conclusions from the chart, FYI:
- The fastest supercomputer in the world gets 100x faster about every 10 years
- The fastest supercomputer in any given year is 100x faster than the 500th fastest
- The sum total compute power of ALL the 500 fastest supercomputers on earth is generally only 10-50x faster than the fastest all by itself.
- within 10 years, the fastest will have been bumped 500 places down the list
- In any given year, the world’s fastest supercomputer is roughly 100,000 times faster than a high-end consumer laptop computer. Put another way, it would take 100,000 laptops, networked, to even approach the world’s fastest supercomputer (not, in fact, an unreasonable scenario… thus, cloud computing)
In a separate article, we compare the performance of traditional monolithic supercomputers to the new powerhouse of global compute: the AI datacenter.
- Supercomputers vs. AI Datacenters : Power, Performance, and Scalability
This might make us ask: given that an iPhone is essentially a mobile supercomputer, and that AI datacenters far outstrip the performance of any single supercomputer on earth, what is the purpose of supercomputers today?
.
prompt:
“draw me a political cartoon. the image is of an anthropomorphized iPhone, looking haughtily at a massive supercomputer (racks of servers fading to infinity), and laughing in contempt (‘BWAHAHAA!!!’)”
engine:
DALL-e3 (OpenAI)