| Zettabits per second (Zbit/s) | Tebioctets per second (Tio/s) |
|---|---|
| 1 Zettabit per second | 113686837.722 Tio/s |
| 2 Zettabits per second | 227373675.443 Tio/s |
| 3 Zettabits per second | 341060513.165 Tio/s |
| 4 Zettabits per second | 454747350.886 Tio/s |
| 5 Zettabits per second | 568434188.608 Tio/s |
| 10 Zettabits per second | 1136868377.22 Tio/s |
| 20 Zettabits per second | 2273736754.43 Tio/s |
| 25 Zettabits per second | 2842170943.04 Tio/s |
| 50 Zettabits per second | 5684341886.08 Tio/s |
| 100 Zettabits per second | 11368683772.2 Tio/s |
| Reference | Zettabits per second (Zbit/s) | Tebioctets per second (Tio/s) |
|---|---|---|
| A dial-up modem | 5.6 × 10-17 Zbit/s | 0.00000000636646 Tio/s |
| Typical home broadband | 1 × 10-13 Zbit/s | 0.0000113687 Tio/s |
| Gigabit Ethernet | 1 × 10-12 Zbit/s | 0.000113687 Tio/s |
| Streaming a 4K film | 2.5 × 10-14 Zbit/s | 0.00000284217 Tio/s |
The zettabit per second is a unit of data transfer rate equal to a thousand exabits per second. Its symbol is Zbit/s. No system on Earth moves data at this rate, and none is planned; the unit exists because the metric system defines every prefix for every unit, whether or not the combination has yet been needed.
To see how far off it is, take the whole internet. Global traffic at present runs at roughly one exabit per second on average, so the entire planet's communications would have to grow a thousandfold to reach one zettabit per second. At the growth rates of the last two decades that would take somewhere between twenty and thirty years, which is precisely the sort of extrapolation that has been wrong in both directions before.
A zettabit per second is 125 exaoctets per second. Since global data storage manufacturing runs at a few hundred exaoctets a year, a link at this rate would transfer the world's entire annual production of new storage capacity in a couple of seconds. Nothing could be stored at the far end; the data would have to be processed and discarded as it arrived.
That last point is not as fanciful as it sounds. Several existing systems already discard almost everything they receive: particle detectors, radio telescope arrays and network monitoring systems all process far more than they keep, because keeping it is impossible and unnecessary. A zettabit-per-second link would be an extreme case of an architecture that already exists.
The physical obstacles are less absolute than they might appear. The theoretical capacity of a single optical fibre is far above what is used today, and the practical limits come from amplifier noise, non-linear effects and the electronics at each end rather than from the glass itself. Aggregating enough fibres would reach a zettabit per second; the difficulty is that nobody has a reason to.
For a converter, the unit matters because forecasts and capacity models are written in whatever unit keeps the numbers legible. A projection that reaches into the 2050s may reasonably state totals in zettabits per second, and a reader needs to be able to convert that into something familiar.
One zettabit per second equals 1,000 exabits per second, 125 exaoctets per second, or about 0.8470 zebibits per second.
The tebioctet per second is a unit of data transfer rate equal to 1,024 gibioctets per second, or two to the fortieth power octets per second. Its symbol is Tio/s. It is the binary counterpart of the teraoctet per second, and the two differ by 10 per cent.
Nothing outside a large machine moves data this quickly. The rate describes the memory bandwidth of an accelerator with stacked memory, the internal fabric of a high-end processor package, or the aggregate throughput of a parallel filesystem spread across thousands of drives. All of these are built from binary structures, and their totals are binary quantities divided by time.
High-performance computing is where the unit is written most often. A supercomputer's storage system is specified by how many tebioctets per second it can deliver to the compute nodes, because that number determines how quickly a simulation can save its state and resume. A machine that computes quickly but writes slowly spends its time waiting.
The ten per cent difference from the decimal unit is significant in that context. A filesystem procured to deliver 10 teraoctets per second and one delivering 10 tebioctets per second differ by a whole teraoctet per second, which in a facility of that size represents a substantial fraction of the hardware budget.
In bits a tebioctet per second is 8 tebibits per second, and in decimal terms about 1.1 teraoctets per second. Expressing the same rate four different ways is routine at this level, because the storage industry, the memory industry, the network industry and the standards bodies each prefer a different one.
For everyday comparison, a tebioctet per second would fill a large consumer hard drive in about twenty seconds. No external interface carries this; the figure describes movement between components inside a single system, where the wires are short and there are very many of them running in parallel.
Graphics processors have brought the rate within reach of a single component. A stack of high-bandwidth memory bonded directly to the processor die delivers well over a tebioctet per second to the chip that uses it, and a card carrying several such stacks passes a few. That bandwidth, rather than raw arithmetic speed, is what limits the training of large models: the arithmetic units sit idle unless the memory can keep them fed. The same reasoning explains why supercomputer designers spend as much effort on the paths between memory and processor as on the processors themselves, and why the rate is quoted in binary units when the memory it describes is addressed in powers of two.
One tebioctet per second equals 1,024 gibioctets per second, 1,099,511,627,776 octets per second, or about 1.100 teraoctets per second.