| Bits per second (bit/s) | Tebioctets per second (Tio/s) |
|---|---|
| 1 Bit per second | 1.13686837722 × 10-13 Tio/s |
| 2 Bits per second | 2.27373675443 × 10-13 Tio/s |
| 3 Bits per second | 3.41060513165 × 10-13 Tio/s |
| 4 Bits per second | 4.54747350886 × 10-13 Tio/s |
| 5 Bits per second | 5.68434188608 × 10-13 Tio/s |
| 10 Bits per second | 1.13686837722 × 10-12 Tio/s |
| 20 Bits per second | 2.27373675443 × 10-12 Tio/s |
| 25 Bits per second | 2.84217094304 × 10-12 Tio/s |
| 50 Bits per second | 5.68434188608 × 10-12 Tio/s |
| 100 Bits per second | 1.13686837722 × 10-11 Tio/s |
| Reference | Bits per second (bit/s) | Tebioctets per second (Tio/s) |
|---|---|---|
| A dial-up modem | 56000 bit/s | 0.00000000636646 Tio/s |
| Typical home broadband | 100000000 bit/s | 0.0000113687 Tio/s |
| Gigabit Ethernet | 1 × 109 bit/s | 0.000113687 Tio/s |
| Streaming a 4K film | 25000000 bit/s | 0.00000284217 Tio/s |
The bit per second is the fundamental unit of data transfer rate. Its symbol is bit/s, often written bps. It counts how many binary decisions a channel carries in one second, and every other unit of transmission speed is a multiple of it.
Because it is a rate, it has the form of a quantity divided by time, exactly like metres per second or litres per second. That makes the arithmetic straightforward: a link running at a given number of bits per second, multiplied by a duration in seconds, gives the total number of bits transferred, and dividing by eight converts that to octets.
The unit must be distinguished from the baud, which counts symbols per second rather than bits. Early modems transmitted one bit per symbol, so the two numbers were the same and the words were used interchangeably. Modern schemes encode several bits in each symbol — by varying phase and amplitude together — so a channel running at 3,000 baud may carry 33,600 bits per second. Only the bit rate describes how much information moves.
Claude Shannon established the theoretical ceiling in 1948. The capacity of a channel in bits per second depends on its bandwidth and on the ratio of signal to noise, and no coding scheme can exceed it. Every advance in modem and radio design since has been an attempt to approach that limit more closely, and modern systems come within a fraction of a decibel of it.
In practice the raw bit rate of a link is never the rate at which useful data arrives. Protocol headers, error-correcting codes, acknowledgements and retransmissions all consume capacity, and the usable fraction is typically 90 to 95 per cent on a wired link and considerably less on a shared wireless one.
Single bits per second are rarely quoted, because almost every channel is faster. The exceptions are deep-space communication, where a probe billions of kilometres away may return data at a few tens of bits per second, and certain low-power sensor networks that transmit a handful of bits at long intervals to preserve battery life.
One bit per second equals 0.125 octets per second, 0.001 kilobits per second, or about 0.0009766 kibibits 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.