bps FAQ (Data Speed FAQ)

modified 1998-08-11.

This file has the data rate of all kinds of devices. I try to normalize everything to the SI standard unit for data rate, bps (bits/second) (should it be b/s ?) and the pseudo-standard Mbps (which sometimes mean 1 000 000 bps but other times 1 048 576 bps). Other common units are Bytes/second (Bps or B/s; often incorrectly abbreviated bps; 1 B/s = 8 bps exactly) and Hz (a common 16-bit bus has a frequency of 8 MHz = 64 Mbps; a voice stream digitized by the phone company has a bandwidth of exactly 64 000 bps and roughly 4 KHz). As you can see, converting between Hz and bps is not trivial.

This is just something I threw together that I thought you might find useful. Comments? Suggestions for this page are welcome (needed!).

2004-12-04:DAV: todo: move this list to http://en.wikipedia.org/wiki/List_of_device_bandwidths . See also http://en.wikipedia.org/wiki/Comparison_of_latency_and_bandwidth .

Related local files:

http://whatis.com/thespeed.htm almost makes this file redundant.

Copper Serial

EIA-562 (used in notebook computers to hook to RS-232E devices): 64 Kbps (though some implementations go at 116 Kbps or more). [Maxim]

RS-232E: 20 Kbps (though some implementations go at 116 Kbps or more). [Maxim]

ACCESS.bus: max 100 KHz clock.[ACCESS.bus Industry Group, _ACCESS.bus Specifications_ V. 2.2]

«Echelon network (I believe it is called LonWorks) ... Neuron Chips implement a complete 7-layer protocol (based on the OSI architectural model), ... plus memory (ROM, SRAM, and EEPROM) on a single chip, and run internode communication at up to 1.25 Mbits/s (depending on the medium: already implemented and available from multiple suppliers are twisted pair, power line, coax, RF, infrared, and optical fiber). The power line solution, for example, uses 100-450 KHz spread spectrum transmission ... twisted pair (which will give you control message bandwidths up to 1.25 Mbits/s, compared with 10 kbits/s over power lines in North America),» [UseNet]

Serial Port Communication

nominal LocalTalk external clock
classic Macs 57.6Kbps 230.4Kbps 500Kbps
Mac SE 57.6Kbps 230.4Kbps 500Kbps
Mac II 57.6Kbps 230.4Kbps 900Kbps
IBM-AT 115 Kbps N/A N/A

-- from _Apple Mac Family Hardware Reference_ book

«EIA RS-485 ... 4000 ft. maximum cable length ... twisted pair cable ... some implementations [ADM1485] up to 30 Mbits/s» [Analog Devices]

«EIA RS-422 ... over LocalTalk ... 230.4Kbps over a maximum distance of 300 meters. ... max of 32 devices» _Inside AppleTalk, 2nd Edition_ by Sidhu, Andrews, Oppenheimer.

PhoneNET (230.4Kbps)

Topology 22 gauge 24 gauge 26 gauge

Max # nodes

daisy chain n/a n/a 1800 feet 24
Backbone 4500 feet 3000 feet 1800 feet 48

22 gauge = (telephone wall cable - twisted pair)
26 gauge = (modular extension cable)
-- Farallon _PhoneNET Connector User's Guide_.

FireWire = IEEE1394 = «a serial SCSI variant developed jointly by Apple, IBM, and others» ... supports bus speeds of 100, 200, and 400 Mbits/sec.

DeviceNet: multidrop (up to 64 nodes). based on the Controller Area Network (CAN). 500 Kbps at up to 100 m. -- Allen Bradley, «An Open Device Network Standard ... DeviceNet» (1994)

"Silver Streak delivers 10Mbps Ethernet ... [over] ... standard telephone cord ... Simply unplug your PhoneNET connectors and replace them with Silver Streak's." -- Tut Systems, Inc. 1-800-998-4888 ad in MacUser Aug. 1994.

MIDI: 31.25 Kbps [anon]

«Mac to Mac file transfer ... up to 750,000 baud ... with our exclusive LapLink Mac Accelerator.» - Traveling Software (1-800-662-2652)

"Ariel ... supports DSP Port, which originated ... as a ways to connect serial devices to the DSP56001 chip. With a maximum speed of roughly 12 M bits / sec ... Ariel indicates it's popular for I/O on Sparcstations. ... On the AT bus, that firm implemented DSP port on a C30 card"

"... the 2.048 Mbit/s (E1) European rate ... the 1.544 Mbit/s (T1) North American rate" -- 1995 March Dallas Semiconductor ad.

«156Mbps (SONET STS-3c) serial interface»[anon]

"synchronous-optical-network (SONET) OC1/OC3 (51.84/155.52 Mbps) ... fiber-distributed-data-interface (FDDI) 125-MHz clock ... OC12 frequencies (622.08 Mbps) ... RG174 miniature 50 Ohm coax cables are recommended for equipment interconnections at OC3 frequencies and below because they are flexible and lie easily on a benchtop. Keep these cables all the same length ... to avoid clock-vs-data measurement errors ... Time skews are equivalent to 1 ns for 20 cm of cable." -- from article " "Tanktwanger" circuit puts a new spin on clock synthesis" article by Glen Chenier, Fitel Photomatrix, in _EDN_ 1997 Feb 3.

Other Serial

info from article "IrDA-Protocol IR links make 35-fold leap in data-transfer speed" article by Bill Travis in _EDN_ 1996 April 11.

Table 1 -- Speeds of data-transfer methods

Transfer rate (Kbps)
	Transfer Method

9.6	Cellular-phone modem
28.8	High-speed modem
115	serial port
115	IrDA 1.0 (asynchronous)
230	Floppy-disk drive
330	Double-speed CD-ROM
600	standard parallel port
660	Quad-speed CD-ROM
1000	Lan cable (typical)
2000	Tape drive
4000	Enhanced parallel port
4000	IrDA 1.1 (synchronous)

(see David's file machine_vision.html#IrDA for more info on the IrDA infrared wireless communication standard )

(see "DAV's generic serial comm FAQ" for more detailed serial port information)

telephony rates

As far as I know, *all* long-distance telephone calls are digitized at the local POP at 8 bits/sample (non-linear ulaw or Alaw ?), 8 000 samples/s ("voice quality", 40 Hz to 4 000 Hz), then shipped digitally (multiplexed with everyone else's conversations) at 64 000 bits/s. (Both the analog 4 000 Hz signal and the digital 8 000 Hz signal are typically shipped via twisted-pair wires). This is sometimes called "T0" or "D0". "T1" (and "D1", basically the same thing the subtle technical differences) is 24 T0 voice conversations (plus some control overhead) muxed together on a single 1 544 000 bits/s coax cable. "T3", "D3", and SONET OC1 carry 28 T1 signals (672 voice channels) run at rates close to 50 Mbps.

Higher-speed protocols (OC-12, OC-48) typically are carried on fiber optic cables.

ATM is a protocol used for this sort of high-speed, guaranteed-data-rate application (as well as also handling bursty data). Current implementations of ATM support data transfer rates of from 25 to 622 Mbps. http://www.atmforum.com/


SCSI-2 Specification http://abekas.com:8080/SCSI2/

Advanced SCSI Programming Interface (ASPI) software ftp://ftp.adaptec.com/pub/BBS/adaptec/aspi_dev.txt

SCSI Physical Layers (parallel and serial):

Data Rate Cable Length max. devices Interface Name

32 Mbps 6m (SI) 8 (8 bit) Parallel SCSI
20m (DI)
80 Mbps 6m (SI) 8 (8 bit) Fast SCSI
20m (DI)
160 Mbps 6m (SI) 16 (16 bit) Fast and Wide SCSI
20m (DI)

160 Mbps (point to point) 40m (copper) 127 SSA (serial storage architecture)
640 Mbps (loop aggregate) 680m (fiber)

800 Mbps (single loop) 30m (copper) 126 FC-AL (Fibre Channel arbitrated loop)
1600 Mbps (dual loop) 10 Km (fiber)

200 Mbps 4.5 m 64 P1394

SI = Single Ended
DI = Differential
chart from _EDN_, 1995 Oct 12, p. 74, "New Peripheral Interfaces" article by Maury Wright.

Old Mac II SCSI: 1.4 to 1.7 Mbyte/sec typical throughput. Mac Quadra SCSI «approaches limit of 6 Mbytes/sec» «Plain SCSI-2 allows transfer rates at up to 5MHz (i.e. 5MB/sec over an 8 bit bus)» [anon]

"Asynchronous SCSI (aka SCSI I) can be up to 5 Mbyte / second. Synchronous SCSI (aka Fast SCSI II) can be up to 10 Mbyte / second. Fast wide SCSI II can achieve 20 Mbyte / second. Ultra SCSI (not the same as SCSI III) can achieve 20 Mbyte / second in narrow form, and 40 Mbyte / second in wide form." -- From: Tony Rogers <proton@NETSPACE.NET.AU> (To: Multiple recipients of list DDK-L <DDK-L@PEACH.EASE.LSOFT.COM>)(Date: Mon, 13 May 1996)

non-SCSI Storage devices

"Disk data rates ... 250,000 bits per second for 5+1/4 inch floppies" - p. 342 of _The IBM PC from the Inside Out: includes the PC AT_ (1986) by Sargent and Shoemaker.

Parallel (and local buses) (Other than SCSI)

The white paper ``New Bus Architectures: How CardBus Fits with IEEE 1394, USB, PCI and Others'' http://www.pcmcia.org/papers/new_bus.htm by Claude A. Cruz describes several parallel and serial busses.

«PCI gives ... data rates of 1 Gigabit per second and above»[_Solutions OEM, a publication of Intel_, Autumn 1994] PCI bus: 33 MHz * 32 bits (32 bit address and 32 bit data multiplexed on same pins ... an optional 64-bit bus)

GPIB evolved from HP-IB; became standardized as IEEE-488 (1 Mbyte/s); HS-488 high-speed protocol extensions push it to 8 MByte/s

Typical bandwidths of modern video cards in a pentium type system could be...

400-640 Mbps	host -> display memory
1600-4000** Mbps	display memory -> ramdac
2000 Mbps	graphics engine -> display memory (partial or complete 3D rendering pipeline.)

(** this is a peak value for high performance VRAM cards... the actual value on any given graphics system can be estimated by [width * height * vertical_refresh * bytes_per_pixel * 8 bits/byte * 1.3]). -- (From: "John R Pierce" <pierce@hogranch.com>) (Date: Mon, 9 Dec 1996) (Subject: Re: SVGA cards on PCI) (To: pci-sig-request@znyx.com)

Resent-Date: Tue, 10 Dec 1996 12:24:12 -0800
From: "John R Pierce" <pierce at hogranch.com>
Subject: Re: SVGA cards on PCI
Date: Tue, 10 Dec 1996 12:24:12 -0800
X-Mailing-List: <pci-sig at znyx.com> archive/latest/4023
Resent-Sender: pci-sig-request@znyx.com
To: Mailing List Recipients <pci-sig-request at znyx.com>

Eric Rehm <eric@scn.org> asks...
> > the actual value
> > on any given graphics system can be estimated by [width * height *
> > vertical_refresh * bytes_per_pixel * 1.3]).
> Why the factor of 1.3?

Well, thats a ballpark fudge factor.  Thats to allow for the overscan amd
retrace (i.e. blanking time), which is generally around 30% of the total.

The actual video data fetched per second will be [width * height *
vertical_refresh * bytes_per_pixel] but during active video the fetch rate will
generally have to sustain the somewhat higher dotclock rate as very few if any
video boards have sufficient buffering to spread the 'overhead' over a full

Video refresh happens via two distinctly different methods, depending on
whether one is dealing with a DRAM or a VRAM board.

On a VRAM board, the [often 64 bit wide] outputs of the VRAM's are connected
directly up to the RAMDAC, and pixel data gets clocked out of the VRAM into the
RAMDAC in real time.  Since most all VRAM's can clock at speeds up to 50MHz or
higher, 50MHz * 8 bytes provides 400MBytes/sec or more of video refresh, enough
to handle most any display mode.  Since the VRAM's do this with a second
'serial' port, this display refresh activity has no impact on drawing

On a DRAM board its quite a bit more complex.  Taking the example of a 64 bit
modern EDO style DRAM card, EDO bursts may be able to fetch somewheres around
250-300MBytes/sec.  This bandwidth however must be shared between display
refresh, host access, and graphics controller memory operations (bitblt, etc).
The display refresh controller will typically have a FIFO of a few dozen 8byte
'words'...   Basically, the display controller will fill this fifo at full
burst speed, then release the memory subsystem to other uses until it needs to
reload the fifo...  The higher the MByte/sec display refresh requirements are,
the less time left over for host access and graphics controller operations.

Since I've explained all this... here's a little kicker.  At lower
resolution/color depth combos,  EDO DRAM is FASTER than VRAM.  This is because
EDO DRAM tends to have faster memory accesses... If the display refresh
requirements are modest, this still leaves MORE bandwidth left over for host
and graphics accesses.  Only when you get into fairly high resolution / color
depth / refresh rates does the inherently slower VRAM catch up due to its
dualported nature.  My general rule of thumb... if you have a 14" or 15"
monitor, stick with DRAM. If you have a 17" and are going to run at 1024x768
16million colors at refresh rates like 75Hz, you'll probably want VRAM.  If you
have a 21" you definately want VRAM (although in fact, newer EDO DRAM cards do
quite nicely at 1280x1024x64k colors).

[footnote... I know this has not that much to do with PCI etc, but once I was
on a roll...]


ISA bus: 8.33 MHz * 16 bits ("286" "AT-bus")

SCSI jaz drives are actually fairly quick.  IOMEGA claims Sustained transfer
rate: 6.6MB/sec. maximum; 5.4MB/sec. average; 3.4MB/sec. minimum...
Obviously OS overhead is going to enter into the picture...

>1 MByte/s is a typical maximum rate on a ISA bus, but you say you're using PCI.

Actually, the ISA bus itself can sustain xfers around 6 to 8MByte/sec [2
clocks at 8MHz per 16bits].

OTOH, the MFM disks used on the ORIGINAL ISA systems were only capable of
maybe 240kbyte/sec sustained.  Somehow, I don't think so.

>I assume you're writing out very large chunks (multiple sectors) with each
>fwrite(), not merely one character. (Obsolete OSes were extremely slow with
>one-char-at-a-time writes; modern OSes buffer so well that the difference
>is negligible).

The overhead of making library calls to move one byte at a time to file
system buffers is not inconsiderable, it easily could be 1 uS or more even
on a 166MHz processor.

---- From: John R Pierce <pierce at hogranch.com>
(To: cary at agora.rdrop.com (David Cary))
(Date: Tue, 07 Jan 1997)
(Subject: Re: Fast Disk Writes)

Bus Specifications http://www.microsoft.com/hwdev/ver4/respec/busspecs.htm

Seagate has some technical papers on Fibre Channel, Ultra2 SCSI (LVD) Low Voltage Differential, and Self-Monitoring, Analysis and Reporting Technology (S.M.A.R.T.). http://www.seagate.com/support/papers.shtml

from _Computer Design_ 1997 Nov p. 102 "Serial bus technologies stake out territories" article by Jeff Child "next-generation peripheral buses ... Source: Molex" figure
<1ADB and ACCESS.bus
4slow USB
12fast USB (Universal Serial Bus)
40SCSI 2
80SCSI 3
2001394 AV
400IEEE 1394 (FireWire)
1 000FC-EL
1 2001394.2
8 000proposed 1394 upgrade, backwardly compatible

From the text and other sources, these numbers are supposed to be in Mbits/s; the figure confusingly uses the abbreviation "MB/s", which is usually interpreted as MBytes/s.

More unsorted misc stuff.

Echelon LonWorks(tm) 78Kbps differential Manchester communication with Neuron(r) chips.

Get info on digital / computer buses
(merge this with "serial" and "parallel" files)

Consumer Electronics Bus (CEBus)
developed by the Electronic Industries Association (EIA)
for home networking.

IS-60 standard uses standard power lines as a communications medium
... also RF transceivers, twisted-pair wires, or fiber-optic cables.
The EIA IS-60 standard specifies ... spread-spectrum technology ... carrier-sense, multiple-access (CSMA) network. Data is transmitted over existing AC wiring at a rate of 10,000 "1" bits per second.
... 100 to 400 KHz spread-spectrum "chirps". Data is encoded by inverting or not inverting the phase of the individual chirps.
... the amplitude of the chirps must be between 2.5Vpp and 7 Vpp when measured into loads ranging from 10 Ohm to 2 KOhm.
... Because the upper end of the CEBus standard's spectrum approaches the lower end of the AM broadcast radio spectrum, you will need a transmit filter to ensure that  CEBus transmissions do not interfere with AM radio reception. The AM radio band starts at 540 KHz ... comply with FCC levels ... the transmit filter's output can't exceed 60 dBuV at 530 KHz into a 50 Ohm Line Impedance Stabilization Network. ...
-- from article "Build a Tester for CEBus Devices" by Robert B. Shepard, _Test and Measurement World_ 1995 June.
article refers to EIA/IS-60, "EIA Home Automation Systems (CEBus)", Electronic Industries Association, Washington, DC, 202.457.4900, Oct. 1992.

parallel port
want info on the IEEE-P1284 spec, “Standard Signaling Method for a Bidirectional Parallel Peripheral Interface for Personal Computers” -- is ECP mode mentioned ? “IEEE Std 1284-19XX”.
see "DAV's parallel FAQ" file.

NuBus 90 documentation
Date Written:  9/17/91
Last reviewed:  2/9/93

Where can we get NuBus 90 information?

NuBus 90 is specified in the following two documents published by the
Institute of Electrical and Electronics Engineers (IEEE):

• IEEE Standard for a Simple 32-bit Backplane Bus: NuBus
• IEEE Std. P1196-R1990

You can obtain these documents by contacting IEEE at the following address:

    IEEE Service Center
    445 Hoes Lane
    Piscataway, NJ  08854

standard master-slave polling protocol.

IEEE 488
“has been around for a long time ... widely used and therefore inexpensive”
“The maximum data rate over the IEEE 488 bus is 1 Mbyte/s, and actual performance is typically 200 kbytes/s or less.”
-- Dan Romanchik , “How to connect Computers to VXIbus Systems” article in _Test and Measurement World_ June 1995.

PC Card

Fibre Channel SCSI
100 MBytes/second and future increases planned
fiber optic or copper

Dallas Semiconductor's clever 1-wire communication protocol http://www.pointsix.com/1wire.htm communicates at 14.4 Kbps.

Bluetooth "The gross data rate is 1Mb/s." The nominal link range is 0.1 m to 10 m, but can be extended to more than 100 m by increasing the transmit power." see machine_vision.html#spread_spectrum for more info on machine_vision.html#bluetooth .


The SCRAMNet http://www.systran.com/scramnet.htm (Shared Common RAM Network) "is a multi-vendor, replicated shared-memory, 150-megabit/second fiber optic communication system designed for use in real-time computer systems." "data latency per node is only 250-800 nanoseconds. Up to 256 nodes (each separated by up to 3500 meters) can be contained on a single fiber optic ring... . ... . The SCRAMNet protocol provides calculated maximum transit times, avoids network collisions, and avoids software delays."

cable modems: network -> computer: typically 4 Mbps; theoretical limit of ~36 Mbps. computer -> network: typically 0.5 Mbps; theoretical limit of 10 Mbps. More cable modem info: http://dmoz.org/Computers/Data_Communications/Cable_Modem/ has user-level set-up installation tips, and links to manufacturers and distributors. http://dmoz.org/Science/Reference/Standards/Telecommunications/ goes into the low-level protocols used by people who design cable modems.

Media Fusion http://www.mediafusionllc.net/ claims to be able to transfer up to 2.5 GB/sec. over ordinary power lines. "powerline area network" (PAN). While I know that many people are already transferring relatively low-speed data over power lines (X-10, etc), 2.5 GB/sec seems too fast to be believable. It would be cool if it really works.

Communications ICs http://www.wnielectronics.com/products/9106.htm ???

Broadband Systems & Design magazine http://www.broadbandmag.com/ ???


1997 Feb 15: html-ized and put on web
Started: 1996 Aug 25 (split off from my main "Bignums" file)
Original Author: David Cary.
Current maintainer: David Cary.

Return to list of DAV's documents

Send comments, suggestions, bug reports to David Cary d.cary@ieee.org. http://rdrop.com/~cary/html/bps_faq.html