Ottawa PC Users' Group (OPCUG)
 
   Home
   Reviews

 

   Copyright and Usage

   Privacy Policy

   Contact Us

 

A tale of three serials

by Dunc Petrie

Yesterday

Historically, serial connections have often been regarded as the poor cousins when speed was the criterion. Their tasks have included printers (almost universally replaced eons ago by the faster parallel port), a mouse port and a modem. Speed was - until recently - the fiat of parallel bus connections: for example, printer parallel ports, IDE/EIDE hard drive connectors and SCSI

Today

Serial connections are reborn: they can boast speeds that shame their parallel bus competition and they are poised to make inroads into other traditional parallel connection territory

Most desktop users today have Ultra DMA/66 as the hard drive interface "flavour of the year." This interface is capable of data transfers up to 66 MB (bytes, not bits) per second (burst mode) although the sustained rate is proportionately lower. SCSI is an option; however, for most users it suffers a significant cost disadvantage. Not only are the hard drives more expensive but also users must purchase the host adapter. Most home and small business users opt to use the built-in IDE/EIDE connectors

As processor power and speed accelerate, data must arrive faster to prevent a bottleneck. To compensate, several design approaches, singly or in combination, are available: for example, faster platter rotation speeds, greater areal density, RAID drive arrays, larger on-drive caches and faster algorithms

Indeed, a faster protocol - Ultra DMA/100 - has appeared and chipsets to support it are becoming available. Already, the next increment, Ultra DMA/133 awaits official blessing

Tomorrow

High speed serial connections were resurrected with the arrival of Universal Serial Bus (USB). Motherboard support is integrated with connectors on ATX form  boards; a wide range of peripherals (for example, keyboards, mouse, printers, scanners and digital cameras) utilize the convenience of hot plugging. USB is only the pioneer; indeed, it is scheduled for an imminent speed boost. Also waiting in the wings is IEEE 1394 (aka Firewire - trademark of Apple Computers, or iLink - trademark of Sony, or high speed serial bus or high speed serial interface) and Serial ATA. IEEE 1394 was originally relegated to digital video interconnections; however, it has gained prominence as other peripherals (chiefly hard drives) arrive in retail channels

Serial ATA

The incremental progression of "Ultra" may soon be replaced by an entirely different protocol: Serial ATA. Gone would be ribbon cables: the hallmark of the present parallel signal interface. The major problem with the present scheme is signal noise and crosstalk between adjacent wires. To minimize this interference Ultra DMA/66 (and beyond) specifies an 80 conductor ribbon instead of the more common 40 wire; for backward compatibility the familiar 40 pin connector remains. These connectors are significantly more expensive locally. Switching to a serial interface allows fewer wires and lower voltages; both are conducive to less signal noise and crosstalk (with less heat and better cable organization as freebies)

Features/Benefits of Serial ATA

  • Increased transfer rates - initially up to 150 MB (again, bytes not bits) per second (MBps) with projected incremental increases to 600 MBps)
  • a round, 8 pin (PS/2 mouse connector style), seven conductor, cable instead of a wide ribbon - promoting better airflow and requiring less motherboard space
  • no need to daisy-chain (each drive will have an individual cable)
  • point-to-point protocol - the end of master/slave jumpers (Each drive would have a single-purpose connector that would connect to a motherboard hub. For those versed in LAN topology this would be a star configuration.)
  • parallel operation of devices (currently, only one drive on a bus can be working)
  • first party DMA (allows device access to host memory - uncertain of details but supposedly improved from the current algorithm)
  • lower logic signal voltages and differential signalling (500 millivolts proposed versus the current 5 volts) - less noise, crosstalk
  • reduced motor voltage (3.3 versus 5.0) for less heat production: a boon for portables
  • transparent support for CDs, CD-R, CD-RW, DVDs, tape devices, high capacity removable media devices, zip drives
  • intended only for internal use - no external connectors anticipated for peripherals (marketing strategy to keep costs down; avoids bogging down hard drive performance with slower peripherals)
  • one meter maximum cable length (ideal for Ultra is less than 0.5 meter) permitting greater flexibility of internal case layout
  • forward and backward compatibility with existing hardware provided by adapters (both new drives to the old interfaces and old drives to Serial ATA)
  • OS and BIOS transparent (Microsoft Windows 9x, Me, and 2000; Linux and presumably Mac)
  • compatible with all current software (register level transparent)
  • device driver changes not necessary
  • hot pluggable at manufacturer's option (less expensive hard drive arrays?)
  • provisional timetable: 1.0 release fall 2001, first commercially available products late 2001, broad adoption in 2002/li>
  • identical connectors for desktop (3.5 inch form factor) and mobile (2.5 inch form factor) drives
  • cost competitive with present ATA/ATAPI

USB

The specification for the current Version (1.1) supports transfer rates of 1.2-1.5 MBps; the proposed 2.0 specification would raise it to 60 MBps (in 2002) while providing backwards and forwards compatibility. That is, a version 2 device would work when connected to a version 1.1 interface (only at the slower rate) and a version 1.1 device would work when connected to a verson 2.0 port (but would not speed up). For external peripherals like scanners this offers the expectation of SCSI-like performance without the additional cost of the host adapter

USB was originally intended for low-impact, narrow bandwidth peripherals (like keyboards and mice). Indeed, the USB chipset and cabling designs are not ideal for wide bandwidth devices like hard drives. Intel, the parent of USB, in conjunction with Microsoft originally indicated that IEEE 1394 was the high-end counterpart and co-existence was the operating strategy. Are the gloves off? This new variant offers comparable performance to IEEE 1394 (SCSI beware) and enjoys de facto motherboard integration that reduces costs (IEEE 1394 is integrated on few systems; most implementations utilize a PCI expansion card)

USB Specifications

  • intended only for external devices
  • integrated on motherboard
  • cable up to 6 meters per link
  • not bootable
  • hot pluggable
  • foolproof, one-way cables (not IEEE 1394 interchangeable)

IEEE 1394

This serial protocol was faster than the original implementation of USB and entered the arena in digital video applications. Given this niche application, higher implementation costs and "no other game in town" it has languished in the PC marketplace. Some manufacturers are offering "Firewire" hard drives that use a conventional hard drive in a suitable external package with this interface but with  a considerable price premium. By increasing USB's data transfer rates into IEEE 1394 territory, the latter would be hard-pressed to break out of its mold given the higher cost of implementation. Still, what it does, it does well and for technical reasons USB is not ideally engineered to provide large amounts of data at high speed (mass storage data transfers)

IEEE 1394 Specifications

  • aimed at high bandwidth peripherals (digital video) and mass storage
  • higher implementation costs (also possibly royalties)
  • originally much faster than USB 1.1; roughly comparable to USB 2.0 (IEEE 1394 is scheduled to increase in 2005  to 200 MBps)
  • intended for external devices only
  • motherboard integration seems unlikely (Intel holds the reins here and a more complicated protocol requires supporting chipsets that are at a cost disadvantage when compared to USB)
  • not bootable
  • hot pluggable
  • foolproof, one-way cables (not USB interchangeable)
  • up to 4.5 meters per link

Conclusions

Given the momentum behind Serial ATA (most major manufacturers are committed) it would appear to be a "no brainer." Faster processors require greater data bandwidth and the current protocol has reached the end of its lifespan. The new approach will offer greater data throughput, headroom for future speed increases, internal design improvements (better airflow, less heat generation) and ultimately lower costs. During transition, inexpensive (hopefully) adapters will allow gradual phase-in with minimal disruptions

USB appears to have the upper hand. It is positioned to support a diverse range of peripherals and motherboard integration offers the benefits of low cost and end-user simplicity. Otherwise, both buses are surprisingly alike. Casual end-users, I suspect, will be quite satisfied with the already present USB. For them, USB fulfills their peripheral connection requirements. IEEE 1394 has made no attempt (to date) to move into the USB arena of connecting a mouse or keyboard

The exotics of digital video aside, consider connecting a scanner. Given the choice, would you purchase: a USB version that you can literally plug in and run the software; a SCSI version that requires obtaining and configuring a host adapter, setting the device ID  and arranging termination in addition to the software; or, an IEEE 1394 version that requires purchasing a card, installing software drivers and then the scanner software. Yes, IEEE 1394 is arguably easier than SCSI but why would you purchase, at additional expense, any add-in board (regardless of protocol), with the complications of setup (tongue in cheek: Plug and Pray) and drivers when your motherboard has all you need?

Surprisingly, IEEE 1394 and USB (Version 2.0) are more alike than different with specifications that are very similar. Marketing considerations originally positioned them to address, respectively, the high end and the low end of the peripheral arena. If IEEE 1394 were integrated on the motherboard, instead of an expansion card, could/would end- users notice much difference? Would they care? In both cases, I doubt it.


Originally published: April, 2001


top of page

 

 

Archived Reviews

A-J

K-Q

R-Z

 

The opinions expressed in these reviews
do not necessarily represent the views of the
Ottawa PC Users' Group or its members.