|
Where once only a handful of connections were made to the PC backplane, now we invariably see a twisted tangle of coax, cords, and cables. A proposed new data-transfer protocol, Serial ATA, promises to simplify the situation while avoiding the data gridlock plaguing high-speed peripherals. What's more, it already has Linux on board.
The cause of that mass of cables and connectors? The growing stable of data-input devices. New hardware peripherals seem to come with their own interface, cables, and
connectors. Internal hard drives use the ATA/66 interface with its 80-pin ribbon cable, while high-speed input devices such as digital video cameras require the IEEE 1394/FireWire interface. Digital cameras and Zip drives want USB. Why this proliferation of pins and protocols?
Before the age of high-capacity data storage and input devices, the only high-speed interface in the computer was between the microprocessor and the internal hard drive. The parallel ATA (also known as IDE) standard was developed for this data interchange. Introduced in 1996, the ATA/33 interface transfers 33 MB of data per second through a 40-pin ribbon cable.
PHYSICAL ATTRIBUTES AND
PERFORMANCE
SPECIFICATIONS FOR I/O CONNECTIONS |
| |
| INTERFACE |
SIGNAL |
CABLING |
SPEED |
APPLICATIONS |
AVAILABILITY |
| ATA/66 |
Parallel |
80-pin ribbon |
66 MB/sec |
PC internal hard drive |
Present |
| ATA/100 |
Parallel |
80-pin ribbon |
100 MB/sec |
PC internal hard drive |
Present |
| USB 1.1 |
Serial |
4-wire cable |
1.2 MB/sec |
Zip, digital cameras |
Present |
| 1394/FireWire |
Serial |
6-wire cable |
40 MB/sec |
Video, hard drives, DVD |
Present |
| USB 2.0 |
Serial |
4-wire cable |
48 MB/sec |
SCSI drives, DSL modems |
Present |
| 1394b |
Serial |
Optical fiber |
80/160/320 MB/sec |
Digital video editing |
Not announced |
| Serial ATA 1x |
Serial |
4-wire cable |
150 MB/sec |
Digital video, hard drives |
Late 2001 |
| Serial ATA 2x |
Serial |
Not announced |
300 MB/sec |
Digital video, hard drives |
Not announced |
| Note: Serial interface speeds are approximate
and depend on processor overhead in any specific implementation. |
With the explosive growth in drive capacity, this data rate quickly proved to be a bottleneck; the doubly-fast ATA/66 was introduced in 1998. ATA/66 included an additional set of 40 wires to reduce crosstalk and noise remaining in the cable after transmission. The ATA/66 was backward-compatible with the ATA/33, the new cables were backward-compatible with the old 40-pin plugs, and motherboards easily upgraded to the 66 MB/sec performance level.
Just two years later, 60- to 80-GB hard drives are again limited by interface bandwidth. Quantum, which recently sold its HDD division to Maxtor, announced the general availability of a patented ATA/100 standard, offering free licensing to competitors in order to gain widespread acceptance.
Linux came on board early with support in the Linux 2.4 kernel. In addition, controller chipsets are already available for the Intel, CMD, and Promise. Nonetheless, as desktop customers continue to require faster performance and greater capacity, ATA/100 is only a stopgap measure.
The race for high-speed interfaces is zooming deeper into the domain of serial protocols. Serial interface bandwidths, generally quoted in megabits per second, can easily be translated into megabytes per second for comparison (assume 8 bits per byte and at least 20% processor overhead).
Serial interface bandwidths have an advantage, as implementing the electrical interface for a serial protocol needs far fewer pins than an equivalent high-speed parallel protocol.
The serial boom began as USB 1.1 gained wide acceptance by offering 1.2 MB/sec transfer rates, which was suitable for consumer electronics such as digital cameras. To handle streaming video, the IEEE introduced FireWire 1394, which offered a thirty-fold improvement over USB 1.1-up to about 40 MB/sec. Moreover, Apple Computer picked up on 1394 as a means to support video and multi-gigabyte hard drives.
Not to be outdone, Intel upped the ante to 80 MB/sec with the announcement of USB 2.0; Intel is also heavily advancing a new Serial ATA interface standard intended to exceed the performance of both USB and 1394. Should the performance of Serial ATA match the claims within the cost structures expected, it may indeed become the industry panacea for both data-transfer gridlock and the interface cacophony caused by the current dizzying number of connectors and cables.
Serial ATA could emerge as cheap and easy enough to accommodate low-speed, inexpensive devices while powerful enough to handle next-generation video and hard drives. As the name implies, Serial ATA will be a serial interface like USB and 1394, but will maintain software and register compatibility with parallel ATA. There will be no immediate
need to rewrite software at the driver or operating-system level.
Under parallel ATA, two devices are daisy-chained together via one ribbon cable using a master/slave communication technique. Under Serial ATA, two devices will use the standard parallel ATA software and device drivers as if they were masters on two separate ports. The drive-interface section of the host adapter uses a new design that converts the normal operations of the software into a serial data/control stream.
As the connection is now direct point-to-point, there's no need for jumper settings or external termination. More importantly, multiple drives no longer share bandwidth and can thereby transfer data in parallel.
Serial ATA uses two differential pairs of wires-one pair to send and one to receive-and three ground connections. These wires fit into a standard connector that will be used across all hard-drive platforms-5.25-inch, 3.5-inch, and 2.5-inch form factors. The elimination of 80-pin flat cables, which are difficult to route and impede airflow inside the chassis, is a distinct benefit to PC manufacturers.
Intel, which has taken a leadership role in promoting the development of Serial ATA despite being heavily invested in USB 2.0, ATA/100, and wireless 1394-has announced the formation of the Serial ATA Working Group (www.serialata.org) to focus solely on the initial specification. Group participants include such industry heavyweights as IBM, Maxtor, Seagate, Western Digital, and Dell. To speed time to market, the group plans to minimize nonessential features while establishing basic functionality and acceptance in the hard-drive industry. Chipsets and hard drives should appear late this year or early next year.
Onboard converters, which will arrange parallel data into the appropriate serial format and vice versa, can handle the backward compatibility. Such converters were used for the first demonstration of the Serial ATA on a Seagate hard drive at the Intel Development Forum last August.
The consortium demonstrated a successful 1.5 Gbit/sec transfer rate. APT also unveiled a DVD-ROM running on the same Serial ATA PCI host adapter, demonstrating the migration of the interface to ATAPI devices. The prototype drive system was built around a natively attached Intel Pentium 4 microprocessor.
The group plans to design Serial ATA to scale well beyond 500 MB/sec. "Serial ATA technology provides the performance headroom to support system and storage device advancements for years to come, as well as backward compatibility to ensure existing software investments are preserved," says Rick Coulson, director of I/O Architecture, Intel Architecture Labs, and the chairman of the Serial ATA Working Group.
"The serial cable and connector will streamline system design while simplifying end-user upgrades of storage devices."
The community will closely follow the group, flush with optimism and technological ambition, to see if it actually delivers on its initial promises. The need for bandwidth is relentless; the group has the industry muscle to make it happen.
| 
