Media Transfer Rates
When we decided to update this particular segment involving media transfer rates, we chose to step away from the norm and not advance anyone’s opinions, theories or supposition, including our own. We chose instead to dig into the subject of media transfer rates, and try and bring into the light of day such issues as “burst” transfer rates as opposed to “typical” transfer rates; what affects transfer rates, and what, from a users standpoint, can be done to improve them; and most important, offer you this information in a format that is easily digested without a myriad of computer jargon. We have reviewed hundreds upon hundreds of pages dedicated to the subject, much of which was filled with opinion, guesswork and sales hype. It wasn’t until we began digging beneath the hype that we found the answers to the questions many people ask, such as “Why can’t I attain the advertised transfer rates?”. What we were looking for was accurate information about all of the issues that affect these transfer rates.
You may be asking why we would go to this extent given all that has been written about faster and faster transfer rates, and what could we possibly add? The answer to these questions is simple. If you peruse the Internet, for example, you will find that most of what has been written is either outdated, overly technical, overly simplified or disjointed. By disjointed, we mean that no one has bothered to tie all of the pieces together and present a single picture.
Beyond this, in all of our years in business we have adhered to principles that have kept us in business, one of those being honesty. Many times a customer, and not necessarily our own, will purchase an ultra-fast hard drive. After installing it, the customer finds very little improvement in system performance, which leads to disappointment and distrust. The customers disappointment was because he or she read the hype as the facts were hidden. Our mission is to present the unvarnished facts!
In order to understand media transfer rates, you must first understand all those things that affect them, and try and determine which of them may affect the performance of your system, and what you can do, if anything, to improve them. Let’s begin with a bit of history surrounding the AT Attachment Interface, or “ATA Interface” for short.
The AT Attachment interface was developed in or about 1981. While some say this effort was led by Western Digital, others suggest it was led by other drive manufacturers. Who led the effort is irrelevant to this discussion!
The fact is, desktop computers are pushing the limits of system performance. Faster microprocessors, multimedia applications, ballooning file sizes, and higher performance hard drives all mean that we are quickly reaching the limits of current data transport pipelines. As we approach these limits, the result will be reduced system performance and slower host-to-storage access. Obviously, there’s a growing need for higher speed I/O connections in order to fully utilize the potential of next-generation devices, especially hard drives. This equates to more disc space, faster performance, more memory, better displays – virtually every component is under relentless pressure to improve. Continual improvement for the disc drive industry means lower costs, improved reliability, higher capacity, and better performance for you.
As computer performance increases, hard drive performance, the central input/output (I/O) device of the computer, becomes increasingly important. Improvement in disc drive performance is a complex area and is measured using several components: seek time, rotational latency, internal transfer rate, cache, and interface speed. Improvements in drive performance, however, do not necessarily equate to improvements in system performance, it must be a combined effort of all components of the system. Adding a high-performance drive requires: An Ultra ATA compatible logic either on the system motherboard, or on an Ultra DMA PCI adapter card, equal to the performance capabilities of the drive; an Ultra DMA compatible BIOS; an operating system that supports Ultra ATA; a DMA-aware device driver for the operating system; and a 40-pin 80-conductor cable. Believe it or not, these are just the basics. There are a few other issues that affect media transfer rates aside from what we have just mentioned, but we’ll get into those after a bit more history.
The hard drive interface, the IDE bus on your systems motherboard, is the path through which data travels between your CPU (processor) and memory and the hard drive. The original ISA-dependent ATA (IDE) interface was limited to about 4 Mbytes/sec in the beginning, but reached as high as 8 Mbytes/sec. Interface protocols, such as programmed input/output (PIO) and direct memory access (DMA) modes, were designed to take advantage of the new local bus architectures that replaced ISA. ATA interface modes have progressed from PIO to DMA and now Ultra DMA, giving burst data transfer rates from 8.3, 11.1, and 13.3 Mbytes/sec up to 16.6, 33.3, 66.6, 100 and now 133 Mbytes/sec.
|Max Transfer Modes||PIO 1||PIO 4
|Max Transfer Rate||4 MBps||16 MBps||16 MBps||33 MBps||66 MBps||100 MBps||133 MBps|
|Max Connections||2||2||2||2 per cable||2 per cable
|2 per cable
|2 per cable
|*Under certain circumstances, and on certain special motherboards, up to 4 IDE cables (8 drives) can be used.|
The original ATA interface was based on transistor-transistor logic (TTL) bus interface technology, which is in turn based on the old industry standard architecture (ISA) bus protocol. This protocol uses a data transfer method called asynchronous. Both data and command signals are sent along a signal pulse called a strobe, but the data and command signals are not interconnected. Only one type of signal (data or command) can be sent at a time, meaning a data request must be completed before a command or other type of signal can be sent along the same strobe.
Beginning with ATA-2 a more efficient method of data transfer called synchronous was used. In synchronous mode, the drive controls the strobe and synchronizes the data and command signals with the rising edge of each pulse. Synchronous data transfers interpret the rising edge of the strobe as a signal separator. Each pulse of the strobe can carry a data or command signal, allowing data and commands to be interspersed along the strobe. In order to obtain improved performance in this type of environment, it would be logical to increase the strobe rate. A faster strobe means faster data transfer, but as the strobe rate increases, the system becomes increasingly sensitive to electro-magnetic interference, (EMI) also known as signal interference, noise or cross-talk, which can (and does) cause data corruption and transfer errors. If you would like to learn more about EMI, follow this link EMI – Electro-Magnetic Interference. We will be discussing EMI issues further into this discussion.
ATA-4 includes Ultra ATA which, in an effort to avoid or at the least reduce EMI, makes the most of existing strobe rates by using both the rising and falling edges of the strobe as signal separators. This allows twice as much data to be transferred at the same strobe rate in the same time period. While ATA-2 and ATA-3 transfer data at burst rates up to 16.6 Mbytes per second, Ultra ATA provides burst transfer rates up to 33.3 Mbytes/sec. The ATA-4 specification adds Ultra DMA mode 2 (33.3 Mbytes/sec) to the previous PIO modes 0-4 and traditional DMA modes 0-2.
ATA-5 includes Ultra ATA/66, which doubles the Ultra ATA burst transfer rate by reducing setup times and increasing the strobe rate. This faster strobe rate increased EMI measurably, which could not be eliminated by the standard 40-pin ribbon cable used by ATA and early Ultra ATA. In order to eliminate (or reduce to acceptable limits) the increase in EMI, a 40-pin, 80-conductor cable was developed and introduced. This cable added 40 additional ground lines, one in between each of the original 40 ground and signal lines. These additional 40 lines help to shield the signal from EMI generated on the cable, as well as reduce EMI affects from external sources. The ATA-5 specification adds Ultra DMA modes 3 (bursts of 44.4 Mbytes/sec) and 4 (bursts of 66.6 Mbytes/sec) to the previous PIO modes 0-4, DMA modes 0-2, and Ultra DMA mode 2. The ATA-6 specification adds Ultra DMA mode 5 (bursts to 100 Mbytes/sec) and the ATA-7 specification adds Ultra DMA mode 6 (bursts to 133 Mbytes/sec).
Changing the ribbon cable alone, however, does not completely solve EMI or the cross-talk issue.
To digress for a moment, initial Ultra ATA implementation resulted in a doubling of the burst transfer rate. By having the hard drive as the source of both the strobe and the data during a read, Ultra ATA/33 eliminated both propagation and data turnaround delays. The elimination of these delays improved the timing margins. Ultra ATA/100 and 133 retain the same strobe frequency but, in the case of Ultra ATA/100 for example, it increases the burst transfer rate three-fold over Ultra ATA/33. Both the perceived, as well as eventual, advantage of Ultra ATA/100 was to increase transfer rates, burst as well as typical. Doing this however, presented a data integrity problem, as the 40-pin interface cable of the earlier Ultra ATA/33 and multi-word DMA interfaces could not handle the shorter cycle times for the strobe rates of the new 66.6 MB/s, 100 MB/s and now 133 MB/s protocols. The 80-conductor cable retains the same connector configuration as the standard 40-pin interface cable, but has ground lines interleaved between all signal lines as noted in Figure 2 above. In other words, the 40 new lines are all ground, acting as shields, and drive manufacturers theorize that no new signals are transferred. This is not necessarily true though, as the new 80-wire cable doesn’t necessarily stop all of the EMI (cross-talk) interference, and the restrictive 18″ ribbon cable length presents other issues to system builders, such as what to do with taller case configurations. As a result of some of these issues, drive and chipset developers enhanced CRC checking even further.
CRC, or Cyclical Redundancy Checking, was first introduced as part of Ultra ATA/33, a “then” new feature of IDE that provided data protection verification. Both Ultra ATA/100 and Ultra ATA/133 still use the same process. CRC is calculated on a per-burst basis by both the host and the hard drive, and is stored in their respective CRC registers. At the end of each burst, the host sends the contents of its CRC register to the hard drive, which then compares it against its own register’s contents. If the hard drive reports errors to the host, the then host retries the command containing the CRC error. But this presents a problem, another bottleneck, as CRC requires additional processing time.
With continued expansion in disk capacity, higher rotational speeds and increased hard drive internal disk rates, the transfer of large files, often written sequentially to the hard drive, are particularly affected by the transfer rate. During sequential reads, the hard drive, because of its fast internal data rate, may fill its buffer faster than the host controller on the motherboard can empty it when using the Ultra ATA/66 or the older multi-word DMA interfaces. This obviously creates a performance bottleneck between the host controller and the hard drive. Hence, improving the interface, from a protocol standpoint, to keep up with internal data rate improvements is exactly what Ultra ATA/100 achieves. As mentioned earlier, fast host data transfer rates help maintain sequential media transfers, but they also accelerate cache hits. In order to understand this a bit better, let’s presume that all of the commands are either a cache hit (data coming from the buffer that has less than 1 ms latency), or a cache miss (data that comes from the media that has a greater than 10 ms latency).
The ATA/100 protocol has the bandwidth capacity to ensure that the hard disk to host data transfer pipeline does not become a limiting factor in next generation system performance. Built-in design compatibility, simplification with tightly controlled trace lengths, slew rates, hysteresis, and other timing parameters ensure optimization of this aspect of media transfer rates. In February of 1999, Intel, in an effort to keep up with and support changing technology, announced support for the ATA/100 protocol with its upcoming 820E (ICH-2) chip set. The combination of the two, the implementation ATA/100 and the development of a supporting chipset with the ICH-2, brought us closer to the full potential of the AT Attachment interface and matching the PCI Bus at 133 Mbytes/sec.
Filling the Pipeline:
Intel has already announced a first look at a 3GHz processor by this fall (or winter 2002) and a release to market in early 2003. Motherboard manufacturers are rolling out some of the most robust designs yet, and memory manufacturers are managing to keep up. Hard drive manufacturers, Maxtor and Western Digital, have announced the releases of huge hard drives, with Maxtor announcing the first ever 320GB hard drive on September 9th 2002. Many technical writers have advanced the idea that all of this development is nothing more than a “clash of the titans”, as in Intel versus AMD for example, vying to be king of the mountain. We think otherwise!
Over the last twelve to eighteen months, there has been an explosive popularity of multimedia applications and the trend towards larger operating systems. Applications such as video on the personal computer, personal video recorders, home video and sound editing and set top boxes have generated a new requirement for massive data files that end users need to access and use at faster rates to ensure smooth performance. Performance problems often appear as data stream interruption, or dropout, which degrades the user experience and is far more obvious than a minimal speed change. With the expansion in media technology, people do not expect to see jerky video playback of movie clips or hear a dropout in their audio playback of music.
If you have ever looked inside your computer to add a component, or to upgrade some parts, you will no doubt recognize the drive and its connection to the motherboard via a flat ribbon cable. This cable is responsible for carrying the data between your hard drive and the motherboard, processor and memory, and was once thought to be one of larger road blocks to high level media transfer rates. The 40-pins 80-wire ribbon cable installed on today’s newest computers is capable of moving data at a maximum speed of 100 MB/s. In reality, there are overhead and timing efficiencies in the ATA interface that slow data transfer rate down significantly. A conservative figure will show a transfer efficiency of 62% (based on a 32K block), so the actual data rate via this cable connection will clock less than 100MB/s.
In mid-2001, the Maxtor Corporation introduced the Ultra ATA/133 interface in an effort to overcome the bottleneck and give users the performance they had been looking for. When this final phase of the ATA interface was developed and introduced by Maxtor, and they had hoped for broad adoption of this faster interface in early 2002, however it doesn’t appear that will happen. Of course, this interface and the newer hard drives will appear in the marketplace, but not nearly on the scale expected by Maxtor and other drive manufacturers. Most of the major component manufacturers, such as Intel, have chosen to promote Serial ATA.
The benefits of the Ultra ATA/133 Interface are 33% faster (burst) data transfer rates of up to 133 Mbytes/s; backward compatibility with all parallel ATA devices, including Ultra ATA/33, ATA/66 and ATA/100; the new interface uses the same 80-conductor, 40-pin cable currently used for ATA/100.
The interface builds upon Maxtor’s patented double-edge clocking technology and cyclical redundancy checking of Ultra ATA/33 and the 80-conductor cable introduced with the Ultra ATA/66 interface. All these interfaces and technologies come from patents held by Maxtor.
*Courtesy Maxtor Corporation
Bringing the ATA interface speed to 133 Mbytes/sec aligns the interface rate with the PCI bus data rate (133 Mbytes/sec). The former ATA/100 interface didn’t quite fill the PCI data pipe, but ATA/133 matches the pipe perfectly. Several have said that ATA data transfer rates faster than 133 Mbytes/sec would be possible. However, there are obstacles to such a speed increase. In order to make this possible, the ATA interface would have to be completely redesigned (read Serial ATA), requiring, among other things, an entirely new cable design, which would virtually eliminate the current backward compatibility provided by the current Ultra ATA interface.
Now that you have a pretty reasonable idea of the basics of media transfer rates, as well as the AT Attachment Interface, let’s take a look at some of the specifics on what affects these transfer rates, what is possible and what isn’t and what you can do to improve performance.
Notice: Windows® 95, Windows® 98, Windows® NT, Windows® 2000, Windows® XP and Microsoft® Office are registered trademarks or trademarks of the Microsoft Corporation.
All other trademarks are the property of their respective owners.