There appears to be some confusion in the market following the recent announcement of USB 2.0. To the casual observer, the role of 1394 appears to have changed significantly; however, to those of us who are close to the action, little has changed. USB has always been, and continues to be, an interface for traditional peripheral devices which operate in a PC-centric environment. 1394 is a multimedia bus enabling high-speed peripherals, convergence between PCs, peripherals, and consumer A/V products -- and enabling new applications for traditional products. Even at faster speeds the basic roles of USB and 1394 remain the same.

The implication of some recent announcements has been that 1394 is not appropriate for PC peripherals. Nothing could be further from the truth. 1394 HDDs and desktop cameras are already in the market with scanners and printers to follow in the second half of 1999. 1394 peripherals entering the market this year have chosen 400 Mbit/s as the entry point, well above the speeds of USB 2.0. The additional cost for 400 over 200 is insignificant and the benefits are enormous. By using silicon which operates at speeds well above that required for a particular PC peripheral, the 1394 bus is not slowed by 100-Mbit/s or 200-Mbit/s transactions. This leaves more bandwidth for all devices connected together via 1394. With 1394, there is plenty of bandwidth for several high-speed devices to work at the same time. It is conceivable that mid-speed peripherals may choose USB initially, but transition to 1394 in search of adequate bandwidth.

One of the benefits of USB 2.0 should be increased bandwidth. USB suffers from over-marketing which has reduced its benefit to the end user. Expectations are far above what USB can deliver today. With USB being promoted for use in speakers, desktop cameras, mass storage, and printers, the end user might reasonably expect to be able to use two or three of these devices at the same time. USB proponents generally agree that USB peripherals are primarily intended to be used one at a time. There are even reports of bandwidth-related problems when playing audio and trying to use a low bandwidth device such as a mouse at the same time. With a faster version of USB, multiple devices could be used at the same time and the benefit of USB would be greatly increased. However, the tendency of USB marketers to over-market has them encouraging higher-bandwidth peripherals to design USB 2.0 into future products. The USB bandwidth problem is just being repeated at a faster speed.

For PC-peripheral companies who are looking to expand their SAM, 1394 is a godsend. With the peer-to-peer capabilities of 1394, peripherals can connect directly to each other such as a scanner and a printer for "photocopies" or a digital camera and a photo-quality printer. 1394 also enables new applications for traditional devices such as mass storage connected directly to an STB, serving the function of a digital VCR but with random access. Now the end-user can record a program and, through time shifting, start watching the beginning of the program while it is still being recorded. This opens the market for peripherals for people who do not own a PC. It also improves the market for selling second peripheral devices into homes which already have printers, scanners, etc. Each of these example products is now in development and will soon be in the market.

The market for 1394 is very healthy today. 1394 is the undisputed winner in the consumer A/V realm. It has been shipping in digital camcorders since the summer of 1995 and is currently in camcorders from Canon, Sony, Panasonic, Sharp, and JVC. Approximately 7 million have been sold to date, with a current run rate of 1 million new units per quarter. 1394 DTVs, D-VCRs, D-STBs, and AV-HDDs are scheduled to enter the market in plenty of time for Christmas 99.

But new A/V products will not be the only new 1394 products attracting attention on store shelves around Christmas this year. Printers and scanners are also scheduled to hit the market in time for the rush. 1394 mass storage devices for Apple G3s are already shipping with devices for PCs scheduled to appear in mid year 99. 1394-enabled PCs are already shipping from NEC, Compaq, Sony, and Apple, with four more coming from the top 10 PC makers this year. Estimated 1999 sales are 8 million for PCs sporting 1394 ports. No lesser an authority than Intel has declared that 1394 will be common in PCs in the near future to enable convergence with consumer A/V devices.

So what about USB 2.0? Preliminary research indicates that USB 1.1 can be sped up, but not easily. There are a number of design hurdles which need to be overcome.

First of all are the cables and connectors. Can they really work at higher data rates? The USB enthusiasts frequently point to super-cheap USB cables and connectors as an indication of how cheap it is to implement USB. Unfortunately, these super-cheap cables and connectors don't match the USB standard, wear out prematurely, and have severe radiation problems. These will obviously not work at faster speeds. They barely work at the 12 Mbit/s standard. Even the suppliers of brand-name USB cables and connectors freely express doubts that their current products can work at higher speeds. There is a high probability that USB 2.0 will have to use different cables and connectors and the USB 1.1 products will become antiquated.

How much will USB 2.0 cost? No one knows, but we do know that the most significant cost adder to 1394 (over USB) is a direct result of the back-end infrastructure, FIFOs, buffers etc. required to run at faster speeds. USB will not be able to increase speed without approaching the cost of 1394. Keep in mind that a 1394 chipset cost $60 five years ago. In 2000, the cost will be below $6 and will be going lower.

How fast will USB 2.0 be? The announced speeds are 120 Mbit/s and 240 Mbit/s, but in reality USB throughput is not nearly as fast as advertised. When shipping data directly from a peripheral to the host, throughput is OK. However, when shipping data from a peripheral to another peripheral, real bandwidth drops in half. All data must be moved twice. It must be moved from the peripheral to the host and then from the host to the target peripheral. USB 2.0 does not really compete with 1394. Even at rates of 240 Mbit/s, USB 2.0 is only as fast as 1394 at 100 Mbit/s. This data rate is fast becoming too slow for everything but consumer A/V products. Of course, consumer A/V products will never incorporate USB. USB requires a CPU to perform the bus master functions while 1394 is peer-to-peer. A D-VCR must be able to talk directly to the D-TV without going through a PC first.

And finally, how will USB 1.1 and USB 2.0 work together? According to the current information, a complete USB 1.1 and USB 2.0 block is required in EVERY hub and host controller. Now each device needs the intelligence to recognize what it is connected to and it must have routing logic to only connect USB 1.1 to other USB 1.1 devices and only connect USB 2.0 to other USB 2.0 devices, forming two completely separate USB buses. If a user's USB 1.1 bus does not have enough bandwidth for its current devices, buying a USB 2.0 host controller won't help. He must buy new USB 2.0 everything (hub, controller, and device) for the data path from source to sink. This is required since a USB1.1 device will bottleneck the throughput where it appears.

The USB 2.0 hubs are now more complicated. Not only do they require an entire USB 1.1 HOST controller and a new USB 2.0 hub controller, but also a high-speed signal repeater, routing logic, dual-function ports, etc. The biggest problem here is that the USB 1.1 host controller must also be a "rate matcher" between USB 1.1 and USB 2.0. This requires the intelligence and FIFO to buffer packets with a 240 (two hundred forty) times difference between data rate (USB 1 Mbit/s to USB2.0 240 Mbit/s.) With USB hubs costing from $80 to $120 today, how costly will USB 2.0 hubs be? Only the two USB ports on the PC are "free." Once the keyboard and printer are connected, the end user must buy a hub before connecting more USB peripherals.

It will be interesting to see how the best and brightest in the USB community will overcome these barriers. They are a talented bunch of engineers and I have little doubt that they will find workable solutions. How much will it cost? More than they are saying. How soon will it be ready? Not as quickly as they think. What is the benefit to the market? I'm not really sure but with 1394 already in the PC, I can't help but ask "Why add a second high-speed interface? Wouldn't it be a better use of engineering talent to simply use 1394?" When USB 2.0 was announced to an IDF crowd of 2000, only one person applauded. I can only surmise that the rest of the crowd was mulling over the significant task ahead of them.