Let's face it, we computer users are greedy. We always want more power, more speed, and more time. Luckily the more advanced people at Apple (not the geniuses who gave us the crippled Classic) think along the same lines and have come up with a new technology called QuickRing, which promises to significantly enhance the Mac's utility in some data transfer-intensive tasks.
Each successive generation of Macs runs faster than the last, but the Macintosh still some notable bottlenecks, including SCSI, the memory subsystem, and the processor itself. One bottleneck that you may not often notice is the NuBus. Currently, NuBus is limited to transfer rates of about 20 MB per second on the Quadras and 10 MB per second on the older machines. A friend calculated that only a 33 MHz 68040 will begin to outpace the this bottleneck.
Apple apparently feels that 20 MB per second is not up to snuff, since snuff-induced sneezes travel pretty quickly, some 200 miles per hour according to a book I read many years ago. Enter QuickRing. QuickRing is a high-speed architecture for data transfer between NuBus cards in the Mac so that they can move data faster than ten times the maximum speed of NuBus, or 200 MB per second. That's something to sneeze at. Keep in mind that although the cards will be NuBus cards, there will have to be a faster connection to the CPU than what the NuBus offers. I anticipate that Apple will either use some sort of direct connection to the CPU (unlikely) or the Processor Direct Slot (why do you think they call it that!).
What's it good for? -- What would you want to do that would require that sort of speed? If you only use one NuBus card, you probably don't need the speed. But, if you use several cards simultaneously, the speed could come in handy. Most of us don't use several NuBus cards at the same time, but that day may come sooner than we think. Here's some examples of what QuickRing will be good for.
Apple's pushing a lot of high-technology announcements out the door these days, which is a heck of a lot easier than actually pushing the high-tech out the door. Whenever you talk about voice recognition or accurate handwriting recognition, you have to think about extra hardware. It's possible to do it in software, but the more processing power you can throw at voice recognition, the better it can do and the more it can do with what it hears. Voice recognition is perfect job for a fast card with a digital signal processor (DSP) chip, though Apple may manage to get it working completely in software on today's high-end Macs.
QuickTime is nice idea and seriously snazzy, but let's face it, watching a five second clip of "Star Wars" on a postage stamp isn't exactly my idea of entertainment. To produce serious QuickTime movies you need hardware, and the more you have, the better. Today's hardware (like a VideoSpigot) lets you do real-time video in a 160 x 120 pixel window - increasing the window to size of an index card will limit the frames per second. However, if you can throw some faster hardware at the problem with one card to bring in video, another to compress it, and a third to display accelerated graphics, you could probably capture full-screen real-time video, assuming you had that sort of hard disk space. Alternately, a 13" color TV and cheap VCR will do the same thing. :-)
Another proposed use is in high-speed networking, although the current NuBus is more than sufficient for even a theoretically-fast Ethernet network, which runs at 10 megabits per second (Mbps), and even a super-fast 100 Mbps network could in theory work with NuBus, although there would certainly be speed-eating conflicts if anything else was using the NuBus at the same time. The main things I can imagine that would require such speeds would have to do with complex graphics or video. Videoconferencing might come into its own with QuickRing, especially since there's very little processing that would have to go on, so the processor and memory systems wouldn't be limiting factors.
Of course, once you've got a tremendously fast network, you will probably want to do some distributed processing of all those 3-D rendered QuickTime movies you'll be making. Sending some of the processing work off to other Macs on your network will work fine, but there's no need for new technology for that. You could also put multiple processors in one Mac and have them communicate very quickly to off-load processing from your primary CPU. Radius is already working on this sort of thing with its Rocket accelerators and RocketShare, and perhaps there are some multiple processing applications that could benefit from this, although the Mac would need much faster memory to really take advantage of it.
Why it's cool. -- I've perhaps sounded slightly dubious about some of these uses, but Paul Sweazey of Apple's Advanced Technology Group pointed out why QuickRing truly is a big step. Although the NuBus doesn't seem like a major bottleneck, you have to keep in mind that it has to carry numerous different tasks. So in a common setup, you might have an EtherTalk card, an accelerated video card, and maybe a Radius Rocket, all in the same Mac. None of those individual tasks will present any threat to the bandwidth of the NuBus, but together, they might come close. Add in a voice recognition board and something to capture real-time video, and you've seriously overloaded NuBus, which would have been more than enough for one of those tasks.
Even more important is the fact that NuBus is fairly inefficient, so the three tasks will bump into each other all the time. No one anticipates needing 200 MB per second of throughput with one task on QuickRing, at least not right away, but in the meantime, it will provide faster real-world speed to multiple slower tasks. Think of NuBus as a two lane highway that bogs down when there are 50 cars all trying to enter and exit at the same time. QuickRing, in contrast, would be the equivalent of a 20 lane highway for those same 50 cars. Plenty of room.
Interestingly, Apple decided to get help with QuickRing, and it was developed jointly by Apple's Advanced Technology Group, National Semiconductor, Molex, and Beta Phase. National Semiconductor designed the controller chip, and Molex and Beta Phase cooperated on designing and manufacturing the interconnect system to go between the cards. Apparently, the hard part was to create the chip and the interconnect system using conventional methods so that the finished products could be produced in high volume and at a reasonably low cost. The problem was that these controllers and interconnect systems have been done in the past, but only to work with Cray supercomputers and the like, and you just don't worry too much about producing anything for a Cray in volume cheaply. There just aren't enough Crays around and they're so expensive that no one worries about the price of a controller chip here or an interconnect system there.
It will be a while before QuickRing products appear, but I believe they will be compatible with current high-end machines, so that won't be a limiting factor. In fact, it seems that without faster processors and faster memory, we won't be able to do much at all with QuickRing. However, processors, memory systems, and disk storage systems have all significantly increased in either speed or capacity in the last few years, and the main area left dormant has been the bus systems. QuickRing may provide more than the rest of the Mac can handle, but so what? Aldus could never have created PageMaker 4.2 if all we had was double 800K floppies. QuickRing pushes the performance envelope, and the rest of the systems will play catch-up with what it makes possible for few years, just as software developers were suddenly able to create huge applications like PageMaker once most everyone had a hard disk.
Look for components to start being available to developers in early 1993, which means products might show up sometime in 1993/94, or a bit before the future fantasy time when Taligent is supposed to deliver Pink. Developers interested in QuickRing can send email to Apple at the AppleLink address QUICKRING or send snail mail to the following address:
Apple Computer, Inc.
Mail Stop: 76-4K
20450 Stevens Creek Blvd.
Cupertino, CA 95014 USA
Paul Sweazey, Apple Computer