Dawn of the Tera Era
It was only a few years after I had bought my first desktop computer in 1983 -- a dual floppy with RAM so low I can no longer remember its size -- that I watched my next- door neighbor install a memory card on the motherboard of his new PC. This top-of-the-line system had a hard drive with half a megabyte of storage. But the memory card that cost $750 held twice that amount -- all by itself. I was awestruck.
It was only a few years after I had bought my first desktop computer in 1983 - a dual floppy with RAM so low I can no longer remember its size - that I watched my next- door neighbor install a memory card on the motherboard of his new PC. This top-of-the-line system had a hard drive with half a megabyte of storage. But the memory card that cost $750 held twice that amount - all by itself. I was awestruck.
Times have changed and this story has taken a seat next to gasoline for $1 a gallon and 5¢ bubble-gum baseball cards. We have entered the Tera Era, where kilobytes, megabytes, and even gigabytes are passé. Hitachi coined the name last week to publicize the release of its second-generation 1-TB hard drive, the Deskstar 7K1000.B. It is faster, more energy efficient, and much smaller than the first-generation drive Hitachi released a year and a half earlier.
As we begin recalibrating our comparisons in trillions of bytes, we will also have to adjust our thinking as to what computers can do for medical imaging. The Tera Era signals a change in not only how much we can store, but what we can do with what we store.
Earlier this year, Intel unveiled a prototype processor that can perform one trillion calculations per second. It is not the first. That honor belongs to a supercomputer built in 1996 that had 10,000 Intel Pro processors sprawled over 2000 square feet. Intel's new processor is the size of a fingernail.
Teraflop computing could lead to extraordinary photorealistic image rendering in a flash. Real-time everyday speech recognition will arrive along with machines seeded with true artificial intelligence. This opens up the possibility of advanced computer-aided diagnosis rather than simply detection.
The day when these advances will arrive is still far off. Intel's new chip is not designed for production but as a proof of concept. It will be used to understand how this chip can communicate with other chips and computers, as well as the software needed to get the most from it. But tera-scale computing is on the way. And with it will come changes in the practice of radiology that can barely be imagined today.
