Design Con 2015
Breaking News
Blog

2020 Vision: Transistors a buck a billion

Blog
7/7/2009 05:00 PM EDT

 3 comments   post a comment
NO RATINGS
View Comments: Newest First | Oldest First | Threaded View
shirazkaleel
User Rank
Rookie
re: 2020 Vision: Transistors a buck a billion
shirazkaleel   10/29/2009 12:13:55 AM
NO RATINGS
With a > billion transistor budget, computer architects should seriously consider going back to undo all the bad decisions made in the history of computer design, such as needing to share and centralize everything. This was needed because hardware was so expensive and needed to be made cost effective. This sharing is the basis of many bugs, which is why it is so difficult to get rid of all the bugs in software. For example, I've seen traffic light controllers that stay green in the cross direction although there is no traffic in that direction. When finally, traffic appears in the cross direction, the lights change against them, and let me through! The problem is caused by aliasing of events - instruction execution is like a tangle of fuses in which only one spot on one fuse is lit. The rest of the fuses are inactive, rendering the rest of the algorithms "asleep at the wheel" and unable to respond to events. Sure, there can be interrupt controllers that can be programmed to field edge triggered or level triggered interrupt events, but most current microprocessors are trying to shovel parallelism into a serial execution device, and in the process putting their algorithm's state machine(s) at the mercy of their processor's hidden, hard-wired state machine(s).

Jonathan.Kang
User Rank
Rookie
re: 2020 Vision: Transistors a buck a billion
Jonathan.Kang   7/15/2009 5:35:28 PM
NO RATINGS
I think the trend is the opposite really. As transistors begin to cost less and less, it won't be the gotta-have-more crowd that will push it but rather, new applications. Think of the PC. Desktop powerhouse computers used to drive innovation. Everyone wanted more speed. Then it got to the point where people just didn't notice the blazing fast improvement in performance as much as they used to. Then someone got the idea "hey, we can take what we made before and shrink it"; and the laptop took over as the primary technology driver. Smaller, lower power, lighter in weight but still good at performance. I think we'll see that the latest and greatest Pentium won't be the hot-button topic anymore in the next 10-years. The latest and greatest ARM (or perhaps Atom) or the that new tiny DSP that runs off power from ambient body heat will be what's coming.

DL Seth
User Rank
Rookie
re: 2020 Vision: Transistors a buck a billion
DL Seth   7/9/2009 10:30:16 PM
NO RATINGS
Reminds me of a quote from a fellow former TI guru, Ronald Shaefer, that I heard some nine years ago?.He told our class of undergraduate EE students to "imagine what you can do with infinite bandwidth, because compared to what we had in the early days of TI, what we have now is infinite", or something like that?.So, to him, what we have now is basically a buck a billion. It?s a great quote, which I?ve thought about often. But, even working for a programmable logic company (founded on sloppy transistor use!), there?s something not quite right about it?.Emerging applications quickly gobble up any performance gains, and nothing, especially in this economy, seems infinite. What would I do with a buck a billion transistors? The same thing I?m doing now with a buck a million. Wanting more!! And realizing that while sloppy use of transistors is the way of the future, exploring the exploding possibilities of a buck a billion is, in the end, still about optimal use of scarce resources. And that a buck a billion will seem nothing next to a buck a trillion.

Radio
NEXT UPCOMING BROADCAST
EE Times Senior Technical Editor Martin Rowe will interview EMC engineer Kenneth Wyatt.
Top Comments of the Week
Like Us on Facebook

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
EE Times on Twitter
EE Times Twitter Feed
Flash Poll