Researchers set new transistor speed record
By Donald Melanson 
posted Aug 18th 2006 3:39PM
posted Aug 18th 2006 3:39PM
Ok, so there's no wind in your hair or chance of an immediate fiery death, but a world speed record's a world speed record, and when it could also lead to better and cheaper cellphones and digital cameras, we're all ears. At least that's what could be coming down the pipe thanks to the speedy new transistors created by researchers at the University of Southampton. What's even better is that they've done it with just a simple modification to existing transistor technology, meaning it should be fairly easy to move it from the research stage to the production line. The speed gain was achieved by adding fluorine implants to the silicon layers of the transistor, making the transistor thinner and more effective, ultimately topping out at a speed of 110 GHz, leaving the previous record of 70 GHz in the nanoscopic dust.[Via Slashdot]
sick! so if the current record was 70GHz and we only have around 3.0GHz in the real world, this technology should bring us only a 1.57x real-world speed increase?
so i should expect a MacPro with Dual Processor Quad Core Xeons clocking in at... say 4.7GHz per core, in a couple years?
sounds good to me. :)
ha... those "records" mean nothing to the general public
i expect minimal changes for the time being
Not really about processor speeds, it's BJT technology not CMOS.
The article is about analog chips; op-amps, oscillators, high speed muxes etc.
Processors use FET transistors, not BJT.
Sorry.
This is for a single transistor only. Not the millions they have in microprocessors. But it can still be useful in telecommunications for example.
Put your MacBooks and thinkpads aside and read this comment: the guy knows things
http://hardware.slashdot.org/comments.pl?sid=194385&cid=15931705
Cheers
This announcement falls into the category of "make the claim narrow and it's easier to claim a record." 110GHz is only about 20% of the actual performance record for a semiconductor-based transistor and less than 1/3 the performance of even silicon-based transistors that have been demonstrated. Indeed, bipolar transistors using silicon-germanium as the base (middle) layer attained 110GHz about 5 years ago. The main claim-to-fame here is that implanting fluorine is simpler and cheaper than growing a film of SiGe. However, once you include the cost of integrating the bipolar transistor (good for high-speed RF/analog circuits as well as low-density, ultrafast logic) with CMOS (necessary for any high level of digital logic integration) or even just the cost of passives (resistors, capacitors, etc.) for making good RF circuits, the difference between a SiGe or fluorine-implanted bipolar is negligible. Hence the dubious nature of this claim. As I've implied above, it's also important to realize that bipolar devices were abandoned for digital logic by the early 90's, when IBM stopped using such technologies for its processors. Energy dissipation is just too high (can't get the heat out) and device size too large for high levels of integration. For RF, such technologies work well. IBM SiGe BiCMOS has made great inroads into WiFi and GPS chipsets, for example, and SiGe Semiconductor is selling cellular/WiFi power amplifiers made from SiGe bipolars as well.
I *chortle* for one welcome our speedy transistor overlords.
Seriously though, with bus and subsystems' I/O speed so pitiful, what's the point?
So I suppose there's some fundamental difference between this and the 500Ghz speed mark recently accomplished at Gatech?
http://gtresearchnews.gatech.edu/newsrelease/half-terahertz.htm
You know, this might be a different technology than processors, but that doesn't remove the fact that if the chip makers wanted we could already be enjoying way higher frequencies than today. But since that doesn't make any sense in a business model, then we get just wii increases in clock speed. It's like petroleum and an alternate 'aka' free source of energy; the big market leaders don't want the technology out since all their business models would have to change, which means loss of money input.
Imagine a chip so fast that not even Bill Gates' Windows 3000 could bring it to it's knees. Instant queries, instant response, no stalling. Of course, as the cpu clock rates increase, so would the interfaces and interconnections to RAM(faster RAM of course), HDDs(holographic) and every clockable chip inside the computer. And best of all, I'm confident that the heat wouldn't be a problem, either by a special mix of elements to make up the "silicon" of the computer or a smaller process. One of these days...
Not bad considering the chip fabrications facilities at the university burnt down a few a while back....
http://news.bbc.co.uk/1/hi/england/hampshire/4390048.stm
Yay. More bandwidth for the FCC to sell!!!