IBM gives Moore's Law a punch in the face

February 28, 2002 1:55 PM

IBM gives Moore's Law a punch in the face by developing a 110GHz silicon germanium microchip. Only for use in ultra hi-tech environments right now (network infrastructure, military, etc.), of course. What other things could these processors be useful for? Finding vaccines? Genome mapping? SETI? And how many years before they're mass-producible and inexpensive enough for consumer use?
posted by andnbsp (11 comments total)

That's a fantastic lead in =)

What I'd want to use these things for - 3D rendering. Mmmmm.
posted by SiW at 2:01 PM on February 28, 2002

Now don't get *too* excited. The article says that IBM already has chips running at 80GHz, so this isn't a huge leap. I think that these chips are much simpler than a general purpose microprocessor like you're using right now, thus they can run much faster. If you built a Pentium 4 using this process, I doubt it would be able to run anywhere near 110GHz, as it would have much longer paths and greater propagation delays due to its complexity.

So to answer your question, andnbsp, the processors won't be useful for much more than high-speed networking. The process they are built with will eventually be used to make general purpose CPUs, but this isn't going to mess with Moore too much.
posted by whatnotever at 2:15 PM on February 28, 2002

Whoa whoa andnbsp. It should be remembered that ICs (Integrated Circuit) are not CPUs. All CPUs are ICs, but not all ICs are CPUs.

ICs often perform very simple functions that, in many cases, don't even require a program to be present in the chip. An example is the 555 timer IC that many people used in their electronics classes at school.
Although most ICs aren't running at Ghz rates, such as IBMs, (they simply don't need to in most cases) they can run at far higher clock speeds that we associate with 'regular' CPUs because the tasks they do are predefined, require no memory access, and/or are extremely simple.

My knowledge of electronics is quite rudimentary, so all corrections are welcomed, and indeed encouraged.
posted by wackybrit at 2:30 PM on February 28, 2002

SiW, thats exactly what I was going to say (both comments).
posted by tomplus2 at 2:36 PM on February 28, 2002

Does anyone have any hard data regarding Moore's Law, specifically dealing with where this chip fits in on the curve with other such chips? I poked around on Google for a bit, but all I found were outdated charts and some chartjunk from Intel.
posted by jkottke at 2:47 PM on February 28, 2002

Hey Kottke, have you read Age of Spiritual Machines by Ray Kurzweil? The first chapter has a lot of good info about Moore's law.
posted by ktheory at 2:55 PM on February 28, 2002

what about this Google search?
posted by badstone at 2:55 PM on February 28, 2002

If you built a Pentium 4 using this process, I doubt it would be able to run anywhere near 110GHz, as it would have much longer paths and greater propagation delays due to its complexity.

Since this is IBM we're talking about shouldn't we be relating this to PowerPC structure and not P4's?

But seriously, unless huge CPU Ghz gains are matched by huge gains in bus and memory speed they are pretty much a moot point - sychronization across these even LONGER data paths is even more of a challenge in many cases than dealing with the CPU die. I'd think that the first place you'll see speeds like this is dedicated processors for graphics engines and the like which are much simpler than general purpose microprocessors and do more number crunching than they do data shuffling.
posted by RevGreg at 2:55 PM on February 28, 2002

Carbon Nanotubes could keep Moore's law alive for at least the next twenty years. Smaller = less power = less heat. Current chip design could reach temperatures as hot as the surface of the sun.
posted by jeblis at 3:22 PM on February 28, 2002

I can't find the originaly article, but I also read an article about the same thing a few days ago, and one of the major problems is that it is only a way-way circuit right now, and only use for communications right now. The article said this chip is roughly equivilant (in nature, not speed) to the chip used in cell phones. Wish I could find that in my stupid cache.
posted by jmd82 at 4:13 PM on February 28, 2002

Wired had this to say earlier in the week.

This was intended more an a display of the silicon germanium fabbing process than anything else. And, as many have already pointed out, this is not a CPU.
posted by briank at 4:39 PM on February 28, 2002