Archived from groups: comp.sys.ibm.pc.hardware.chips (
More info?)
On Wed, 09 Feb 2005 23:31:30 -0500, George Macdonald
<fammacd=!SPAM^nothanks@tellurian.com> wrote:
>On Wed, 09 Feb 2005 09:33:47 -0500, Robert Myers <rmyers1400@comcast.net>
>wrote:
>
>>On Wed, 09 Feb 2005 08:21:39 -0500, George Macdonald
>><fammacd=!SPAM^nothanks@tellurian.com> wrote:
>>
>><snip>
>>
>>>
>>>>I made an intense effort to understand what was going on with process
>>>>technology when all the surprises came down at 90nm, but since then
>>>>I've lost track of process technology. If Intel really has lost the
>>>>playbook, that would be news, but I don't really believe it.
>>>
>>>What's going on with process tech does not really have to be understood at
>>>the detail level to see the picture. IBM chief technologists, among
>>>others, have told us of the "end of scaling" - Intel has demonstrated the
>>>effect with 90nm P4. We know, as Keith has said right here, that the two
>>>critical issues involved are power density and leakage. OTOH nobody is
>>>talking of abandoning 65nm and lower, though they do talk of increasing
>>>difficulty.
>>>
>>But I don't know whether to take "the end of scaling" seriously or
>>not. What about nanotubes?
>
>I think it's serious OK -- we already have evidence -- and I'm not sure
>which part of the problem nanotubes solves...
Mobility. Faster gates at lower voltage, smallest possible voltage
being the goal of low power operation.
http://www.eetimes.com/at/news/OEG20031217S0020
I've got a decent physics education, but I'm not a solid state
physicist and certainly not a device engineer. I am pretty quick with
google:
nanotube transistor mobility electron OR carrier.
Carbon nanotubes also have very attractive thermal properties. They
also currently cost about as much, pound for pound, as industrial
diamonds.
>besides which major change in
>material like that is bound to have an extended development time.
>
Don't know how to evaluate that. There's a company nearby I could
walk to that thinks it's going to revolutionize memory (memory always
comes first, doesn't it?) surviving on venture capital. They'd better
come up with something pretty quick.
>>It doesn't matter, anyway? Hell, I don't know. Suppose you could
>>raise the computational density by a factor of a thousand. What kinds
>>of robotic widgets might we see as a result, for example?
>
>Given the way govt. is working those days I'm very suspicious of the way
>robotic anything gets abused. Take a look at the last 3 paras here
>http://www.edn.com/article/CA185948.html for what CARB was experimenting
>with 3 years ago. Take a drive through the U.K. and you'll see their
>highways lined with electronic snitches - the latest models are buried in
>the road so you can't even see them.
>
Creepy. Embedded microelectronics in cars already don't work.
Mustn't confuse what you can do with existing embedded electronics
with what would be possible if the rules really changed. Advances in
AI would be nice, but there is, as far as I can tell, an esentially
inexhaustible demand for cycles in the business of motion dynamics and
kinematics.
>>>I wouldn't go so far as to say that Intel has "lost the playbook" but their
>>>ego seems to be getting in the way when technology sharing is the way the
>>>rest of the industry is moving.
>>>
>>Intel is a cash cow. It's a weak defense, but they do behave better
>>than M$, which completely substitutes market domination for
>>competence.
>
>See my post on RHEL 4.0 on dual AMD64 - M$ and maybe even Sun must be
>worried.
AMD-based server doesn't even make it into the top ten on $/tpmc:
http://www.tpc.org/tpcc/results/tpcc_results.asp?print=false&orderby=priceperf&sortby=asc
Power doesn't. Xeon does. Itanium does. I doubt very much that
anybody at Intel is in a panic.
>As for Intel, I wonder how many $billions they've pissed away on
>efforts to proprietarize the architecture?
>
Watch for Intel to push RAS. And push, and push, and push. Think
Centrino.
>>>>As to performance, which I've also lost track of, Power5 and Itanium
>>>>seem to have run away from the pack on CFP2000. That's the horse race
>>>>that Intel wants. As to the pack, AMD is in the hunt, but only just.
>>>
>>>Really? I'm baffled as to why that's what Intel wants.
>>>
>>
>>Keith and I effectively already had that discussion. Intel wants
>>enterprise applications locked onto Itanium the way they are locked
>>onto IBM mainframes. In that horse race, x86 is a sideshow--or Intel
>>wants it to be a sideshow. ;-).
>
>Keith likely said the same but.... ain't gonna happen. Anyway, FP matters
>little for "enterprise applications" - sorry, I don't see it.
>
No, but the $/tpmc numbers do. RIP Ken Olsen. I didn't mean to imply
the Keith agrees with me, but we have discussed Intel's mainframe-envy
and how it plays out as a business strategy.
>>>>CFP2000 not a realistic measure of real-world performance? Probably
>>>>not, but then what is, other than your own code? Yes, it is easier to
>>>>write naive code for AMD processors than it is to write naive code for
>>>>NetBurst or for Itanium.
>>>
>>>Horses for courses!
>>>
>>Yes, indeed. And that's why I get so bent out of shape about some of
>>the choices our esteemed national assets, er, laboratories, have been
>>making in hardware. Problems can define hardware, but it can (and
>>actually does) work the other way around.
>
>Don't you think this is just a fact of the change in economics: (nobody*
>can afford a modern day equivalent of a Cray. Even though the Japanese
>have done it, it's basically a boat-anchor.
>
Oh, the cluster poster to comp.arch who has such contempt for my
wisdom got on my case for dissing the Earth Simulator, too.
Megabureaucrat projects: think Donald Trump. Real estate, staff,
power, ego. Earth simulator plainly does well on my touchstone
calculation, the FFT, but I've been told that on less ideal
calculations that require global communication it self-partitions into
"islands of performance."
As the alternative, think tiny, low power stream processors with a
sizzling commodity interconnect, not yesterday's embedded processor
with an undersized interconnect and a custom router. Custom Cray
processors don't make any sense? Probably not anymore. Commodity is
the right word. The national lab's latest pick just chose the wrong
commodity (out of date embedded microprocessor) to build on.
<snip>
>>>The way I see it, unless some mind-boggling new technology is discovered,
>>>the only "something else" to switch to from oil is nuclear. Of course, the
>>>way the media has put it, the masses seem to have this weird idea that oil
>>>supply is just going to dry up one day/year/decade... which is absurd.
>>>
>>If you ignore the ravings of the Malthusians and just look at what the
>>U.S. govt. is putting out, there are some interesting things
>>happening, and I'm never quite sure what's real and what's show.
>>There is an ORNL report that says, effectively, that, if you include
>>things like tar sands, you can forget about ever seeing a peak in oil
>>production, unless something dramatic happens to affect human
>>longevity.
>>
>>OTOH, some of the noise about the hydrogen economy is coming from
>>within the U.S. govt., and some from contractors funded by the DoD.
>
>To paraphrase Eisenhower: beware the academic-bureaucratic complex!
>
The academic-bureaucratic complex can move the DoE and the national
labs, but not the energy industry.
>>If you look over the history of oil since 1974, it's been a history of
>>the kinds of surprises that you (and Keith) apparently favor: small
>>individually but important in the sum. Because it has been immensely
>>profitable (much more profitable than developing renewable energy
>>sources) people have just gotten smarter and smarter about finding and
>>extracting oil. As far as I can tell, none of the technology
>>developments that have reshaped the industry (albeit very quietly)
>>were foreseen in 1974. Meanwhilst, the revolutions that were supposed
>>to happen (nuclear, for example) still haven't happened.
>
>The technology of the petroleum industry hasn't really changed that much in
>30 years - a couple of new processes to bolster fine tuning of existing
>octane production... and banish aromatics and other unsaturates. The
>biggest change has probably been the disappearance of the small
>"tea-kettle" refiners.
>
Seismic tomography just keeps getting better and better, the cost of
finding new oil has *dropped* over the last quarter century or so
(because people have gotten much smarter about where they look), and
people keep revising their estimates of what can be extracted upward
(sometimes, admittedly, not always with complete honesty).
>>One read is that "renewables" and "the hydrogen economy" coming from
>>Washington are really a message to oil-producing states: "We don't
>>need you." For all I know, some deeply cynical person inside the
>>government was thinking that way in 1974. The fear of even the
>>possibility of realistic alternatives to oil is what has kept OPEC in
>>line.
>
>Could be but OPEC has scientists too - they must know that renewables and
>hydrogen fail on umpteen fundamental, scientific/economic counts.
>
I read somewhere (National Academy of Sciences, I think) that hydrogen
as a fuel has about the same time horizon as a mars mission. Biofuels
are more realistic. They're not getting the emphasis because they
don't keep the attention of the weapons scientists (hydrogen
economy==hot fusion). If the price of oil were stable at a
sufficiently high level, renewables would become much more of a
player.
>>The lesson I draw from the oil business is one that Keith thinks I
>>don't understand: money drives everything. Until someone can count on
>>making the same kind of money displacing oil they can make by
>>producing it, people will continue to get smarter about producing oil
>>than to look for ways to displace it.
>
>And of course the petroleum companies are well placed to make that decision
>to "displace".
>
Maybe not as well placed as they'd like, but, to the extent they're
playing, that's the game.
><<snip>>
>
>>>As I've said before, steady progress with the odd discontinuity is fine
>>>with me; it's also the way that the application of science to engineering
>>>solutions has traditionally worked, with few exceptions.
>>>
>>
>>It's hard to argue with a statement like that since steady progress
>>with a finite number of discontinuities covers a pretty broad class of
>>functions. You do seem to be ruling out functions that aren't
>>Riemann-integrable. ;-).
>>
>>>As for AMD, we'll see if they can come up with something to tackle the
>>>notebook market... but there's nothing about Centrino which changes or
>>>defines any rules.
>>
>>Oh, but I think it did. Everybody's got a wireless laptop, and
>>Centrino is the brand of choice. Big marketing score for Intel at a
>>time when they did just about everything else wrong.
>
>OK - marketing score.... au suivant!;-)
>
Itanium-based Enterprise servers have all the RAS features of z-series
and they are much, much less expensive.
>>Centrino isn't tied into connectivity in any kind of fundamental way,
>>but the drumbeat of the message is there: it isn't the processor
>>that's important, anymore, it's the whole platform. That's the battle
>>Intel has defined, and PCI-Xpress, Advanced Switching, and heaven only
>>knows what else are going to stomp Hypertransport. I understand why
>>the crowd here isn't pleased emotionally, but, unless those emotions
>>gain wider acceptance (something like what probably is happening to
>>Microsoft), Intel will do just fine.
>
>But Hypertransport and PCI-Express play together - stomping is not
>required. When Intel does its on-chip memory controller they'll need
>something equivalent to HyperTransport;
>no doubt AMD will develop from what
>they have. Basically Intel has not been allowed to proprietarize their
>"platform"... the game is open for the foreseeable future.
Right. As long as you're willing to stick a bridge in there, you can
hook up to infrastructure that's driven by Intel architecture.
RM
Facebook
Twitter
Instagram