Intel follows lead of AMD, introduces model numbers

[Tony Hill]

Aren't IBM making Xilinx's Spartan 3 FPGAs at 90nm? I have a couple on
my desk, but don't know if they're from IBM or UMC.

Hmm.. you may be right on that one. I did a bit of quick searching,
and it definitely looks like the bulk of the Spartan 3 chips are
coming from UMC, however they have signed IBM up as a sort of
second-source for 90nm production. I'm not sure how many (if any) of
the chips IBM is shipping now, but I get the impression that they are
at least shipping some 90nm chips to Xilinx.

This is a bit of an interesting case that I totally missed, Xilinx
REALLY looks like they got the jump on the rest of the industry with
that Spartan 3 chip.

 

[Tony Hill]

If you want a _lot_ of free cooling, as well as high-bandwidth
connections, I have a suggestion for you:

Place your cluster in Longyearbyen, on Svalbard: N78 13'

This is almost certainly the northernmost point in the world with the
infrastructure to support such a facility.

While it may be far north, the average temperature is not much
different from some closer destinations in the North of Canada.
Yellowknife (N62 47') has about the same annual average temp (about
-5) as Longyearbyen, though the latter is somewhat more consistently
cold (~ -15C for January and 6C for July vs. -28C and 16C for
Yellowknife).

Still, I'm not sure that either Yellowknife of Longyearbyen are likely
to get any big computing clusters exported there any time soon. The
costs of cooling are likely rather small as compared to the costs of
the people to run the thing.

 

[Tony Hill]

That's a bit strong IMO. There are some who believe there's more to this
than meets the eye. I know I read the initial article at Electronic News
and, after spending much too long trying to refind it, I even found a URL
link to it elsewhere. Turns out Electronic News has removed the bloody
article from their site.<GRRR> The only quoted para from it I can find
does mention .13u specifically for IBM's own parts but 90nm was part of the
story.

There seemed to be a few different stories being melded together and I
clearly remember that the 5% number came from someone mixing one set
of numbers from one story with another bit from a different story.

Low first time yield? Having to redesign to fit the process will
possibly drive the customer to his alternate supplier... as appears to be
happening with Qualcomm, nVidia and Xilinx... at least for the moment.

Certainly. I'm not saying that this isn't a problem, just that it's a
different problem than yields and therefore has a different solution.
Of course, IBM also seems to have some yield problems as well to go
along with this design issue.

90nm is proving to be a tough nut to crack, IBM's having some
problems, Intel's 6+ months late and still having lots of trouble, AMD
has delayed their parts for a good 4-6 months and it still remains to
be seen how well they'll deliver. Definitely more problems going
around this time than with the 130nm process.

No - 90nm *has* been part of the yield story in general. TSMC & Qualcomm
claim to have working 90nm production. IBM has definitely stumbled here
and TSMC and UMC are getting a windfall as a result.

It is important to remember that TSMC and UMC are after a slightly
different goal with their production. For them it's largely a minimum
cost function, while top-end performance is not as important. I would
guess that a company like Intel could get about as much performance
out of a design at 130nm as TSMC/UMC could get at 90nm. However the
latter would achieve that performance at a much lower price point.

TSMC and UMC are still coming out ahead here because they can offer
similar performance for lower price, but I don't see them taking over
all the world's production just yet. A lot of companies still seem
very interested in using IBM as a foundry for high-end parts.

 

[Rob Stow]

If you want a _lot_ of free cooling, as well as high-bandwidth
connections, I have a suggestion for you:

Place your cluster in Longyearbyen, on Svalbard: N78 13'

This is almost certainly the northernmost point in the world with the
infrastructure to support such a facility.

That is one thing Canada does not have - a place that is cold
all year round that also has the necessary infrastructure.
You couldn't get much further north than Edmonton before the
infrastructure disappears - and Edmonton would be sufficiently
cold only about 6 or 7 months per year. There are many towns
and a few small cities in Canada that are further north than
Edmonton, but not with the necessary infrastructure.

Kind of funny that we curse the cold for half of the year
when perhaps we should be developing the infrastructure
necessary to sell it

 

[Robert Myers]

On Wed, 17 Mar 2004 08:55:30 -0500, Tony Hill

Still, I'm not sure that either Yellowknife of Longyearbyen are likely
to get any big computing clusters exported there any time soon. The
costs of cooling are likely rather small as compared to the costs of
the people to run the thing.

Dunno about Yellowknife or Longhorn, but I'll bet there are existing
hydro power sites where the average temperature is pretty low. How
much bandwidth can a 765 kV power line carry? Rights of way exist.
Power plants exist. If you can staff a hydro power plant, you can
surely staff a computer center.

RM

 

[Martin Heller]

That's a bit strong IMO. There are some who believe there's more to this
than meets the eye. I know I read the initial article at Electronic News
and, after spending much too long trying to refind it, I even found a URL
link to it elsewhere. Turns out Electronic News has removed the bloody
article from their site.<GRRR> The only quoted para from it I can find
does mention .13u specifically for IBM's own parts but 90nm was part of the
story.




Low first time yield? Having to redesign to fit the process will
possibly drive the customer to his alternate supplier... as appears to be
happening with Qualcomm, nVidia and Xilinx... at least for the moment.




No - 90nm *has* been part of the yield story in general. TSMC & Qualcomm
claim to have working 90nm production. IBM has definitely stumbled here
and TSMC and UMC are getting a windfall as a result.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

nVidia and Xilinx where probably put on low priority by IBM in favor
of AMD, see:
http://www.theinquirer.net/?article=14752
which really sounds more reasonable than the low yield rumors.
Yours,
Martin Heller

 

[Nick Maclaren]

|>
|> nVidia and Xilinx where probably put on low priority by IBM in favor
|> of AMD, see:
|> http://www.theinquirer.net/?article=14752
|> which really sounds more reasonable than the low yield rumors.

That implies that IBM were short of capacity, whether that be
production or process engineers to help the customers. Given
what I know of IBM and AMD's imminent requirements for 90 nm,
that seems a little unlikely. Not impossible, just unlikely.


Regards,
Nick Maclaren.

 

[Del Cecchi]

On 15 Mar 2004 20:18:48 GMT, (e-mail address removed) (Nick Maclaren) wrote:



Practically any exposition that isn't just a snow job is going to
assume that you understand effective mass:

http://en.wikipedia.org/wiki/Effective_mass

If you swallow hard, that page contains a big piece of the puzzle. A
hole or an electron in a solid can be imagined to be accelerated by an
imposed electric field as if it were a free particle with positive or
negative charge, respectively with an effective mass m* (equation
given).

The effective mass can be calculated as a second derivative of the
dispersion relation E(k), where E is energy and k is the wavenumber
(equation also given).

If you need a deeper understanding of effective mass, you probably
want to consult a book on solid state physics, like Kittel, but, since
you gave up on qm, that's probably not going to get you very far.

What does the straining of silicon have to do with this? The crystal
structure scatters waves just like the rulings of a diffraction
grating. The details of that scattering produce the shape of the
energy surface E(k).

Just as you can change the scattering properties of a diffraction
grating by changing the spacing of the rulings, stretching or
compressing the silicon crystal lattice changes the scattering
properties and thus the dispersion relation and thus the effective
mass of holes and electrons.

If you do it right, you can reduce the effective mass, thus increasing
the acceleration of the hole or electron for any given imposed
electric field--the "increased mobility" that is constantly being
referred to in the snow jobs.

The stretching and compression are achieved by growing the strained
silicon on a crystal substrate whose natural spacing is greater or
lesser than the natural spacing of the silicon crystal. Assuming the
substrate is thicker than the strained layer, it will tend to impose
its spacing on the strained silicon layer, thus changing the
dispersion relation, thus changing the effective mass.

The floor is now open to nit-pickers.

RM

Dredging up memories of long ago, when holes and electrons had just been
invented, I recall something about the wave equation in a periodic
potential. The resulting solutions show "particles" which are much larger
than the lattice constant, and lead to the notion of the conduction band and
the valence band. Later on, phonons showed up. But it seems plausable that
mucking with the lattice constant will change the properties of these
solutions to the wave equation. It is not clear that the optimum for
electrons is the same as for holes. Seems to me that there was something
about the product of the mobilities being constant.

del cecchi

 

[Stephen Fuld]

On Wed, 17 Mar 2004 08:55:30 -0500, Tony Hill



Dunno about Yellowknife or Longhorn, but I'll bet there are existing
hydro power sites where the average temperature is pretty low. How
much bandwidth can a 765 kV power line carry? Rights of way exist.
Power plants exist. If you can staff a hydro power plant, you can
surely staff a computer center.

Interesting idea. Plus, you probably don't need it to be that cold, as you
have a ready source of water for cooling. Just put a heat exchanger in the
water and you can cool a lot of air. Or use a water based heat pump system.
And, you might arange a lower per kW price for power given the lack of
distance transmission required. Furthermore note that there is a push by
the FCC, and companies such as Intel (IIRC) to support IP over power lines
in a standard way. Lots of possibilities.

Now if the power companies (at least in the US) weren't so unimaginatively
managed. :-(

So, Terje, would Norsk Hydro consider going into the ISP or similar
businesses?

 

[Robert Myers]

On Wed, 17 Mar 2004 10:57:39 -0600, "Del Cecchi"

Dredging up memories of long ago, when holes and electrons had just been
invented, I recall something about the wave equation in a periodic
potential.

The way that periodic potential affects solutions to the wave equation
really does have alot in common conceptually with the physics of a
diffraction grating. The lattice is just a stack of cross ruled
transmission gratings spaced just so.

The resulting solutions show "particles" which are much larger
than the lattice constant, and lead to the notion of the conduction band and
the valence band.

Bloch waves. The way that Bloch waves are usually introduced, with
Brillouin zones and reciprocal lattices and all the baggage of
crystallography makes my head hurt. I found a presentation that is
relatively free of that stuff:

http://www.techfak.uni-kiel.de/matwis/amat/semi_en/kap_2/backbone/r2_1_4.html

Later on, phonons showed up. But it seems plausable that
mucking with the lattice constant will change the properties of these
solutions to the wave equation.

Plausible? You're a hard man, Del.

It is not clear that the optimum for electrons is the same as for holes.

Plainly, it is not.

Seems to me that there was something about the product of the mobilities being constant.

Not in my memory banks, and Kittel is hidden somewhere in my mountains
of books. ;-).

RM

 

[RusH]

Balderdash. Everyone seems well aware that a P-M will perform
on a par with a P4 clocked about 40% higher. What is stopping
them is the high prices for the P-M and tremendous difficulties
in getting a decent motherboard for the damned things.

Decent ? Asus P2B with iBX will do just fine I think. It works with
tualatin, so little pin tweaking should make P-M happy.

Pozdrawiam.

 

[RusH]

A 1.4 GHz or 1.6 GHz Pentium-M is far more than enough horsepower
for the overwhelming majority of office workers. If it wasn't
for the price and the lack of motherboards, what would make it
the ideal processor for office workers is that it can be used with
passive cooling and it has excellent energy efficiency.

You are talking Via erza and other VIA crap, go buy some - this will
change your mind

If you could use 1.6 GHz P-Ms instead of a 2.6 GHz P4s in the
desktops of office workers, you would suffer no loss in computing
power and between the cpu and the air conditioning you would save
about over one KWh per machine per day.

buy some VIA miniITX boards, go ahead, dont be shy

All that is needed is P-M motherboards and ideally also lower P-M
prices.

It is not. Almost nobody* cares about VIA superlowpower boards.

* being few weirdos sticking those thingies into lunchbox.


Pozdrawiam.

 

[Iain McClatchie]

Tony> I would guess that a company like Intel could get about as much
Tony> performance out of a design at 130nm as TSMC/UMC could get at
Tony> 90nm. However the latter would achieve that performance at a
much
Tony> lower price point.

I think this is a very interesting comparison.

(From http://www.eetimes.com/printableArticle.jhtml?doc_id=OEG20020312S0038&_requestid=33837)

bit cell channel
Intel 130nm: 2.09 um^2 70nm
TSMC 90nm: 1.3 um^2 65nm
Intel 90nm: 1 um^2 50nm

Due to different thresholds (at the very least) the 130nm-generation
Intel xtrs are way faster than the 90nm-generation TSMC xtrs.

But different applications respond differently to xtr speed and onchip
SRAM capacity. In a power-limited chip, especially, big SRAMs may
make
a TSMC-specialized chip cheaper at the same performance than the
Intel-
specialized chip.

It would be interesting to see a graph of SRAM cell size versus a flop
over multiple process generations. My guess is that SRAM cells are
actually getting smaller relative to flops in more recent processes.

 

[Rob Stow]

You are talking Via erza and other VIA crap, go buy some - this will
change your mind

A C3 can't come close to a Pentium M in performance, so
why the heck did you think you needed to introduce VIA to
this thread ?

And I have actually used about 3 dozen C3's so far.
If you want a silent machine for something like driving a
projector in a board room they are a great processor.

buy some VIA miniITX boards, go ahead, dont be shy

And what exactly does VIA make that will take a Pentium M ?

It is not. Almost nobody* cares about VIA superlowpower boards.

Mostly because when you get a C3 you end up with a system
that is adequate for a secretary using MS Office or for watching
DVDs, but good for nothing else. A Pentium M gives you energy
efficiency *and* decent computing power - so I repeat: why the
heck did you think you think you needed to introduce VIA into
this thread ?

 

[Rob Stow]

Decent ? Asus P2B with iBX will do just fine I think. It works with
tualatin, so little pin tweaking should make P-M happy.

LOL. You really think a little pin tweaking will let you put
a 478 pin Pentium M into a Socket 370 ?

 

[Nick Maclaren]

LOL. You really think a little pin tweaking will let you put
a 478 pin Pentium M into a Socket 370 ?

Use a bigger hammer.

Followups set appropriately.


Regards,
Nick Maclaren.

 

[Toon Moene]

If you read the history of the development of quantum mechanics, which
you surely have, it becomes more bearable. Poisson brackets are
mathematically elegant. Commutators are mathematically elegant.
Making an analogy between Poisson Brackets and Commutators was an act
of inspired courage. Presenting that analogy as a formal mathematical
theory long after better ways of understanding what was really going
on (which is what I was confronted with) is a travesty. Teach
history. Teach mathematics. Don't jumble them up and call them
physics.

I recently had to help a 16 year old through the understanding of
logarithms. The school had taken a novel approach: Use history to teach
logarithms - i.e., they offered a treatise based on the research done by
Napier and Euler.

However, the people who thought up this approach probably didn't realise
that those mathematicians were the top-of-the-line in their time and
that, therefore, their reasoning would be hard to understand for a 16
year old who has to do more than just mathematics to succeed at school.

It didn't really help that the Latin Napier and Euler wrote in was
translated and their examples dumbed down.

Math is hard - especially back in the time it was invented.

 

[Terje Mathisen]

On Wed, 17 Mar 2004 08:55:30 -0500, Tony Hill



Dunno about Yellowknife or Longhorn, but I'll bet there are existing
hydro power sites where the average temperature is pretty low. How

Not that low actually: Hydro power requires water that on average is
liquid. :-(

much bandwidth can a 765 kV power line carry? Rights of way exist.

That one is easier: All high-voltage cables in Norway have fiber twisted
around them for comms, originally just for power plant remote control,
but now also for regular TCPIP.

Power plants exist. If you can staff a hydro power plant, you can
surely staff a computer center.

See above: Most power plants in Norway are unstaffed.

Terje

 

[Terje Mathisen]

So, Terje, would Norsk Hydro consider going into the ISP or similar
businesses?

No, for two reasons: The Hydro isn't due to hydro power but the
company's start as The Norwegian Hydro-Electric Fertilizer Company, i.e.
we mostly use the power we generate, instead of selling it.

Besides this, we seem to be 'concentrating on core competencies' by
divesting all non-core businesses, including the worlds largest
fertilizer corporation which will keep the viking ship logo and start
trading as Yara instead of Hydro.

Much closer to home, they are also (sort of) getting rid of the IT
department, which includes myself. :-(

Terje

 

[Sander Vesik]

They are far and away Intel's best *desktop* cpu. I am
amazed that their refusal to market it accordingly does
not run afoul of any industrial laws, particularly in
places like California where energy shortages are a huge
concern and getting worse every year.

I keep expecting places like California to start putting
caps on the amount of power a desktop cpu can draw - much
the same way they have minimum fuel efficiency standards.

By this logic, AMD should also have been marketing the Mobile
Athlon XP for 1U servers, something they haven't done so far
AFAIK, even though it would make for nice cool machines.

Speaking of low power and lots of bang / watt - does anybody
have spec numbers for AMD-s Alchemy line?