Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Introducing "No-Moore's Law"

Author: Robert Hyatt

Date: 20:46:51 03/07/03

Go up one level in this thread


On March 07, 2003 at 17:23:26, Matt Taylor wrote:

>On March 04, 2003 at 14:40:53, Robert Hyatt wrote:
>
>>On March 04, 2003 at 13:03:23, Matt Taylor wrote:
>>
>>>On March 04, 2003 at 00:24:27, Robert Hyatt wrote:
>>>
>>>>On March 03, 2003 at 22:33:57, Jeremiah Penery wrote:
>>>>
>>>>>On March 02, 2003 at 23:24:16, Robert Hyatt wrote:
>>>>>
>>>>>>On March 02, 2003 at 22:42:59, Jeremiah Penery wrote:
>>>>>>
>>>>>>>On March 02, 2003 at 10:17:08, Robert Hyatt wrote:
>>>>>>>
>>>>>>>>On March 02, 2003 at 00:33:11, Matt Taylor wrote:
>>>>>>>>
>>>>>>>>>On March 01, 2003 at 20:10:49, Robert Hyatt wrote:
>>>>>>>>>
>>>>>>>>>>So if Intel ships 3.0 today, that's what I buy.  If they ship 2.8, that is
>>>>>>>>>>what I buy.
>>>>>>>
>>>>>>>That's exactly the point.  Why release the 3.0, when you'll buy the slower 2.8?
>>>>>>
>>>>>>_what_ is "the point"???  if we believe your scenario, then it costs the same
>>>>>>to produce _either_ so their profit is the same whether I buy a 2.8 or a 3.0.
>>>>>
>>>>>Exactly, again.  Why should they produce the 3.0 today if they make the same
>>>>>profit from the 2.8?  As long as they're faster than the competitor, they don't
>>>>>need to be any faster.
>>>>
>>>>I suppose it is time to give up, but one more time.  If vendor A produces a
>>>>processor at 2.5, and vendor B produces a processor at 2.55, I will buy the 2.55
>>>>processor, but only if it is close to the 2.5 in cost.  If Vendor B offers me
>>>>a 3.0, I will not only buy it, but I will pay _more_ because it is significantly
>>>>faster than the competition.  The wider the gap, the more I will pay.
>>>>
>>>>I'm not sure why _that_ is hard to understand...
>>>>
>>>>You can sell 2.55 at X dollars today, and in 6 months you can sell 3.0 at X
>>>>dollars since your competition is now at 2.95.  Or you can sell 3.0 today and
>>>>with the wider performance gap, you can widen the price.
>>>>
>>>>Simple economics, IMHO.
>>>
>>>It also raises the prices on lower-clocked chips. Many people don't buy high-end
>>>because it's too expensive. By staggering release, Intel avoids bidding too low
>>>on the price of their chips. The desktop market is a -big- market. If Intel is
>>>able to make $10 more per chip, that's a lot of profit simply because of sheer
>>>sales volume.
>>
>>I understand.  But they will make _more_ than 10$ per chip on the top-end as
>>the faster they go, the more folks are willing to pay.  Low end can stay where
>>it
>>is.
>
>If that's true, then why haven't the prices for high-end chips changed over the
>years? A better market strategy would involve keeping prices the same for
>lower-clocked chips and introducing more expensive high-end chips. Every time
>Intel releases a new high-end chip, the prices for their other chips drop.
>
><snip>
>>>Who is to say that AMD isn't doing the same thing? Don't you find it a little
>>>peculiar that they both release new chips at the -same- time? It is almost like
>>>one waits for the other to catch up.
>>
>>
>>If AMD does it, they _are_ grossly incompetent, for obvious reasons.   #2 wants
>>to
>>out-do #1 to capture part of its sales share.  barely keeping up or falling
>>behind certainly
>>would question their marketing intelligence...
>
>Either AMD's marketting dept. is incompetant or they disagree with you. They
>felt it would be better to delay Clawhammer until September rather than release
>in January. The same thing happened with the K7, and it has happened with
>various strains of the K7 too.
>

Yes, but what is the reason?  Marketing?  Or production issues?


>>>>>>I do _not_ see any reason to "hold back" on the top end.  But this is an
>>>>>>argument that can _never_ be ended because there is no way to prove it.  You
>>>>>
>>>>>It's an argument that can never be ended because you absolutely refuse to admit
>>>>>that you're wrong about it.
>>>>
>>>>
>>>>Or vice versa?  pot, kettle?
>>>>
>>>>You offer _zero_ evidence.  Engineers say they push as hard as they can.
>>>>You say they don't.  So why would I admit I am wrong when you offer _nothing_
>>>>to _prove_ that I am???
>>>>
>>>>I can't find an engineer to support your theory.  Because it makes no sense from
>>>>any angle.  You might try asking a couple to see if you get different answers
>>>>from what I got.
>>>
>>>No, we say engineers report the limits to marketting, and marketting staggers.
>>>If they get you to buy a 2.8 instead of a 3.06, they haven't made any profit
>>>(yet), but Intel is also looking toward the long term.
>>
>>
>>This argument just doesn't work.  They have to capture _me_ today, when I am
>>ready to buy.  And then for 3 years, or 4 years, I am _not_ going to replace
>>this
>>machine unless some overwhelming requirement comes to the forefront that
>>forces me to upgrade for something other than obsolescence.
>
>Maybe AMD does, but Intel doesn't. You aren't a major buyer. Dell and Gateway
>are. The teeming masses buy prebuilt computers from major vendors, and most
>major vendors don't even sell AMD. The ones who do usually sell AMD as the
>"low-end economy PC" rather than being on-par in performance for less money.
>Since most people believe in the idiom, "You get what you pay for," they'll buy
>the 3.06 GHz Intel because it's more expensive.
>
>>If they release 2.8 today, and that is the best around, I will buy it.  I will
>>_not_ buy
>>a 3.06 in 6 months.  Nor will I buy a 3.6 in one year.  I will probably buy what
>>they
>>are shipping in 36-48 months.
>
>If they were selling a 5 GHz today, they would probably be selling a 5 GHz in 2
>years, and you wouldn't be enticed at all to buy more of the same stuff. The
>average lifespan of an x86 core is something like 5 years. You will upgrade
>twice during that time, but you probably wouldn't upgrade twice if they hadn't
>produced anything better.

That makes no sense.  If they can produce 5ghz today, they can _definitely_
produce 10ghz in 2 years.  Moore's law is still functioning to an extent.


>
>>If they want to choose today between selling 2.8 and 3.06, they can do it and
>>make _more_
>>money simply by pricing the 3.06 higher, which they _are_ doing.  And those that
>>want the
>>fastest will pay dearly for it.  Delaying the 3.06 for a few months makes no
>>financial sense
>>whatsoever, unless you believe there are folks that upgrade every time the clock
>>jumps.  I'm
>>sure a _few_ do this, but not even .001% (and that would likely be a high
>>estimate).
>>
>>Now granted, if they want to sell a cadillac without leather seats today, at a
>>cheaper price,
>>and then start selling leather seats 3 months from now at a higher price, that's
>>ok.  But if
>>I buy one with cloth today, I will _not_ be buying one with leather in 3 months.
>> They won't
>>sell me again for _many_ years.  The profit will be higher on the leather seat
>>version, so it
>>makes more sense to sell 'em _now_ at an even higher price than the non-leather
>>version and
>>let folks pick what they want and what they can afford.
>
>On a Caddy that may be true. When the same manufacturing process yields a 2.8
>GHz Pentium 4 and a 3.06 GHz Pentium 4, they do better by appealing to supply &
>demand and letting people buy the 2.8's at a hefty price and then rolling in the
>3.06's. I know a fair number of people who upgrade every year which is the
>difference between the 2.53 and the 3.06.

It should be at least 2.53 to 3.8...  if we are going to double every 2 years.


>
>>I don't doubt that they _might_ delay a new chip to deplete old inventory.  But
>>when they
>>release 3.06 I do _not_ believe that they could ship 3.2 that same day if they
>>wanted.  It
>>just doesn't make financial sense.
>
>Not in the short-term, but in the long-term it does.

How?  if they can make 3.06 today, they ought to make 3.8 in 6 months,
assuming linear growth at 2x every 24 months.


>
>>> If they want you to buy a
>>>4.0, they want to widen the gap so you'll be more enticed. If the 5.0 had been
>>>available when the 2.8's were, you would grab the 5.0 instead of the 2.8. Given
>>>that you do have a 2.8, you're more likely to buy the 5.0.
>>>
>>>I think overclocking scenarios are quite interesting. If Little Tommy overclocks
>>>his Pentium 4 1.5 GHz (in the era of the 2.0 GHz chips) to 2.6 GHz, that is a
>>>lot of headroom. Perhaps 2.6 GHz is actually a little much and it's silently
>>>failing. Still, the chip was surely capable of more than 2.0 GHz. And what if
>>>this is commonplace?
>>>
>>>There really is nothing more than circumstantial evidence, but it seems too
>>>fantastic to be coincidence.
>>
>>
>>What about overclocking the _top-of-the-line_ chips?  That seems less successful
>>than
>>overclocking slower chips that are probably nothing more than remarked faster
>>chips off
>>the same fab line.  I've never said overclocking the "slow boys" is that bad
>>although there
>>is obviously risk.  But overclocking the front-line offerings seems unsafe.
>
>I haven't done any overclocking, but I do read reviews from time to time which
>generally include all the overclocking info. It seems to me that slow ram, slow
>chipsets, and locked multipliers are more the issue than what the chip is
>capable of running at.
>
>I've heard of many cases where someone takes a processor and clocks it well
>beyond the fastest offering for both Intel and AMD. The case I gave above it
>real -- someone I know overclocked a 1.6 GHz Pentium 4 to 2.66 GHz. Others have
>clocked Pentium 4's to 4 GHz. When 2 GHz was the limit on the P4, some were
>clocking it to 3 GHz.
>
>-Matt

I have seen that.  A company in Huntsville, Alabama was clocking a 16mhz 386
to 40mhz way back when, with a secret cooling system.  I was asked to evaluate
it by some investors in Birmingham.  When I asked the right questions (IE how
many chips does it take to find one that will run at 40ghz?) the fallacy in
their company plans became apparent.  It failed.

It was impressive (that one chip/machine) but the idea failed as did the
company as they would _never_ deliver 40mhz to the masses.  The chips could
simply not take it.




This page took 0.03 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.