Dual-Celeron Project Update
Scrap-heap. I was hoping to be able to get a pair of 300A PPGA Celerons pre-tested at 450 MHz @ 2.0v. The dealers I know of that were selling the tested slot 1 Celerons aren't able to get the PPGA's at all. So I was stuck trying to decide whether to buy some untested PPGA's. (The reason I don't want the slot 1's is that they are too difficult for me to perform the dual-modification on: You can buy a pre-wired "slocket" card for $22 that is dual-ready, and allows you to plug the PPGA into a slot 1 board). I called several dealers, trying to get any customer feedback on over-clocking success. All were similarly and suspiciously tight-lipped. I asked them all whether they had signed some kind of non-disclosure agreement as part of their distribution contract. One of them said, yes. I don't think it's any coincidence that none of dealers would comment on over-clockability, and that the two dealers who had advertised pre-testing on the slot 1's were now unable to get the PPGA's. I decided Intel didn't want me doing this, and that they were probably going to have their way like it or not, so I've let it go. Duals will probably come into vogue and I'll get it when it hits the "sweet spot".
As for Intel's reasons for squashing the over-clocking bug, I have to admit that they have some very justifiable reasons. This article explains pretty well:
hardocp.com Click on "Letters from Karl"
wily
_______________________________________________________
Date: Tue, 08 Dec 1998 09:41:16 -0800
From: Karl Andrews <kandrews@Removed.com>
To: DV-L@dvcentral.org
Subject: Why not to overclock
Message-ID: <366D64BC.65F948A@Removed.com>
Let me say up front that I work for Intel, and that this is not an official Intel statement, only my opinions. That out of the way, I help design microprocessors, so maybe I can shed a little light on this subject.
Yes, with the currently available crop of chips, you can usually get away with moderate overclocking. What you are trading off is reliability and chip life. There are certain semiconductor failure modes that are speed sensitive. We design chips with these in mind for a certain (rather long) lifetime. Overclocking reduces this, but many people don't care as they will have traded in their machines long ago. We care, because we honor the warranty on that chip, no matter who has it now, so overclocked chips cost us extra money in support.
Also, faster chips can be sold for higher prices, right? When we test manufacturing batches, we sort them by maximum reliable speed. If a 333 MHz chip was capable of running reliably at 350 or 400, don't you think we would be selling it at that speed, with it's correspondingly higher price? Whatever you may think of Intel, we aren't stupid. The speed-sensitive error causing that lower speed rating may or may not be significant to your application, but how do you know? We don't label the chip with the type of error, only the maximum reliable speed.
Anyway, the newer chips coming out soon have more effective speed control methods built into them, so overclocking will soon become a dead subject. Why do we go to such great lengths to prevent hobbyists from experimenting with their personal property? Ah, if that were all it was...
There are people who make their living by grinding off the speed labels on our chips, and reselling them at higher prices. The unsuspecting buyer of a system with one of these chips doesn't know this, they were only interested in maximum speed at minimum price. Then when problems pop up, who has to pay for replacing that chip? Not the overclocker, they are offshore somewhere; and not the screwdriver shop that assembled the system with grey-market components, they too are long gone. Who is still around to catch the blame? Intel. That's why we care. It comes out of my paycheck in the form of a slightly smaller year-end bonus, so I care, personally.
- Karl
2nd Letter from Karl
Thanks for the heads up. I guess I'm on the way to notoriety.
- Karl
--
The avalanche has already started.
It is too late for the pebbles to vote. - Kosh Naranek, Babylon 5
3rd Letter from Karl
This shouldn't really be a surprise. While not advertised widely, it is
mentioned in the public developer info at www.intel.com. Just search for
"overclock" at the main page and you will get at least three hits, one of which
says the following:
"The Pentium® II processors which are optimized for the 100 MHz system bus have
circuitry built in which prevents overclocking. If the speed is set too fast on
these processors, the system will attempt to boot and then reset itself. When it
comes back up the processor speed will be reset to the default value and will
need to be reset to the correct speed in BIOS."
While this refers specifically to the T440BX server board, I have no reason to
think other chipsets would behave differently.
The specific chip I am working on will not be on the market for several more
years, but the speed at which the individual chip is rated is hard coded at test
time by setting something similar to an EPROM. I believe the Pentium II used a
similar process.
You have a popular board. I have already gotten four messages about this.
Several of the authors have rather strong opinions on the subject and seem to
feel that I can do something about it. I don't speak officially for Intel, and
have no influence on corporate policies, I only design chips and sometimes
express a personal opinion. Anyone who wishes to influence Intel policy on any
matter should participate in one of the forums accessible from the corporate
home page, vitriolic messages to me are a waste of net bandwidth.
- Karl
--
The avalanche has already started.
It is too late for the pebbles to vote. - Kosh Naranek, Babylon 5
4th Letter from Karl
(here I asked Karl if he would like the letter taken down, due to me posting it in the first place without even asking - KB)
If you take my letter down soon, I don't think there will be a problem. Intel
has pretty effective methods of dealing with email bombs and such. It isn't
like this is the first case. I'm at 23 letters now, and there is certainly a
wide range of response types. A few have even thanked me for the insight into
the situation.
If I had anticipated that my letter would have made it to this forum, I
certainly would have phrased things differently, and perhaps put in a little
more technical data, but I was addressing people who do digital video
professionally or as a hobby (which is what I do) and aren't really interested
in manufacturing details.
On the other hand, I probably would not have said anything at all, as I do not
have time to respond to all that mail. Many I would not respond to anyway, but
there are some who were reasonable and wanted to discuss the subject or had
reasonable questions that I could answer without getting into sensitive areas,
and I feel badly that I cannot afford the time to correspond with them.
Thanks for handling the situation in such a civil manner,
- Karl
--
The avalanche has already started.
It is too late for the pebbles to vote. - Kosh Naranek, Babylon 5
5th Letter from Karl
You might make a link to a page containing all of our correspondence, or at
least my side of it. Some calmer heads might be interested in the whole
picture, and others could see that I am not the proper person to talk to and
find out where to go to try to change Intel's mind about whatever they disagree
with. One appropriate place to discuss overclocking is the Intel.etc forum at
support.intel.com.
Some people may not be aware that Intel operates about 40 public forums on our
support web pages. Most are dedicated to specific products or technologies, and
have specific engineers assigned to them to respond to questions. We even have a
person whose responsibility is to monitor the outside newsgroups and generate a
summary of what is going on relative to Intel (talk about a dream job - imagine
being paid to cruise the wires!).
- Karl
--
The avalanche has already started.
It is too late for the pebbles to vote. - Kosh Naranek, Babylon 5
|
