|
|
Symmtex LLC, Shimon Yanowitz, President Tel: +972-54-5740025 • Fax: +972-4-8673957 e-mail: shimon_y [at] barak . net . il |
Note: This document is relevant for the period prior to, and ending at Q2 (end of June) 2005.
As of this date, Intel and the rest of the industry, have taken correctional steps, and progression steps,
which have addressed, and have improved upon, the status and the issues described herein.
Intel and the Semiconductor Industry
Analysis for Press and Media Release
By
Shimon Yanowitz
April 15, 2005
3. My Non-Disclosure-Agreement with Intel
4. A Good-Word before the Bleak Outlook
5. Intel Spins Hyper-Threading into Dual-Core
6. Why is My Analysis Different from many Others'
7. Intel Put on Probation in Japan
8. Intel and Asus Sever Business Ties
9. Intel Recalls Flawed 915 & 925 Chipsets
10. Intel's Chain of Chipset Fiascos
11. Intel Fails to Deliver Long-Awaited Mobile Chipset
12. An Even Bleaker Outlook for Intel's Mobile solutions.
13. Intel Fails to Deliver DDR2 Technology
14. Intel Is Hiding Flaws in Sold Chipsets
15. Intel's Motherboards Ridden with Defects
16. Customers Are Walking Away from Intel
17. New Competition Outperforms Intel
18. Intel Losing Ground to Video and Graphic Microchip Vendors
19. Intel's Non-Existent Customer Care
20. The Secret of the Elusive Yield
21. Fighting a Lost Battle against Physics
23. Financial Implications of Production Impairment
25. A Demonstration of the Over-Heating Problem
26. My Background and Credentials
27. Photo of My NDA with Intel
Recent events, mishaps, glitches and a chain of fiascos associated with Intel (see following) are indicative that business is far from usual with Intel.
This analysis surveys the current status of Intel from several viewpoints such as Research and Development, Manufacturing, Marketing and level of Service & Support, and provides the details behind what may evolve into a major collapse of the giant.
For details about me and my background and for my credentials, see My Background and Credentials.
The material, information and analysis brought in this document, are only partial. More detail is available, should interest be expressed.
I have a Non-Disclosure-Agreement in place with Intel. See Photo of My NDA with Intel. Therefore, all the material and information presented in this analysis encompasses what is available in the public domain, as well as an analysis based upon my long and intimate acquaintance with the semiconductor industry.
This analysis does not include information that was disclosed to me in confidentiality.
For the sake of fairness and balance, I would like to start by complimenting Intel for the HT (Hyper-Threading) solution, which has been with us for some time now.
When Intel had introduced HT, it has offered an intelligent and creative move forward in an otherwise progress-less situation.
In essence, what HT meant was this: "We cannot offer you a more advanced processor, or CPU, at the moments, but we can sort-of give you two CPUs in the same package, instead."
It should be noted that not all software applications can benefit from the HT architecture, though.
See The Secret of the Elusive Yield, Fighting a Lost Battle against Physics and The Over-Heating Problem.
Intel has announced on
http://www.washingtonpost.com/wp-dyn/articles/A57608-2005Apr15.html?nav=headlines
This announcement, which seems much too much like an attempt to recycle the Hyper-Threading idea, is far from convincing.
It seems to be a marketing spin attempt, indicative that the bag of tricks of Intel, has, in fact dried-out.
It doesn't take much understanding to figure out that such a spin-type announcement may be driven by some major deficits in the delivery department, and perhaps by some desperation in view of what may be a shaky financial own-outlook.
Many Analysts do not consider Intel's future to be so bleak. Why is my own analysis so different, then?
The plausible answer is that while there are many Analysts out there who have good understanding of the Global Markets and Economy, there seem to be very few who have adequate understanding of the Semiconductor Physics and Technology.
As regards the obvious indicators that many seem to have missed, or to have misinterpreted, well, I don't know is the answer.
In February, the Japanese administration has issued a
warning to Intel, for improper business conduct. The warning indicated that
Intel was abusing its power, to bullying competitors or customers in
Although no immediate measures were taken against Intel, the warning has specified that Intel's business activities and conduct are going to be closely monitored and scrutinized, and that such misconduct and abuse of power will not be further tolerated.
It goes without saying that for a company such as Intel to
arrive in such a situation, is an indication that Intel's business status (at
least in
At about the same time of the
Intel Designs and sells computer motherboards based on its microchips. However, Intel out-sources these boards from established board manufacturers.
After severing ties with Asus, Intel has named three other (much smaller and less reputable) companies as its board suppliers.
Asus is not just an Intel Supplier. It is Also an Intel Customer. It manufactures and sells boards of its own design, some of which employ Intel Chipsets.
Although no explanation has been given to the severance of business ties between Intel and Asus, the plausible explanation is that Asus has been having enough of Intel's Chipset fiascos, which have been inflicting some significant financial damages upon it, as well as seriously compromising its reputation, and has started relying more than ever on alternate CPU manufacturers (AMD) as well as supporting-Chipset manufacturers.
In particular, the 915 Chipset, which should have delivered a significant boost in performance, has failed to do that due to design flaws, and Asus (along with other board vendors) was left alone to incur the reputation consequences, which they were not responsible for. See Intel Is Hiding Flaws in Sold Chipsets.
See also Intel Recalls Flawed 915 & 925 Chipsets, Intel Chain of Chipset Fiascos, Intel Fails to Deliver DDR2 Technology and What is a Chipset?
On
Intel has recalled all of the affected chipsets.
The implications of this recall go far beyond these chipsets. It implies a recall of all motherboards affected i.e. all motherboards manufactured and sold worldwide that have been employing these chipsets.
The affected motherboard, in turn, may have ended up as integrated into full computers sold to customers, or "End-Users".
Since Intel has always refused to deal with End-Users directly, the recall has forced all links in the chain of agents, integrators, dealers and resellers to deal with the consequences, which imply financial, as well as reputation damages.
Furthermore- this policy of Intel has always implied that customers who have lost contact with their former link in the chain, or that have been refused support, have been left without any solution- being forced to write-off their purchase and incur the financial (and potentially- also the reputation) damages.
In this particular incident, there was no estimate of the losses that Intel has incurred as a result of the recall.
This recall has not been an isolated case. Intel's products have been marred with design and/or manufacturing defects both before and after this event. See Intel's Chain of Chipset Fiascos, Intel Fails to Deliver DDR2 Technology, Intel Is Hiding Flaws in Sold Chipsets and What is a Chipset?
The famous "Pentium Bug" that has created a worldwide shockwave at its time, has been long forgotten. When it seemed that Intel had learnt its lesson, a series of flaws and defects in its Chipsets has started to emerge.
In May 2000, Intel has announced that it has discovered a major defect in its 820 Chipset. It has ordered a recall of all motherboards and computers employing this chipset and has promised financial or material compensation to all affected customers.
It has been estimated that over one million computers were affected worldwide.
Although Intel has never released the exact figures, it has been estimated that this recall has cost Intel somewhere between 300 million to half a billion US Dollars in direct costs.
These losses exclude the enormous costs that have been inflicted upon agents, integrators, resellers and dealers worldwide, who have been forced to handle all the logistics and incur the secondary damages. See Intel Recalls Flawed 915 & 925 Chipsets above.
Here are some links describing the 915 recall:
http://www.tomshardware.com/hardnews/20040625_110502.html
http://www.dvhardware.net/article2719.html
http://www.theinquirer.net/?article=16840
And following are some links about the 820 recall:
http://www.computerwriter.com/Star/2000/jun/cw062100_intel_i820_mth_recall.htm
http://www4.tomshardware.com/blurb/20000510/
http://www.my-esm.com/digest/story/OEG20000512S0088
In July of 2004, just after the 915 & 925 Chipset recall, Intel has apparently decided that its new mobile chipset, Codename- Sonoma, which was to deliver similar advantages as the formers, but specifically to mobile computers, was also flawed. The chipsets that were already manufactured and due for release by that time were all binned, and Intel has announced that it would take a delay of at least 9 months to deliver the chipset.
Here is a sample link about that affair:
http://www.computerweekly.com/Article132052.htm
Intel has not revealed what losses it had incurred as a result of binning the existing inventory, as well as the indirect losses of promised sales. Some losses were in turn incurred by Intel's customers who had already been preparing to integrate that product.
It should be noted that such a delay in delivering this products is indicative of major design and fabrication difficulties that Intel is facing.
See also An Even Bleaker Outlook for Intel's Mobile solutions.
A particular area
that is expected to be hit even harder is the
Apart from the general Yield problem, which affects the entire PC spectrum, the Over-Heating problem is particularly hard on smaller PCs and devices.
The reason is obvious: Small devices simply cannot employ bulky, heavy and noisy cooling solutions.
This is not all, however. The larger computing power and speed that generate all that heat, consume a lot more power from the already over-stressed batteries of portable PCs and devices.
Add to that the requirement for more powerful cooling solutions, and you end up with a device that drains out the battery upon power-up.
See also Intel Fails to Deliver Long-Awaited Mobile Chipset, The Over-Heating Problem.
It has become a standard marketing habit of Intel to every now and then "draw the map from underneath the play-tools", to create new rules and hasten the rate at which existing platforms become obsolete, thus increasing incentives to upgrade, and opening-up the market.
One such move was an attempt to make all existing memory-chips redundant by employing a new type of memory that supports double transfer rates. This type of memory is called DDR2.
Unfortunately for Intel, this move has boomeranged and hit back.
The enabling support behind this move was supposed to be the brand new 900-Series chipset.
As already mentioned, this has turned into a major fiasco.
See Intel Is Hiding Flaws in Sold Chipsets.
See also Intel Recalls Flawed 915 & 925 Chipsets, Intel's Chain of Chipset Fiascos, Intel and Asus Sever Business Ties and What is a Chipset?
Despite recalling early defective 915 Chipsets, Intel has made several revisions of the product in (seemingly desperate) attempts to get it right.
It is not clear whether Intel knew at the time that all these chipsets were flawed, and has simply decided to sell them anyway to minimize revenue loss, or that it has only later found that all these chipsets were not functioning as they should, and is now too embarrassed, or too afraid to face the consequences.
When you take a look at the 915 Chipset page on Intel website, you can find long details of each 915 Chipset revision. Image 1 shows the top of (the long) 915P Chipset page.
When you follow your way down to the bottom of that page, you find some scary stuff in very small print.
Image 2 shows the bottom of that page, as seen in "
To make it look better I have switched to "Largest" font. See Image 3.
What is written there is an unclear description of defects, which Intel calls Errata (a Latin word in plural form, as opposed to Erratum, meaning- oops…).
What this Errata means in a nutshell is this: Your computer won't crash, or hang. However, it will not perform as you and we have thought that it should.
More specifically, the intended performance gain promised by the DDR2 technology is washed away.
Computer owners worldwide: If you have a 915 Chipset based motherboard, and you haven't yet purchased the memory for it, then don't waste your money on expensive DDR2. You will probably not see any performance gain.
In any case, turn to Intel for attempted compensation.
Apart from Chipsets, another major source for fiascos, embarrassment and damages has been Intel's Motherboard niche.
When one views Intel's own documentation and specifications of many of these boards, the elaborate Errata information becomes very obvious.
I can provide a first-hand experience here, associated with my very own Intel D865PERL motherboard.
See Intel's Non-Existent Customer Care and Intel Is Hiding Flaws in Sold Chipsets.
Unlike what many may believe, Intel is not about the Pentium processor (or CPU) only. For some time now, the Chipsets are becoming at least as important as the CPUs, if not more so. Intel has previously dominated the Chipset market, as well as the CPU market.
A third business niche of Intel has been its motherboard business. These boards are used by large and small integrators to create complete Computer Systems.
Intel also has a business niche in networking infrastructure.
A personal survey that I have conducted has indicated clearly that more customers, large and small alike, are walking away from the three former mentioned Intel solutions, products and components, and are looking elsewhere for cheaper and better choices.
The driving forces behind this appears to be, among others, a growing disappointment in view of Intel's poorer performing components, its abnormal number of design and manufacturing defects, its failure to deliver and live upon its declared specifications, as well as problems such as extensive heat, noise and vibrations emitted from its over-heating components.
Additionally, Intel's documentation of its products appears to imply, at times, some sort of cover-up attempts in view of the design and manufacturing defects. The documentation appears to be, at times, poor, unclear and self-contradictory (as I shall point out in examples).
Finally, a growing number of emerging manufacturers and components offer new choices to customers, which are cheaper and often outperform Intel's products and components.
See also Intel and Asus Sever Business Ties, Intel Fails to Deliver Long-Awaited Mobile Chipset and Intel Fails to Deliver DDR2 Technology.
Intel has traditionally experienced some competition in the CPU market, from its archrival- AMD. It appears that more than ever before, AMD has taken the initiative. AMD has had a gigantic leap forward in its introduction of a 64-bit CPU to the market, and more importantly- its capability to deliver it.
At the same time Intel seems to struggle in its manufacturing capabilities, seemingly being ridden by a diminishing Yield, the industry's biggest and best hidden secret. See The Secret of the Elusive Yield. This is what AMD has seemingly started to capitalize upon.
However, as mentioned earlier, the CPU niche is constantly ever-loosing its dominance in the importance hierarchy, whereas the Chipsets are ever-increasing in importance status. This is due to the fact that regardless of how fast CPUs have become, they are not the bottleneck in performance anymore. CPUs are dependant on other components, such as memory (RAM) cards, video cards, bus connection speed to other cards, hard disk drives and optical disc drives to deliver data fast, to and from them. This is where the Chipsets come in.
In fact, the Chipset are responsible for the overall performance of the computer as a whole, as well as the features it offers.
In view of a chain of fiascos involving Intel Chipsets, a vacuum has been created in the industry. This vacuum was jumped-upon by other manufacturers, much faster to respond that Intel, and very eager to nibble into Intel's diminishing dominance in this market.
See also Intel Fails to Deliver Long-Awaited Mobile Chipset and Intel Fails to Deliver DDR2 Technology.
The most notable niche that Intel was never able to master, despite some very aggressive, but ineffective attempts, has been the Video and Graphic accelerator chips. Intel simply was never up to competition with the likes of ATi and nVIDIA who now dominate the market.
On the other hand, these same vendors, who were never impressed with Intel's futile attempts to take market share away from them, have now overturned the wheel, and have been quick in a successful attempt to fill-in the vacuum caused by the ailments of Intel's Chipsets, and are now offering some very impressive Chipsets, rapidly taking away market share from Intel. These new players join some veteran vendors in this field, such as VIA.
Additionally, these new players have learned to leverage their strength in making superior Graphic Accelerators, in combining those into comprehensive, attractive and feature-rich Chipset solutions that give Intel a much more aggressive run for the money.
See also Intel and Asus Sever Business Ties, Intel Fails to Deliver Long-Awaited Mobile Chipset and Intel Fails to Deliver DDR2 Technology.
I have had the opportunity to witness at first-hand the total dismissal of any Customer-Care applications by Intel.
I have built a computer system based on the Intel D865PERL motherboard. I then discovered that Intel has issued a document which it calls "Errata" (a nice Latin word implying- "oops…") which outlines an incredibly long list of design errors associated with that board.
I applied to Intel to either provide me with a error-free revision of that board, or (generously on my behalf, I reckon) exchange the board for a newer board, for an extra charge.
Here is the essence of the communication that followed:
Me: Can you please replace my flawed board?
Intel: There's nothing wrong with that board.
Me: Yes there is. Please read your own document.
Intel: Hmm… This is the
Me (to
Intel (UK Office): There is nothing wrong with that board.
Me: Yes there is. Please review previous correspondence.
Intel (
Me: Great! The secret document reveals that in addition to the defects, I am missing a special add-on for the board.
Intel (
Me: I have lost contact with my dealer. Please assist me directly.
Intel (
Me: But this is your responsibility. There are defects and flaws in that board.
Intel (
Me: The board hasn't been broken! This is not a warranty event!
Intel (
Me (after a few days): Any news regarding my case?
Intel (
Me (after a lot more than 10 days): Hello, is there an answer yet?
Intel (
Me: Hello, is anyone there?
Intel (
Me (to both US &
Intel (both US &
One of the most tightly-kept secrets of any semiconductor manufacturer is the Yield.
Basically, the Yield is the ratio between successful microchip-product manufacturing completions and attempted microchip-product manufacturing starts. Obviously, the goal is to arrive at a Yield of 100%, which means that every single microchip progresses through the fabrication process to a successful finish in final testing.
With costs in excess of 2.5 Billion US Dollars of creating a new modern single Fabrication Facility, not to mention the astronomical operational costs of maintaining and running such a Fabrication Facility (a.k.a. Fab), it is understood that the Yield is of the utmost importance with regard to a manufacturer's financial balance.
At the early days of the industry, the Yield was in the order of magnitude of about 3% - 6%. Just imagine, for example, a Bread-Bakery, having to bin 97% of its produced loafs of bread, rather than selling them. No such bakery can possibly survive.
And this is exactly what has happened to Intel, who has started as a Memory-Chip manufacturer, only to be driven out of business by the much cheaper and more efficient Asian manufacturers.
Luckily for Intel, it has made a quantum-leap forward and has turned to developing and manufacturing processor-microchips, or CPUs.
Until recently, Intel has managed to keep its Yield much closer to the 100% limit, than was ever possible in the early days of the industry.
Unfortunately, this is no longer the case.
Intel has been traditionally boasting proving
A simple calculation shows that for the past 40 years component dimensions have shrunk by a factor of 1,000, whereas the number of components has increased by a factor of 1,000,000.
The world has been accustomed to ever-increasing computer speed and performance, and to ever-increasing memory sizes. Unfortunately, this trend cannot go on forever.
Limiting factors are the size of the tiniest individual component inside a microchip. Obviously, nothing can be smaller than an Atom (of silicon) or a larger Molecule (of the compounds used).
Even if by some miracle things were progressing smoothly
according to
But it is long before anyone even comes close to a size of an atom that problems sneak in.
As individual component sizes get smaller (much smaller than the wavelength of visible light), the difficulty in getting all tens of millions of components (transistors) of even a single microchip working, rises to such levels that it is not at all certain that a single defect-free Pentium processor, Chipset or any other product- can be delivered.
However, it's not all about size. As microchips become faster and more densely populated, they create more and more heat. That heat has to be handled, or the microchip would simply fry.
The over-heating problem and the diminishing Yield seem to have finally caught up with Intel, seriously compromising its ability to deliver its planned products.
For the sake of fairness and balance, it should be noted that everybody, not just Intel, is subject to the barrier of the rules of Physics.
However, as many analysts have noted, AMD for one, as well as new chipset makers have managed to achieve much better solutions within the current limitations.
See New competition Outperforms Intel.
A major factor that Intel has been struggling with, has been the overheating problem associated with its CPUs.
As CPUs are becoming faster and more complicated, and their micro-components multiple in number and density (so as not to increase the overall chips size), much more heat is generated inside the chip. This heat must be radiated and ventilated away, or the microchip will simply become sizzling hot and fry.
Intel, much more than AMD, seems to be facing some unexpected difficulties in driving all that heat away.
The result is a huge and bulky heat-sink that is attached to the chip, and is constantly ever-increasing in size and weight.
On top of that, there comes a cooling fan, which again, is constantly ever-increasing in size and power.
To prevent all the heat dissipating from the CPU, Chipset and other components such as cards, drives etc. from accumulating inside the computer case, increasingly powerful and noisier ventilation fans must be used to pump air into and out of the case.
The net result is that computers are becoming a source of heat and ventilation which is now beginning to rival a domestic heat radiator, accompanied by some very disturbing noises emerging from the powerful cooling fans inside the computer, and the additional vibrations that they create.
It should be noted that what is currently being marketed is the output of the R&D and production pipeline. Obviously, at the input of this pipeline there are much more advanced products, awaiting successful development and manufacturing efforts, with varying levels of success (or failure).
The operational costs and losses Intel is currently incurring are as tightly-kept a secret as the Yield.
It is difficult to estimate the financial losses in direct operational costs, as well as in lost sale opportunities and reputation damages associated with binning entire or partial production-line outputs. They may be in unimaginable orders of magnitude, in Intel's case.
As if this isn't enough, Intel may be facing some colossal damage-compensation claims resulting from the flaws and defects found in the components that it should have binned, but has sold instead. See Intel Is Hiding Flaws in Sold Chipsets.
See also The Secret of the Elusive Yield and Intel Fails to Deliver Long-Awaited Mobile Chipset.
Unlike common belief, the Chipset is now much more important than the processor in determining the overall performance of a computer.
The Chipset is a microchip (or more than one) that is responsible for interfacing the CPU with the "rest-of-the world". I.e. it connects the CPU to the main memory, to the graphics adaptor, to hard and removable-media drives, to network adaptors, to sound and audio devices, to any other "cards" and so on.
A processor, or CPU, cannot deliver any performance whatsoever, without the enabling support of its chipset.
Image 4 and Image 5 show a particular example of the Intel D865PERL motherboard, featuring a relatively "slow" Pentium Processor of 2.8 GHz speed.
Latest attempts by Intel are to market a 3.8 GHz processor, which presents a much more severe heat-dissipation problem.
See Image 6 for a view of a "modern" cooling fan.
As a Scientist, an Engineer, a Vendor, a Customer and Marketing and Business Manager, I have had a close and intimate first-hand acquaintance with the semiconductor industry in general, and with Intel- in particular, since the year 1990.
My fields of expertise include:
Hardware Engineering
Software Engineering
Signal Processing
Machine Vision & Imaging
Semiconductor Engineering, Physics & Technology
Semiconductor Manufacturing, Automation & capital Equipment
High-Tech Marketing
High-Tech Business Management
Among other publications and patents, the US Patent Office has recently approved my patent, US 6,807,022, which IMO has a significant potential in promoting the Semiconductor Technology.
This invention pertains to Semiconductor Metrology (ultra-high-precision sub-micron measurements).
My Patent description is available at the following links:
United States Patent - 6,807,022 – by Shimon Yanowitz
United States Patent - 6,807,022 – by Shimon Yanowitz – at FreshPatents.com
See also my Inventor page at freshpatents.com