Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Of course, as I write this, Intel is currently capacity constrained, so who is the Operations guy to blame!?!
Otellini, who is beloved by so many here, was the Chief Operating Officer until May 18, 2005. Intel's just released report was for the quarter that ended July 2, 2005.
Leno and Letterman are pretty impressive if I want entertainment. Screwing up a company that used to be rooted in engineering is also pretty funny if you own AMD.
"BTX's layout isn't exactly very friendly to AMD's on-die memory controller":
http://www.anandtech.com/showdoc.aspx?i=2070
JMKel, Re: How about double the revenues and a GPM of 59.5%?
That would be very very nice. Except Intel's Q3 2003 revenue was $7.8B. The same year-over-year 18% increase as Q2 (Q1 was 20% year-over-year) would put Q3 2004 revenue at $9.2B. But Intel just said: "between $8.6 billion and $9.2 billion"
Duke: Depreciation has started on all those new plants as they are put into service. This will lower paper profits but look at the cash mounting up.
Actually, it seems depreciation has been flat. It was about $4.6B for both 2002 and 2003. For 2004:
http://www.intel.com/pressroom/archive/releases/20040713corp.htm
"Depreciation is expected to be between $1.1 billion and $1.2 billion in the third quarter and approximately $4.6 billion for the year."
Sure, free cash flow has been way higher because of much lowered capital expenditure.
Meanwhile, I just don't understand the inventory issue. Is Intel having excess capacity? What was explained during the conference call didn't quite match those in the report: "slight reduction in microprocessor average selling prices and a slower than expected reduction in microprocessor unit costs" (lower ASP, higher production cost -- terrible!).
Duke: Intel now controls the motherboard market, for the first time in a long time. What is a lousy 38 Mill?
I would be great if you could share data relating to this. In recent years, I always assumed the leaders are companies like Asus. I am not sure if Intel even makes motherboards in Intel's own factories (?).
Duke: NOW!, the board says, we need a guy like Otellini to sell this stuff.
I just have problems trusting Otellini. For example, is he really so talented that, as the Intel's CEO-to-be, has the time to also be a director at Google (or is Intel not giving him enough options already?):
http://sap.ittoolbox.com/news/dispnews.asp?i=116829&t=99
Wbmw, Re: Recalls are expensive. Besides the parts that Intel has to replace for their customers, they also have to junk anything they already made that was sitting in inventory.
But it was very early and there shouldn't have been too many units:
http://www.investorshub.com/boards/read_msg.asp?message_id=3428902
Re: You are referring to Paul Otellini, but you are making an incorrect connection. Otellini is the one responsible for the "Right hand turn" to lower power processors (Centrino), not "Netburst". Albert Yu advocated Netburst, just like he did with Rambus, and that's why he was exiled to technical Siberia (also known as Opto-electronics).
I think Rambus, had it succeed (it had its advantages, especially pre-dual-DDR) and became wide spread, would have allowed Intel to control the platform (given the special arrangements between INTC and Rambus). AMD had nothing related to Rambus that I can remember. It was the right business strategy, but not well executed (not with product development or the technology, but marketing the idea to the industry).
i820 worked with Rambus just fine. The problem was with having to support SDRAM "the last minute".
P4 was severely delayed. Part of it was caused by Intel having to spread resources around all in sudden (Celeron, Xeon, etc).
I am sorry I am not aware of Otellini's involvement. I always assumed Grove had something to do with that team being in Israel. Timna had Rambus.
Re: I think Otellini's vision is good for Intel. Going towards a marketing direction is not bad when you are going up against the commoditization of PC processors, which is Intel's bread and butter. Intel needs more ideas like Centrino to take margins up to the 60% and above range, and I think Otellini is the right guy.
I don't know... If PC processors really become a commodity, I wouldn't even consider INTC at all. I am not aware any company really having a competitive advantage by relying on marketing to sell a commodity.
Even Coke, which is fairly close to being a purely marketing company, offers a taste that some people prefer over Pepsi.
I started holding INTC since 1997. For the first time I am seriously considering getting out. The only thing positive from the report is "The company expects faster growth in products such as flash memories, chipsets and motherboards that have lower margins." Yeah, more revenue, but LOWER MARGIN products.
Even the chipset recall, which I didn't think would even cost $8M, turned out to cost $38M. Why?
This is extremely worrisome. Please don't take this personally people. But I have been extremely bothered by the kind of investors that INTC has recently attracted. For example, following a post "IA-64 was not meant just for server/workstation", someone would then rebut with "Itanium was meant for server/workstation". And the use of empty words like "officially". They just all sound like playing with words.
I was at first concerned about something like "NetBurst" marketing b.s. The fact that a marketing type is going to soon become CEO for the first time in Intel history is probably going to the last straw for me.
What happened to "Be open and direct"?
http://www.intel.com/jobs/workplace/values.htm?iid=jobshomenav+values&
I hope Grove comes back and clean house. I think Barrett made a serious mistake choosing the marketing type over Albert Yu.
I don't know if we simply have a miscommunication of terms. I refer to IA-64 as the architecture. Itanium is one version of IA-64.
Itanium is a brand name. It WAS and IS targeted to the server and workstation market. There was no "Itanium" until the name was announced in 10/1999:
http://www.intel.com/pressroom/archive/releases/sp100499.htm
So yes, "Itanium was originally intended to compete with Sun and IBM". But I personally don't believe IA-64 was originally intended to only "compete with Sun and IBM". IA-64 began to take shape in 1994.
More about Yu's involvement here. I just want to point out that at the time, I would guess the biggest microprocessor threat in the server/workstation market was actually from Digital. Alpha was already 64-bit.
http://zdnet.com.com/2100-11-529889.html
So Albert Yu, who was involved with IA-64 from the beginning, actually wrote about planning for the future. I don't know how to respond to "You are stretching".
By the way, Yu was GM of the Intel Architecture Group (I think that sounds a bit different from "worked in the desktop group"):
http://www.intel.com/pressroom/archive/releases/20020627corp.htm
Respectable companies simply don't "officially" announce something many years into the future. When did Microsoft "officially" announce that Win32 (as in NT) would be on every desktop?
http://www.microsoft.com/presspass/features/1998/winntfs.asp
It would be more interesting if you could post a link showing Intel has officially stated that IA-64 will NEVER go into desktop.
No, that article from Microprocessor Report was a prediction. He was NOT reporting those as facts.
I have already tried to find what Grove apparently has said (again, sorry I couldn't find direct quotes). Albert Yu was working right under Grove. He was still with Intel when he wrote "Creating the Digital Future":
http://print.google.com/print/doc?isbn=0684839881
"PLAN THE COMPUTING FUTURE
"These 64-bit microprocessors represent the next giant leap in performance and capability that will allow the microprocessors to power the full spectrum of computing from its current base of personal computers to mainframes, supercomputers, and beyond.
Yu didn't use the word "desktop", but I don't know what else "its current base of personal computer" could mean.
Please think about it (and just read the first paragraph of the excerpt). I am running out of facts to present :)
Again, that was from the 11/10/97 issue of Fortune. I am sorry I couldn't find a direct quote. But here it is again (1999 + 5 years = 2004):
http://www.cnn.com/TECH/computing/9807/15/mystique.idg/
"Eventually, Intel will filter 64-bit processors down to desktop computers, but not for at least four or five years after Merced comes out, Intel Chairman Andy Grove has said.
As to "2004" and predictions in general, I thought this was interesting. I came across this (Microprocessor Report, 6/20/94) as I was trying find a direct quote from my archive:
"Setting the stage for new architecture battles at the end of the decade, Intel and Hewlett-Packard have announced a partnership to develop a next-generation CPU architecture that will eventually replace the current x86 and PA-RISC architectures...
"With this announcement, Intel for the first time has admitted that the x86 architecture will run out of steam around the end of the decade. At the same time, the company revealed its plan to solve this problem by switching to a new post-RISC architecture...
...
"We expect that, in about 10 years, Intel will stop making pure x86 chips in favor of P86 chips... Intel's P6, due in late 1995, probably will be the last pure x86 core that Intel develops...
C'mon, we should feel very disappointed if the architects at Intel hadn't thought about designing a processor (to completely replace x86) from scratch and shrewd business people like Grove hadn't thought about something like this:
http://www.investorshub.com/boards/read_msg.asp?message_id=3332497
It's no different from software people keeping thinking about rewriting legacy code from scratch and Buffett saying this (Fortune, 11/22/99):
"The key to investing is not assessing how much an industry is going to affect society, or how much it will grow, but rather determining the competitive advantage of any given company and, above all, the durability of that advantage. The products or services that have wide, sustainable moats around them are the ones that deliver rewards to investors.
Well, *if* Intel projects something lower than 62%, something must have gone really wrong in just the last 2 months.
Based on Q1 (4/13/04), I suggested "it was as though Intel also said that GM in Q3 and Q4 are expected to be higher than 62":
http://www.investorshub.com/boards/read_msg.asp?message_id=2880258
Within Expectations on June 3, 2004:
http://www.intel.com/pressroom/archive/releases/20040603corp.htm
GM was below 50% in 2001 and 2002. I assume you were still holding, but just concerned? :)
http://www.investorshub.com/boards/read_msg.asp?message_id=754233
I don't think Itanium was ever aimed at what you think it was. I believe it was aimed at the 10% of the server market that Intel had no entry in.
I am not sure if you are playing with words. True, "Itanium" as a brand was aimed at the server market. But IA64 was aimed at, well:
Eventually Merced will begin to displace Pentiums in mainstream desktop and laptop PCs. Grove doesn't expect wholesale replacement until 2004, but Merced gives Intel the luxury of thinking that far ahead -- and beyond. AMD's Dham is only half-joking when he says: "Merced will ensure Intel's dominance of the microprocessor business for the next 50 years."
That was from November 10, 1997 issue of Fortune. My take from what Grove said would be that IA64 was aimed at 100% of the microprocessor market :)
And even if true:
http://biz.yahoo.com/rc/040629/tech_amd_dell_1.html
Gateway used to be Intel-only. Then Gateway started to sell AMD:
http://www.theregister.co.uk/1999/12/01/gateway_to_start_shipping_athlon/
GTW closed at $74.75 on 12/1/1999. Dell ought to look at what has happened:
http://finance.yahoo.com/q/bc?s=GTW&t=5y
Of course, there have been other factors. But I think there is certain mentality that cause people to go with AMD.
By the way, I am not sure if this "Tai Nguyen" character even used the term "channel check" correctly. "Financial health" (such as how well *final products* are doing) versus a rumor of something in the design stage?
http://www.investopedia.com/terms/c/channelcheck.asp
OK, I will try not to be so terse in the future cause I am NOT on drugs! :)
The old rumor was that Tejas was designed for desktop and Jayhawk for server:
http://www.xbitlabs.com/news/cpu/display/20040507000306.html
When Tejas was canceled (along with Jayhawk, but it was less often mentioned), the rumor was that "all chips are to feature Pentium M-like architecture":
http://www.xbitlabs.com/news/cpu/display/20040507095407.html
So it was EXCELLENT news that the 2 GHz Dothan performed so well. Hey, if one is so good, having two Dothan cores is going to be even better :)
http://www.xbitlabs.com/news/cpu/display/20040427080257.html
I am not sure why Prescott has such as bad name. I keep thinking it is very good -- I never would have imagined a chip having 30 pipeline stages (Celeron D) outperforming one with 20 (Northwood Celeron):
http://www.xbitlabs.com/articles/cpu/display/celeron-d_15.html
Seriously, I am not on drugs. Are you having a bad day?
Sorry if it sounds like gibberish. But people have been talking about "Tejas fiasco", etc. I have been saying it's because P-M is so good:
http://www.investorshub.com/boards/read_msg.asp?message_id=3435846
http://www.investorshub.com/boards/read_msg.asp?message_id=3042187
So given the Dothan performance, I shouldn't have suggested looking forward to Jonah?
Yes, please allow me to repeat myself. There is no frequency scaling problem and there is no "Tejas fiasco".
It’s simply because Dothan is so impressive.
Now let’s imagine when Jonah hits the market!
Well, I was being sarcastic. :)
More recent events give people more vivid memory. I thought those "screw-ups" of the past were much worse!
Unlike 386/486SL, Timna never became a real product, and we got Dothan and future across-the-board P-M out of it.
I am not sure if there has been anything remotely comparable to FDIV.
I also had a different view on Tejas:
http://www.investorshub.com/boards/read_msg.asp?message_id=3042187
Intel was going through iterations of chipsets when Intel entered this market. The most recent recall was discovered early, whereas i820 caused a $200 million reserve.
I think the real screw-ups would be something like why did Intel allow Microsoft to create an OS.
From Warren Buffett: "In the short run, the market's a voting machine, and sometimes people vote very unintelligently. In the long run, it's a weighing machine, and the weight of business and how it does is what affects values over time."
I know, Intel really needs to learn from prior crew-ups!
> Timna cancelation
Yeah, why didn't Intel learn from 386/486SL.
http://ascii24.com/news/columns/10107/article/2001/12/28/print/632480.html
(Try http://babelfish.altavista.com/babelfish/)
> 1.13GHz Pentium recall
Why didn't Intel learn from the Pentium. P-60 was huge, ran hot, and full of errata’s. Wait, then there was FDIV. At least I got a free key chain out of it!
http://www.redhill.net.au/c-3.html
Then Intel had the audacity to do the same with the PPro!
http://www.redhill.net.au/c-4.html
> Rambus
Can anyone think of something similar in early '90?
> Tejas fiasco
Intel bet half the farm on RISC, and we only got i960 out of it. Why didn't Intel learn from that!
http://www.businessweek.com/1989-94/pre88/b30711.htm
From "Inside Intel" page 346: "BiiN ... was commonly referred to as Billions Invested In Nothing".
> now Grantsdale
I know, those chipsets keep having bugs:
http://www.os2forum.or.at/english/info/os2hardwareinfo/pci_chips.html
You'd expect Intel to have learned from what Intel had to do for the i820:
http://www.pctechguide.com/27chipsets.htm
Sigh, at least Grove admitted a couple real mistakes:
http://www1.people.com.cn/GB/channel5/28/20000823/199237.html
(Try http://babelfish.altavista.com/babelfish/ Chinese-simp to English)
Imagine losing MSFT's market cap!
Thanks for the eweek article.
"Intel has discovered a glitch involving some of the I/O controllers"
So it's just the ICH6.
"The product bug has apparently caused Intel to recall a portion (approx 25%) of the several hundred thousand chipsets shipped thus far," Root added. "This suggests a total magnitude for the recall of 100,000− 200,000 chipsets, or less than $8 million of product on a Q2 revenue base of $8.1 billion."
Intel sells the whole chipset for at least $37. So the ICH cannot be worth more than that, and even less to make. So even the $8 M estimate is probably way too high. I bet Intel lost many many times that in market cap when the market saw the news headline.
http://www.intel.com/pressroom/archive/releases/20040621comp.htm
"Intel's own−brand motherboards did not appear to have this issue."
This is curious. I thought Intel ships chipset and has companies like Foxconn make the Intel branded motherboards. So why should it be different than, say, Foxconn making its own branded motherboards.
the important thing for Itanium to do is kill off sparc and POWER. If it turns out Xeon and opteron do that trick, fine.
I think it's inevitable that China will spend lots of money (while forgetting return on investment) building fabs. Unlike Taiwan or Korea, China has far greater basic research resources. The equipment suppliers will also offer greater levels of services and integration.
So in addition to fab capacity and leading edge R&D, I would still like to see Intel pushing IA64 throughout the computing platforms. Intel owns 85% of the X86 market and 100% of the IA64 market. The quicker we shift from X86 to IA64 the better!
It's already been overclocked to 2.40 GHz:
http://forums.anandtech.com/messageview.cfm?catid=51&threadid=1307342&enterthread=y
Thermal Guideline for 1.7 GHz, 1.8 GHz, and 2 GHz all say 21.0W. That's even lower than the old 1.3 GHz part at 22.0W!
http://processorfinder.intel.com/scripts/default.asp
Let me try posting the URL again:
http://www.realworldtech.com/forums/index.cfm?action=detail&PostNum=2297&Thread=1&entryI...
Now, when IBM puts an "Athlon 64 at ~1ghz (under power now) is 10 to 12 watts" (where is the data sheet anyway?) into a Txx, I may consider buying it :)
http://forum.thinkpads.com/viewtopic.php?t=141
The slide is in Chinese and *looks* to be from Intel. I will try to do a quick translation for those that cannot wait. Because the source cannot be confirmed, treat this as *pure speculation* and *pure rumor*. Also, I never signed anything and cannot be responsible for anything. :)
http://www.realworldtech.com/forums/index.cfm?action=detail&PostNum=2297&Thread=1&entryI....
Page 2
Dothan: Next generation Intel Pentium M
- 90 nm, strained silicon
- 0.14 billion transistors
- 2MB power saving cache
- standard voltage, low voltage, ultra low voltage
- WHOLE NEW micro architecture enhancements: improved register data visit (?) and data prefetch
Intel PRO/Wireless2200BG
- 802.11b/g
Page 3
Page 4
2H 2004: Alviso chipset, 802.11 a/b/g, Sonoma?
Page 5
0.14 billion transistors is almost twice previous generation
Improved register management (?) supports mixed length read/write
Improved prefetch loads data that CPU may use into L2 cache
Page 6
90 nm benefit: more efficient, lighter, better fan-less design with ultra low voltage
Strained silicon benefit: better frequency scaling
Page 7
SPECint_base2000 (standardized to P4-M 2.6 GHz)
1.00 P4-M 2.6 GHz
1.30 Banias 1.7 GHz
1.45 Dothan 1.7 GHz
SPECfp_base2000
1.00 P4-M 2.6 GHz
1.17 Banias 1.7 GHz
1.31 Dothan 1.7 GHz
Performance higher than Banias by 11-12%
Performance higher than P4-M 2.6 GHz by 31-45%
Page 8
Dothan Banias
Standard Voltage 21W 24.5W
Low Voltage 10W 12W
Ultra Low Voltage 5W 7W
Dothan Banias
Standard Voltage 1.31V 1.48V
Low Voltage 1.05V 1.18V
Ultra Low Voltage 0.89V 1.0V
Page 9
Through 5 walls:
Intel 802.11g throughput 19.9 Mbps
Competitor 1 throughput 10.8
Competitor 2 throughput 13.1
At 50 M:
Intel 802.11g throughput 20.7 Mbps
Competitor 1 throughput 15.6
Competitor 2 throughput 10.8
borusa, we should know more about 90nm Dothan on Monday:
http://www.theregister.co.uk/2004/05/06/intel_dothan_debut/
mas, I am not a gamer :) More interested in a cool, stable machine. Don't know if I deserve an IBM T41 (or T42) though:
http://fatwallet.com/forums/messageview.php?start=23&catid=18&threadid=299734
After the exercises, I wonder if there is a correlation that the more reliable vendors tend not to use AMD :)
http://www.pcmag.com/article2/0,1759,1186140,00.asp
http://www.pcmag.com/article2/0,1759,1186135,00.asp
Thanks mas again.
So 35W and 62W are for different Mobile Athlon 64 processors.
P4-M now goes from 21W to 35W:
SL6P2 2.50 GHz 400 MHz 0.13 micron C1 512 KB 35.0W
SL5ZW 1.40 GHz 400 MHz 0.13 micron B0 512 KB 25.8W/20.8W
SL6F5 1.40 GHz 400 MHz 0.13 micron B1 1 MB 22.0W
P-M now goes from 7W to 24.5W:
SL6N9 1.70 GHz 400 MHz 0.13 micron B1 1 MB 24.5W
SL6NC 1.10 GHz 400 MHz 0.13 micron B1 1 MB 12W
SL6NJ 900.00 MHz 400 MHz 0.13 micron B1 1 MB 7W
I am pretty sure I won't buy an AMD "mobile" processor based laptop with these numbers.
Does anyone know how the "model name" will work? With 2 digits, it's just a matter of time that we run out of model names even if they are each less than 5 apart.
For example, what's the model name higher than 395? Do we go to 3 digits while keeping the first digit?
But 3005 (a Celeron that’s better than 395) is much much greater than 595 (a fast P4 that's most likely richer than the 3005 Celeron).
Another interesting thing to note from the article is that IBM now says frequency scaling is "dead". This is very different from Intel Research's position.
This doesn't leave much for AMD. Obviously AMD cannot rely on IBM. So AMD cannot rely on frequency scaling. But AMD doesn't have a lower-power core either.
Thanks, mas.
So Mobile Athlon 64 consumes 62W (or as low as 35W -- cannot quite understand that X-bit paragraph).
Then "Intel’s Pentium M processors specifically tailored for notebooks consume up to 24.5W at 1.70GHz. Intel’s Pentium 4-M processors at 2.40GHz and 2.50GHz typically consume 30W and 35W respectively under maximum load."
When it's time to jump on the "get two for the price of one" bandwagon, which lower-power core is AMD going to use? The K6 isn't even competitive with P3.
We heard about "TeraHertz transistor" as early as 11/2001:
http://www.intel.com/pressroom/archive/releases/20011126tech.htm
"Our goal is to overcome these barriers and produce chips that have 25 times the number of transistors of today's microprocessors at ten times the speed with no increase in power consumption."
"Power consumption as a limiting factor"
"Intel is expected to begin incorporating elements of this new structure into its product line as early as 2005."
Things looked even better in 11/2003:
http://www.intel.com/pressroom/archive/releases/20031105tech.htm
"Intel Corporation today announced it has identified new materials to replace those that have been used to manufacture chips for more than 30 years."
"Intel said the new high-k material reduces leakage by more than 100 times over the silicon dioxide used for the past three decades."
"Transistors with these new materials are an option targeted to be integrated into future Intel processors as early as 2007, as part of the company's 45-nm manufacturing process."
At first, I think Prescott was engineered with frequency scaling in mind. Prescott was very well engineered -- same IPC even with 50% longer pipeline. The first iteration even at 103W is tolerable. But by 2005, some elements of "TeraHertz transistor" for frequency scaling need to be there especially for Tejas per the 11/2001 announcement. So is there really going to be a frequency scaling problem?
Let's say there is. Then the problem could from engineering failing to replicate research results, bean counters failing to buy new equipment, or marketing killing something because they cannot think of a way to market it. Intel has always overcome engineering problems. But Intel has also always been run by Ph.D.'s. So the later two would be more worrisome. In fact, I seem to remember Otellini saying that a feature wouldn't be put in unless it can be marketed. (And to this day, I still cannot figure out how "net" or "burst" has anything to do with Hyper-Pipeline.)
Or there actually isn't going to be a frequency scaling problem. This means Jonah really is going to be a better solution. Or, at least it will be easier for marketing to tell people "Buy Jonah, you are getting two processors for the price of one". This is great news to shareholders. In the old days Intel just had two design teams. Now the best ideas can be shared among and the best result be picked from at least four design teams (California, Oregon, Israel, India, Texas?, Arizona?).
(OK, I am bad mouthing marketing too much. My guess is that it's a combination of several things, including Tejas is late. But I really don't think frequency scaling is the main cause.)
What's more curious is the IA-64 and IA-32 merge plan. My impression has been that Prescott has 64-bit X86 support that can be turned on if necessary. But the real stuff is that Tejas will have IA-64 support. Now Tejas is no more, how/when is a X86 processor going to be able to start executing IA-64 code?
I wonder if shifting to P-M architecture will give Intel additional competitive advantage? It seems Intel is moving to lower-power dual-core as early as next year. Yet K9, which is two years away, will still be single core?
http://freespace.virgin.net/m.warner/Roadmap20XX.htm
I looked up the thermal guideline of a few sSpec's:
http://processorfinder.intel.com/scripts/default.asp
P3
SL5FQ 1.00 GHz 133 MHz 0.18 micron cD0 256 KB 29.0W
SL64W 1.40 GHz 133 MHz 0.13 micron tA1 256 KB 31.2W
P4
SL62Z 1.70 GHz 400 MHz 0.18 micron D0 256 KB 67.7W
SL6WT 2.80 GHz 800 MHz 0.13 micron D1 512 KB 69.7W
SL79K 2.80E GHz 800 MHz 90 nm C0 1 MB 89.0W
SL7B9 3.40E GHz 800 MHz 90 nm C0 1 MB 103.0W
P-M
SL6F8 1.40 GHz 400 MHz 0.13 micron B1 1 MB 22.0W
SL6N9 1.70 GHz 400 MHz 0.13 micron B1 1 MB 24.5W
I couldn't find equivalent AMD information except the following:
http://www4.tomshardware.com/cpu/20040419/images/cpu_history_big.gif
http://www.amd.com/us-en/assets/content_type/DownloadableAssets/K8_Electrical_Spec_Rev_ENG.pdf
http://www.amd.com/gb-uk/assets/content_type/DownloadableAssets/K7_Electrical_Specification_Rev_ENG....
Athlon XP 2800+ 68.3W
Athlon 64 89W
Mobile Athlon ?
In August 2002 I wrote the following letter to Andy Grove, Craig Barrett, Andy Bryant, and Intel investor relations. Except for Craig Barrett, they all replied (I was surprised!). In different ways, they all said expensing stock options was bad accounting (I agree) and they were working on providing better information (I agree they have done).
In contrast, Applied Materials investor relations didn't reply to their version the same letter. My e-mail to James Morgan failed delivery.
My personal take is that because stock options is so hard to evaluate, just don't do it. If your employees have any brain, they will use their salary to buy (or not buy) company stock on their own. But I just voted with the Intel board recommendations because I don't think I am nearly as capable as they are.
---
My respect for Intel's conservative accounting and honest management is a major reason for my being a long-term Intel shareholder. I have become more and more concerned that Intel may be lumped with other possibly less respectable companies that do not expense stock options.
I am concerned that the investment public, in order to account for the lack of information from individual companies, may eventually apply a "fixed discount factor" when viewing ALL companies that do not expense stock options. I fear that this "fixed discount factor" may become quite punishing because expensing stock options can be very detrimental to many other companies; some of these other companies may possibly even have something to hide.
Because Intel is a highly respectable company, it simply won't be fair for Intel to be lumped with "companies that do not expense stock options". As the list of companies that do expense stock options grows, I fear that if Intel does not start to report a (much less severe) discount factor on our own, Intel's share price will be punished unjustly.
I ask Intel's management to do everything possible so the investment public will NOT apply this "fixed discount factor" to Intel. Meanwhile, I do understand that expensing stock options has its disadvantages. So it seems that a happy medium can be reached if, in addition to the current set of numbers, Intel simply starts to report another set of numbers with stock options expensed.
On 4/13/04, Intel said "Gross margin percentage in the second quarter is expected to be approximately 60 percent, plus or minus a couple of points" and "The gross margin percentage expectation for 2004 is unchanged at approximately 62 percent, plus or minus a few points"
Because GM was 60 in Q1, it was as though Intel also said that GM in Q3 and Q4 are expected to be higher than 62.
Q4/02 Actual 52
Q1/03 Expected 50, Actual 52
Q2/03 Expected 50, Actual 51
Q3/03 Expected 54, Actual 58
Q4/03 Expected 60, Actual 64
Q1/04 Expected 60, Actual 60
Q2/04 Expected 60
Tenchu,
From outside looking in, I thought Willamette was originally meant to be released in 1998 (or even 1997 with the "two-year cycle"). Instead, Intel had to introduce Katmai, which was very short lived. PII and Katmai were costly to make because of the module and external cache (which could also be explained as a capacity issue). Coppermine finally saved the day with respect to cost. From an outside perspective, because Coppermine came so shortly after Katmai, I actually thought Coppermine was, if anything, done right.
So perhaps I was wrong because someone like you with an internal view think differently. Perhaps you meant that Katmai should never have existed but Coppermine should have been introduced instead? If so, Intel would still have the same problems (not meeting demand and the recall) in 2000 -- the real problem was still with Willamette being so late...
Andy
I think it was around the first part of 2000. You mentioned two somewhat related issues:
1. Coppermine manufacturing difficulties
2. market was booming
I think the real problem was that Willamette was too late and too big. Coppermine/PIII uArch wasn't meant to be carried that far. So the "manufacturing difficulties" was caused by trying to make something that wasn't supposed to be made at that time.
Yes, the market was booming, and Intel did leave money on the table. But that was exactly the wise thing to do! To counter the semi cycle, and if you can afford to, you exactly want to under-invest during boom times but turn up spending when others are just trying to survive.
When Intel made the announcement on 9/21/2000 (the stock dropped 30% over the following week or so), it was a blessing that Intel didn't have as much excess and soon to be outdated capacity. The stock would most certainly have dropped even more if Intel had the extra capacity to satisfy the peak demand. Intel indeed spent heavily the last few years. And with Prescott and follow-ons, the real payoff in the x86 business is just about to begin.
By the way, I am still reading people talking about Intel should do this or that about 64-bit. Don't these AMD cheerleaders (not Tenchu) realize that Intel must have studied all sort of options ten (yes, 10) years ago (when AMD was perhaps worrying over the internal K5 schedule and warying of NexGen?)? The top managers and architects decided to go after IA64 and I know I am not nearly as competent to judge them. Part of the Willamette problem must have had been caused by allocating resources to IA64.
I would quote this from Anand:
Given that we’re at the very beginning of the 90nm ramp and we are already within reach of 4GHz, it isn’t too far fetched that Prescott will reach 5GHz if necessary next year. From an architecture perspective, it is impressive that Prescott remains in the same performance league as Northwood despite the fact that it has a 55% longer pipeline.
Perhaps I am an optimist. But I just want to put this so called "f*&k up" into perspective. Not too long ago:
- Willamette was late enough that PIII couldn't keep up with AMD's offerings.
- Willamette was much slower in general than PIII at the same frequency.
- Intel under-invested in equipment relative to others (because others had had all the money to over-build during boom time).
Things eventually worked out quite well, as we all saw the market share gains during year 2003. Intel (certainly already had all kinds of Prescott data) just said on 1/14/04 that "Gross margin percentage for 2004 is expected to be approximately 62 percent, plus or minus a few points, as compared to 57 percent in 2003." Here are some reasons why 2004 will be even better than 2003:
- Northwood is still doing just fine relative to AMD's offerings.
- Prescott is NOT slower than Northwood at the same frequency.
- Intel now has the equipment (because others didn't have the money so they under-built).
I would say that going from 20 to 31 pipeline stages while maintaining IPC (on non-recompiled code) is very very impressive. AMD went from 10 to 12 pipeline stages and it sounded like a big deal (I am sure it was). I'd love to know AMD architects' reaction when they first heard of "31 pipeline stages". The thought of how much work they need to do just to keep up with Intel's frequency scaling must be very depressing.
Over two years ago, Grove said one third from each of three spaces: client, network, and server. Around the same time, Barrett said networking and communications to account for 50%. Assuming that the client space to be 50% PC and 50% handsets, then Grove and Barrett were consistent with each other.
I once read that Grove and/or Barrett were known to show big goals that a subordinate had signed up for but couldn't quite deliver. With a smirk, he would ask the poor fellow, so what happened? (Or something like it.)
I hope I won't have to show the following news clips to Grove and Barrett in 2010. Anyone else interested in making some projection as to INTC's revenue based on their big goals?
---
Intel Co-Founder Grove on Semiconductor Industry: Comment
3/6/1 17:34 (New York)
Santa Clara, California, March 6 (Bloomberg) -- Intel Corp. Chairman and co-founder Andy Grove comments on the state of the semiconductor industry during an interview with Lehman Brothers Inc. analyst Dan Niles broadcast over the Internet.
Grove told investors to ``keep the faith'' and said that five to 10 years from now, Intel will get about a third of its sales from each of three markets: personal computers and handheld devices, servers that run Web sites and databases, and networking gear that connect the two groups.
On problems predicting the economic slowdown in the U.S.:
``The viciousness of the down cycle was made more so by all the (supply-chain management software.) It blew through the supply chain in a much faster ripple than previous cycles. Nothing in supply-chain management can read minds. End demand is what end demand is.''
On when demand could recover:
``I don't think end demand is going to snap back. For a number of years, technology had a huge momentum. Over some period of time, the investment cycle got ahead of itself. When demand softened some, it exposed the fact that we have built out excess capacity in the whole system. That leads to a slow recovery. I don't think end demand is going to snap back. It is going to ease back in a gradual fashion.''
On whether the U.S. slowdown will spread to Asia:
``The Asian market (decline in the late 1990s) didn't affect the U.S., but the U.S. market is going to affect the Asian market. It is not possible for the U.S. and Europe to have a softer end demand cycle without it affecting some of the Asian export activities.''
--Cesca Antonelli in the San Francisco newsroom (415) 743-3532, or at fantonelli@bloomberg.net
---
Intel's Grove doesn't see rapid snap-back in demand
(UPDATE: Adds analyst comments, background)
By Duncan Martell
SAN FRANCISCO, March 6 (Reuters) - Intel Corp.'s (NasdaqNM:INTC - news) Chairman Andrew Grove said on Tuesday he does not expect demand for semiconductors to recover quickly, noting it will take some time for the downturn in the chip industry to end.
``I don't expect the end demand to snap back,'' Grove said on a conference call and Webcast hosted by Lehman Bros. and its semiconductor and PC-company analyst Dan Niles. ``We are in this state for some period of time.''
Grove declined to be more specific.
The semiconductor industry has undergone a vicious reversal of fortune since the fall of 2000, with virtually every type of semiconductor company warning that first and even second- quarter sales in 2001 will be less than forecast. But most are holding out hope the second half of the year will bring more robust growth in the U.S. economy.
NOT OUT OF THE WOODS YET
But even while the Philadelphia Semiconductor Index gained 5.5 percent on Tuesday, its fourth straight gain, Grove's comments seemed to suggest that the chip industry isn't out of the woods yet.
Indeed, on Monday a semiconductor trade group reported that worldwide sales of computer chips fell 5.7 percent in January from December, the third consecutive monthly decline.
``What I think Dr. Grove's comments speak to is, don't let the stock prices dictate what the fundamentals are, it's actually the other way around,'' Niles said. He added that he has been fielding frantic calls from clients this week who wonder if they've missed the ``bottom'' in the chip cycle.
``There are just too many people not paying attention fundamentally to what's going on,'' Niles said. ``Semiconductor stocks are up for our days in a row and I could care less.''
While the short term may be rough sledding, Grove, an American business icon and Time magazine's Man of the Year, remains bullish about the long-term future of the high- technology industry.
'FUNDAMENTAL INDUSTRY OF OUR TIMES'
An Intel spokesman said, in response to Grove's comments, that the remarks did not represent any change to the Santa Clara, Calif.-based company's business outlook for the current second quarter.
``This industry is the fundamental industry of our times -- not just chips, not just Intel, but the high-technology industry,'' Grove said. ``For those of you that have a stake in there, I'd like to share my confidence and enthusiasm for this industry long-term. So keep the faith.''
Grove also defended the 33-year-old chipmaker's plans to spend $7.5 billion this year on capital spending, more than double the amount it spent just two years ago. He cited a downturn in the chip industry in the 1970s when Intel pared back on capital spending, but Japanese chipmakers did not.
When demand did snap back, Intel did not have the capacity in place to build enough memory chips to meet demand, Grove said. Intel later partially exited the dynamic random-access memory, or DRAM, business as a result.
``My biggest regrets were always that we were too conservative on capital,'' Grove said on the conference call. ``As a result we weren't ready to follow the demand curve when it snapped back.''
Intel and the chip industry are undergoing one of the biggest changes in decades. Not only are chipmakers such as Intel moving to new process technology, called 0.13 micron, they are also simultaneously moving to dinner-plate-sized silicon wafers from salad-plate-sized ones. These changes help to cut costs and boost production markedly.
``We are in the middle of a very important technology cycle,'' Grove said. ``The companies who cut cut back on investment now, when the recovery comes, will be stuck with 8-inch capacity ... compared to other players who moved to 12 inch.''
TOO MUCH SPENDING ON TECHNOLOGY?
In addition, Grove, born in 1936 in Hungary and who joined Intel in July 1968 when it was founded, said that in the past few years there may have actually been too much spending on information technology. That runs somewhat contrary to the thinking that there can never be enough bandwidth.
``For a number of years, technology had a huge momentum -- technology buying and manufacturing had a tremendous investment cycle going,'' Grove said. ``I think people loaded up with not just physical inventory but got ahead of themselves in capacity building and network capacity building.''
``We built in an overcapacity of all physical things,'' Grove said. ``Maybe demand only changed in a relatively minor way but it exposed we had built up an excess capacity in the total system.''
To eliminate this overcapacity, Grove said that demand has to come up or investment levels have to come down until ``the two levels cross, or the investments of the past several years have become obsoleted'' by new technology.
``Both of these require some time,'' Grove said. ``If you accept that that's what's happening in the economy all together, that leads to a slow recovery.''
---
Tuesday March 20 7:01 PM ET
Intel CEO Still Hopes for Second Half PC Rebound
By Yukari Iwatani
LAS VEGAS (Reuters) - Craig Barrett, chief executive of No. 1 chip maker Intel Corp., said on Tuesday he continues to hope for a recovery in demand for personal computers in the second half of 2001, despite the current major slump in the industry.
The European market is ``okay'' and, as expected, the United States was the company's ``primary weak point,'' Barrett told Reuters in an interview at the Cellular Telecommunications Industry Association conference here.
Barrett's comments come after Intel (NasdaqNM:INTC - news) issued its third consecutive sales warning earlier this month saying that first-quarter sales would fall as much as 25 percent to about $6.53 billion from the previous quarter as an economic slowdown spread beyond personal computers to networking, communications and server components.
``We're still hoping for a recovery in the second half,'' Barrett said, who also added that the Asian market should be strong but didn't provide specifics.
Some analysts have gone so far as to argue that the U.S. technology industry has entered a recession and that the PC market would indeed shrink from a year ago in 2001 for the first time in its history.
US Bancorp Piper Jaffray analyst Ashok Kumar said in a note to clients on Monday that the second half of 2001 will in fact not be a period of recovery and should ``result in negative unit growth for the PC market in 2001, a first in its history.''
``We do not believe that there will be a sustainable cyclical recovery until'' the second half of 2002, Kumar wrote.
Also, Salomon Smith Barney analyst Richard Gardner wrote in a note to clients on Tuesday that the ``most recent U.S. PC retail sales data indicates ongoing weakness in the consumer market'' for the week of Feb. 25 through March 3, with growth declining a steep 21 percent.
Banking On A Rebound, New Businesses
But even as Intel and other firms in the high-tech sector suffer through a slump, Intel says it still plans to spend $7.5 billion this year to move to more efficient and cheaper chip-making technology and larger silicon wafers, which also help to cut costs.
``We know it's is going to recover,'' Barrett said of the economy.
Intel has also been making aggressive pushes in recent years to broaden sales beyond microprocessors for personal computers, servers and laptop PCs, which account for about 80 percent of overall Intel revenue and virtually all its profit.
For example, Barrett said that its networking and communications group, which combines two of its previously separate units, now contributes 20 percent of the company's revenue. In the long-term, which Barrett didn't specify, he expects that figure to grow to 50 percent.
Also, Intel's Pentium IV chip, its newest chip, will hit its stride by the end of this year in desktop computers, Barrett said, which is when Intel plans to roll out the Pentium 4 with 0.13-micron technology in large quantities.
``I expect it to be the fastest growing microarchitecture ever,'' he said.
Samsung and Toshiba -- because of the boom in digital storage? But not as bright days ahead for the two?
http://www.siliconstrategies.com/story/OEG20030223S0002
The growth is coming from digital storage and cameras on the NAND-based side of the flash market, and handsets from NOR-oriented devices.
But prices for NAND-based devices are expected to fall 50-60% in terms of price-per-megabyte in 2003 over 2002, Cantore said. In 2002, prices for NAND fell 15-20% in terms of price-per-megabyte, he said.
One of the reasons for the potential price drop is the ongoing battle between the two leaders in the NAND arena--Japan's Toshiba Corp. and Korea's Samsung Electronics Co. Ltd. Samsung is the leader in the NAND market, but Toshiba is aggressively expanding its capacity at an alarming rate--a move that could cause a severe downward pressure on prices in 2003, he said.
“NOR should have a better year than NAND in 2003,” he said. The NOR-based flash market is tied to the handset market. Some 60% of all NOR devices end up in the handset, according to iSuppli.
http://www.siliconstrategies.com/story/OEG20030228S0001
http://biz.yahoo.com/ibd/030310/tech_1.html
Intel is the top seller of a type of flash chip called Nor. Intel sold $2.06 billion worth, or 37% of Nor flash chips worldwide, in 2002. Nor is the type of flash memory that cell phone makers most often use.
For PC Buying, a New Picture
http://www.nytimes.com/2003/03/06/technology/circuits/06upgr.html?pagewanted=print&position=top
This link doesn't require registration(?):
http://www.nytimes.com/2003/03/06/technology/circuits/06upgr.html?position=top&ei=5040&en=cc...
I found three articles. The first article says that worldwide chip sales really have been great, especially for handsets; there are also lots of older PCs that "are due for upgrades". The second article says that iSuppli projects both the PC and handset markets to grow about 10% in 2003, and PC would have been better if not for the first half. Also, Intel's Flash (NOR) should do well because of its tie to handsets. However, the third article says that handset sales actually have NOT been very good, and "analysts expect cell phone sales to pick up".
So, does this mean that the increase in units have been to replenish stock and finished products (at the end-device makers, NOT yet purchased by end-users). Unfortunately, these units have been sold to end-device makers at low ASP. On the other hand, Intel believes it will be able to sell units to end-device makers at HIGHER ASP, it's just that "maybe Intel was one or two months ahead of the call".
If so, it would seem that Intel is actually doing the right thing. (In other words, you want to supply units when you can get a good price for them. Let others supply units for cheap when end-users don't need them yet. This reminds me of how Buffett writes insurance policies.) As soon as end-device makers start to actually sell (both PCs and handsets) through to end-users, Intel stands to gain big? And the fact that Intel hasn't seen higher ASP/margin is simply because end-users haven't really started buying handsets and PCs?
By the way, I would expect CPUs to be even more "just-in-time" than Flash. For example, Dell only needs to drop in a CPU after an order has been placed (but the motherboard built in advance). So CPU unit volume should correlate even more closely with end-user demand.
Another thing, Osha speculated in http://biz.yahoo.com/rb/030307/tech_intel_2.html that Intel "hurt its market share in its older flash memory products". Things may not actually be that bad if: the loss in market share has only been in older products (presumably with lower margin), and/or an immediate revenue miss creates a future pent-up demand (from end-device makers) in high margin products.
1. http://story.news.yahoo.com/news?tmpl=story&u=/nf/20030307/bs_nf/20945
But others say chips are due for a rebound. Last week, the Semiconductor Industry Association said worldwide chip sales were up 22 percent in January over 2002 levels, driven by wireless devices and what the group termed "a recovery in information technology spending."
The SIA estimates that as many as 180 million older personal computers are due for upgrades, promising a strong uptick in the highly cyclical chip business.
2. http://www.siliconstrategies.com/story/OEG20030223S0002
iSuppli originally projected that the PC market would grow by an estimated 13.8% in 2003 over 2002. Now, the El Segundo, Calif.-based research firm predicts 9.8% growth for the year. The reduced forecast reflects a slump "in corporate buying in the first half of 2003," said Joe D'Elia, an analyst with iSuppli, in an interview with SBN.
...
"NOR should have a better year than NAND in 2003," he said. The NOR-based flash market is tied to the handset market. Some 60% of all NOR devices end up in the handset, according to iSuppli.
The cellular-phone handset market is expected to grow 10% in 2003, according to iSuppli. In 2002, the worldwide market for handsets was just under 410 million in terms of unit shipments, according to iSuppli.
3. http://biz.yahoo.com/ibd/030310/tech_1.html
Intel is the top seller of a type of flash chip called Nor. Intel sold $2.06 billion worth, or 37% of Nor flash chips worldwide, in 2002. Nor is the type of flash memory that cell phone makers most often use.
Other makers of Nor flash include No. 2 Advanced Micro Devices Inc., Sharp Electronics Corp. and STMicroelectronics NV.
Although cell phone makers expect to sell more phones this year than last, they're off to a slow start.
"Cell phone sales to end customers were up only 5% to 10% last quarter over the previous quarter," said U.S. Bancorp Piper Jaffray analyst Ashok Kumar in Menlo Park, Calif. Normally, he says, the fourth quarter is stronger due to holiday sales.
As a result, Kumar says, a lot of excess cell phones are sitting around unsold. "That excess inventory is what's impacting Intel's first-quarter performance," he said.
Kumar says this provides a short-term opportunity for Intel rivals like AMD and Swiss firm STMicroelectronics. "Who's really aggressive is STMicroelectronics. I wouldn't be at all surprised if some of it was falling to them," Kumar said.
...
Analysts expect cell phone sales to pick up. Cell phone makers have said they see sales of 420 million to 440 million units this year.
Semico Research Corp. in Phoenix says cell phone makers sold roughly 390 million phones in 2002. It's even more bullish than the phone makers about this year's numbers. Semico forecasts sales will grow 25% to 486 million units.
Timing Wrong
Intel kind of jumped the gun, says Semico analyst Jim Handy.
"Maybe Intel was one or two months ahead of the call," he said.
Handy agrees with Intel's Bryant that cell phone sales will pick up soon.
That should make Intel more competitive.
"Intel should be well positioned when the market turns upward," Handy said.
Just how badly is flash doing?
On 1/14, Intel was expecting the mid point of revenue to be $6.75 billion ("between $6.5 billion and $7.0 billion"). Now, Intel says, "Intel Architecture business is trending slightly above expectation". Because the Intel Architecture business is still so large, *assuming* Flash was/is trending as expected, I would expect the mid point to be about $6.8 billion (Wall Street was expecting higher -- Dan Niles implied $6.85 billion).
Instead, Intel is now expecting the mid point to be $6.7 billion ("between $6.6 billion and $6.8 billion"). That $0.1 billion miss is about 1.5% of total revenue.
In Q4 2002, Wireless Communications group accounted for 9.2% of revenue. In Q3 2002, it was 9.0%. But Wireless Communications also sells XScale. This article says, "Flash products make up about 8 percent of Intel's total sales":
http://story.news.yahoo.com/news?tmpl=story&u=/nf/20030307/bs_nf/20945
1.5% out of 8% is a drop of almost 19%. If Wireless Communications were a separate company, that kind of miss would cause major price declines. Assuming that 19% decline in revenue is after a 30% unit price increase ("Intel raised flash prices by 20 percent to 40 percent in January"), it means a drop in unit volume of almost 38%!
By the way, I also found this ("Flash unit shipments are estimated to rise 21% to 1.8 billion units this year"):
http://www.siliconstrategies.com/story/OEG20030228S0001
-38% versus +21%?
It just doesn't feel very good when Flash is supposed to be the centerpiece of Intel's assault into the wireless space. Anyone has a more cheerful analysis?
Well, I would say that Intel's gross margin has been reasonably consistent with "50% plus or minus a few points". Of course, year 1999 and 2000 were the "bubble" years.
For comparison, I looked up Corning's numbers. Year 1999 through 2001 numbers are from year 2001 annual report. Year 2002 numbers are from Q4 report last month.
Year Net Sales Cost of sales Gross margin
2002 3164 2562 19.0%
2001 6272 4380 30.2%
2000 7127 4131 42.0%
1999 4741 2930 38.2%