Tuesday, February 28, 2006

INTEL's last hope: marketing

The Centrino team at Israel are hyping up their 32 bit Core Duo and 32 bit Merom technology like crazy. You feel a sense of desperation, even the usually soft speaking Nathan Brookwood of Insight 64 said "[o]nly a psychoanalyst would be fully qualified to analyse Intel's behaviour".

First let's analyse the hype on Core Duo, a 32 bit mobile processor with 2 Pentium M cores.

INTEL's Corporate VP of the Mobility Group Mooly Eden had a presentation on Core Duo which I reported here. According to his presentation slides at here, Core Duo had the following selling points:
  • Revolutionary processor
  • 70% performance increase
  • Over 28% longer battery life

What's revolutionary about the 32 bit Core Duo? According to Mooly Eden, there were two technologies, Dynamic Intel Smart Cache Sizeing and Enhanced Intel(R) Deeper Sleep. Frankly, I don't see anything revolutionary here, these two can be best described as minor improvements, Core Duo is pretty much two Pentium M cores glued together and put on a shared front side bus. After so many years, the Israeli amateurs have not figured out how to do 64 bits and Mooly Eden's excuse was that "It may take many years for enterprises to demand it (64 bit)".

What about so called 70% performance increase? Let's look at an independent benchmark done by AnandTech. As we can see, the 2GHZ 2MB cache Core Duo is generally slower than AMD's lowest Athlon 64 X2 3800+ (2GHZ, 1MB cache). The X2 3800+ is 20% faster than Core Duo in DivX encoding, X2 3800+ is 17% faster than Core Duo in Windows Media Encoding, 11% faster in playing Battlefield 2. Overall, the X2 3800+ won 16 benchmarks, the Core Duo won 6. In tests the Core Duo did win, the margin of victory is always less than a few percent. Keeping in mind that the 2GHZ Core Duo is INTEL's top of line and the Athlon 64 X2 3800+ is AMD's lowest entry level, you see although Core Duo represented a quantum leap from Pentium 4 D, its performance is still far below AMD64.

What about the so called "28% longer battery life"? Again, it's hype with no factual foundation. According to this Tomshardware test, Core Duo consumes more than twice the power of a Pentium M or Turion when doing low load work such as reading and office work. According to INTEL's design data sheet, Core Duo has an estimated Thermal Design Power of 67 watts, which is consistent with my estimate that Core Duo is at least 53 watts based on an AnandTech measurement of system power. In comparison, the Opteron 870HE is only 55 watts max. No wonder DELL is only making 17inch, 8 pound desktop replacement notebooks with Core Duo.

So, we fully analysed INTEL Israel's hypish marketing on all three aspects. Core Duo is no revolutionary chip, is just a modification of Pentium III; Core Duo is not 70% faster, but 10-20% slower than AMD's lowest entry level Athlon 64 X2 3800+; Core Duo does not run cooler, but runs hotter than Opteron 875HE server chip (2.2GHZ, dual core, 8 way SMP).

I am impressed by Israeli's ability to hype, but I am equally unimpressed by their ability to deliver.

Now, what about the "revolutionary" NGMA called Merom designed by the same Israeli team? Again, from the all hype no beef messages, I don't see anything revolutionary there for the so called NGMA. The only new feature we know is the so called 4-issue core. However, the PowerPC 970 has a 4-issue core, yet Steve Jobs was not impressed. Can INTEL Israel do better than IBM? I seriously doubt it.

When Mooly Eden went back to Israel to talk to some of his old collegues, they told him: 'You're only one year in marketing, and already you're brain-damaged."'

Friday, February 24, 2006

INTEL's collapse will be sooner than I expected

In this article, I proposed a way out for INTEL in 2006. I was trying to be in INTEL's seat and find a route to survival, I also explored some alternatives. The conclusion was there was ONLY one way for INTEL to tough out 2006 and hope for a better 2007.

It is simple but inescapable logic: INTEL's unit share loss is inevitable due to its inferior technology, even Joe Osha sees that today. The only way for INTEL to maintain constant revenue is to reduce production and hike prices. This is a workable plan, because AMD can only supply about 30% of the market, INTEL can dictate the prices on the remaining 70%.

By reducing production and hiking prices, INTEL can keep inventory low and maintain constant revenue and even higher profit for 2006 and hope to come back in 2007 with vengeance.

But INTEL execs seem have chosen instead to expand production on soon to be obsolete Netburst and 32 bit Core Duo products and slash prices, despite the fact that INTEL's manufacturing efficiency is only a fraction of AMD's. The result of this is the following:

1) Since unit loss is inevitable, reduction in price leads to revenue shortfall and crash of stock price.

2) Inventory build-up of Netburst (P4 and Xeon) and 32 bit mobile (Core Duo, P-M) will perpetuate into 2007, causing profitability problems for the next couple of years.

For INTEL folks, it will be a long and harsh winter ahead.

Lenovo Notebook for SMB will be AMD only

Pervasive mobile 64 bit computing offered by Lenovo, cool! The oudated INTEL 32 bit Core Duo 32, Core Solo 32 and Centrino 32 still rule the 3rd world PC market though.

BTW, HP has some cool products, HP Compaq t5720 Thin Client (Geode NX1500, total system power < 50 wat), Athlon 64 1500 blade PC (peak power 35 watt).

Opteron 890 being tested

It seems the AMD+IBM's new process can indeed boost performance by another 40%, now it's reported that Opteron 890 is being tested in CompAmerica workstations. But, first things first, we shall see Opteron 885 for sale in just a few days.

Opteron blades rule

HP's server blade revenue grew 58% year/year, driving the growth is the AMD Opteron processor. HP's blade offering is now mostly Opteron based, it has retired the Xeon based BL30P and BL40P. Apparently, the 150 watt Xeons are just too much to handle. The Opteron is cool and can have more memory than the Xeons.

Thursday, February 23, 2006

Dual Core Desktop Duel: INTEL won two medals

CNET Dual-Core Duel

Round 1: Day-to-day computing: INTEL wins bronze medal
Round 2: Multitasking: INTEL gets 4th place
Round 3: Photo editing: INTEL succeeds in 4th place
Round 4: MP3 encoding: INTEL achieves 5th place
Round 5: Video encoding: INTEL grabs bronze medal
Round 6: Gaming: INTEL's takes 5th place, its Extreme Edition slower than AMD's lowest chip
Round 7: Price versus performance: INTEL's D820 deserves honorable mentioning for being the cheapest

To my surprise, INTEL's Pentium snatched two bronze medals. It did better than India in Olympics, in Sydney 2000 games, India got only ONE bronze medal, in Athens 2004, India got one and only silver medal.

My congrats to INTEL engineering and marketing.

Top INTC and DELL dumpers and AMD buyers

Logically, AMD buyers must be INTEL dumpers, and INTEL dumpers must be DELL dumpers. This is the winning strategy I employ and it would be interesting to see which big houses are playing the same game.

So let's go to www.nasdaq.com, and check the most recent institutional ownership reports.

The top 5 AMD buyers are:

FMR group: added 11.755 million AMD shares to 20.9 million
CALAMOS ADVISORS LLC, from 0 to 10.249 million
JANUS CAPITAL MANAGE... : added 8.559 million to 30.992 million
AXA : added 5.7 million to 24.83 million
ARTISAN PARTNERS LTD... : from 0 to 3.16 million

I should also emphasize Ben Lynch's DEUTSCHE BANK AG\ which downgraded AMD to sell increased its AMD position by 32% to 5.78 million shares.

The top 5 INTC dumpers are:

CITIGROUP INC : dumped 27.772 million shares, or 68.8%
FMR CORP : dumped 34.353 million shares
AMERIPRISE FINANCIAL... : dumped 9.5 million
OPPENHEIMER FUNDS IN... : dumped 8.39 million
GOLDMAN SACHS GROUP ... : dumped 7.76 million

The top 5 DELL dumpers are:
AXA : dumped 53.822 million shares, or 77.53%
CITIGROUP INC : dumped 26.804 million shares, or 78.11%
GOLDMAN SACHS GROUP ... : dumped 9.074 million shares
FMR CORP : dumped 7.75 million shares
TIAA CREF INVESTMENT... : dumped 4.954 million shares

Looks like FMR group is the most consistent in this period. However, many of the AMD buyers had dumped INTC and DELL in the previous reporting period already.

Lenovo launches commercial PC in Americas, DELL in trouble

AMD and Lenovo just announced their global partnership. Lenovo is going to sell AMD64 powered PCes to SMB customers in United States, Canada, Mexico, the United Kingdom, France, Germany, Russia, Australia, New Zealand and Southeast Asia. According to the press release, already 50% of Lenovo desktops are AMD64.

As I pointed out here, since 52.6% of DELL's revenue and 68.3% of DELL's profit come from corporate market in Americas, and AMD is going to penetrate US corporate market in a big way, if DELL failed to go AMD before 2Q06, then it would be too late. Now it turned out that both HP and Lenovo have already booked AMD's capacity, DELL is out of the AMD64 game. Lenovo will definitely beat DELL on price and performance with ADM64 and its China based manufacturing.

During DELL's CC, analysts asked tough questions about its falling margins, now it seems DELL has no answer to those questions at all.

Enron time may be sooner than I expected.

Wednesday, February 22, 2006

DELL plays hide-and-seek

In the earnings conference call, DELL CEO Kevin Rollins dodged some questions and offered to provide more details in its April analyst meeting. Now DELL has postponed the meeting to September 2006, saying there will be no news until then.

My take is this: DELL is in a very tough situation:

0) Outdated INTEL FSB based technology is a tough sell.

1) INTEL will have to cut co-advertising funds and reduce special kickbacks to DELL, as INTEL itself is in deep inventoy build-up and has been downgraded to a target price of $16.

2) But AMD capacity is all booked by HP, SUN, Lenovo, Acer, DELL won't get allocation even if it wants to pay higher prices.

Lenovo is going to launch PCes in US soon, targeting small and medium businesses.

Socket 754 Athlon 64 benchmark duel against Xeon EM64T

I have an Athlon 64 3000+ (socket 754, 2GHZ, 512KB L2 cache) box running 64 bit Linux, it has 1GB PC3200 RAM and two 7200RPM SATA drives. I bought the CPU+Motherboard combo at Frys for about $80 on sale, the board is an ECS K8M800-M2. Since all AMD64 processors share the same key characteristics (embedded memory controller, hypertransport, PowerNow, etc), I consider this Athlon 64 Socket 754 a poor man's Opteron, even though it only supports single channel DDRI.

So let me do some basic tests. First we run some commands to show system info:

%uname -a
Linux 2.6.9-22.0.1.EL #1 Thu Oct 27 14:29:45 CDT 2005 x86_64 x86_64 x86_64 GNU/Linux

%more /var/log/dmesg
time.c: Detected 1999.800 MHz processor.
Memory: 960900k/982976k available (2398k kernel code, 21328k reserved, 1303k data, 164k init)
CPU: L1 I Cache: 64K (64 bytes/line), D cache 64K (64 bytes/line)
CPU: L2 Cache: 512K (64 bytes/line)
CPU: AMD Athlon(tm) 64 Processor 3000+ stepping 00

powernow-k8: Found 1 AMD Athlon 64 / Opteron processors (version 1.39.04)
powernow-k8: cpu_init done, current fid 0xc, vid 0x2
eth0: VIA Rhine II at 0xee001000, 00:11:5b:fd:6b:8d, IRQ 193.

09:39:13 up 108 days, 13:48, 4 users, load average: 0.08, 0.15, 0.17

The server is live and running quite some web sites, with little CPU load. Now, let's do some OpenSSL benchmarks on our little Socket 754 friend. OpenSSL is used for secure internet communications and its performance mostly depends on CPU speed.

%openssl speed rsa
OpenSSL 0.9.7a Feb 19 2003
sign verify sign/s verify/s
rsa 512 bits 0.000381s 0.000027s 2626.1 37486.3
rsa 1024 bits 0.001228s 0.000067s 814.6 14886.2
rsa 2048 bits 0.006972s 0.000200s 143.4 5006.9
rsa 4096 bits 0.044612s 0.000677s 22.4 1478.0

% openssl speed dsa
sign verify sign/s verify/s
dsa 512 bits 0.000226s 0.000238s 4430.7 4208.7
dsa 1024 bits 0.000566s 0.000677s 1765.3 1476.7

As we can see from the result, the Athlon 64 3000+ can do 1478 verifies per second for 4096 bit RSA keys. I don't have access to an Xeon box now, fortunately, AnandTech has the OpenSSL benchmark results for Xeon 3.6GHZ EM64T here, allow me to copy part of it (for Xeon 3.6GHZ EM64T):

sign verify sign/s verify/s
rsa 512 bits 0.0004s 0.0000s 2376.9 28489.3
rsa 1024 bits 0.0017s 0.0001s 604.8 11277.6
rsa 2048 bits 0.0092s 0.0003s 108.9 3682.0
rsa 4096 bits 0.0601s 0.0010s 16.7 1034.1

sign verify sign/s verify/s
dsa 512 bits 0.0003s 0.0003s 3364.6 2862.7
dsa 1024 bits 0.0008s 0.0009s 1260.6 1057.9

Surprised? Not to me! As I expected, the Athlon 64 3000+ (Socket 754, 2GHZ, 512KB L2) met my expectation as a poor man's Opteron and handily beat INTEL Xeon EM64T 3.6GHZ by approximately 40%.

Now, let me do some Apache stress testing on our poor little Socket 754 friend. I will use the "ab" command to retrieve a SHTML web page, since it's shtml, the Apache server will have do some processing, so this is a tougher test than static HTML. I am running the ab program on the Athlon 64 box:

[root@my ~]# ab -n 50000 -c 10 http://domain-name-replaced/forum.shtml

Server Software: Apache/2.0.52
Server Hostname: domain-name-replaced
Server Port: 80

Document Path: /path-replaced/forum.shtml
Document Length: 267354 bytes

Concurrency Level: 10
Time taken for tests: 41.925014 seconds
Complete requests: 50000
Failed requests: 0
Write errors: 0

Total transferred: 13381450000 bytes
HTML transferred: 13367700000 bytes

Requests per second: 1192.61 [#/sec] (mean)
Time per request: 8.385 [ms] (mean)
Time per request: 0.839 [ms] (mean, across all concurrent requests)

This is not bad at all. From this test, under 32 bit Windows, the dual core Xeon 2.8GHZ handles 1199 requests/second. Although GamePC's test did not provide information such as file size, our tests was on a 260KB page, which is bigger than average. Our tests were under 64 bit Linux, so there must an advantage there.

The conclusion is obvious: Socket 754 Athlon 64 3000+ is 40% to 100% faster than the fastest single core Xeon processor in web environment, and is probably faster than dual core Xeons too.

Tuesday, February 21, 2006

Real men have 64 bit laptops

Once a guy showed off his new notebook computer to me, it looked kinda cute but was an INTEL 32 bit Centrino. He asked me what kind of laptop I have, I said: "I have a 64 bit notebook".

I took out my notebook, powered it on, the GRUB boot screen appeared, the first option was CentOS 4 for AMD64, the second option was Windows XP X64, the third was Windows XP 32 bit. And I said: "Yours uses an INTEL CPU. INTEL has not mastered the AMD64 instructions yet".

I hit the enter key, the beautiful Linux boot sequence scrolled along, quickly, the 64 bit kernel booted up, powernow-k8 started, X Window started, I promptly logged in as root like a seasoned UNIX sysadmin, typed "openssl speed rsa", in a brief moment, the results were produced, and I told da guy: "check the score, this is much faster than your 3GHZ DELL PowerEdge 1850".

I forgot to run the Apache bench and say: "Real men have 64 bit laptops".

VIIV will be fragged by AMD Live!

This page gave the specifications for INTEL's V//V and AMD Live! Allow me to copy it here:

AMD LIVE! Desktop PC
AMD Athlon 64 X2 dual-core processor (65W energy efficient CPU)
Windows Vista Premium (includes Away Mode support)
Discrete or UMA graphics (DX9, HDCP required if HDMI used)
5.1 HD audio with SPDIF output
RAID 0. 1 with SATA (NCQ support recommended)
2 to 4 DIMMs of DDR2 667/800MHz
Ports: 6 (minimum) USB 2.9 and IEEE 1394 (FireWire 400)
GigE LAN and optional 802.11b/g (a is optional)
Slots: 16L & 1L PCIe and 1x PCI minimum
Low Acoustic Noise level (to be determined)

AMD LIVE! Entertainment PC
Same as LIVE! Desktop PC except:
AMD Athlon 64 X1 dual-core processor (35W and 6W energy CPU)
SATA HD support
Support for low profile form factors

Intel Viiv Platform
Core Duo Processor / Pentium D / Pentium Extreme Edition
Intel 945 series or 955 series chipset
Pro 100/1000 PM/VE/VM network adapters

Clearly, INTEL's V//V is just another effort of bundling outdated INTEL 32 bit FSB based hardware for more profit, while AMD Live! is a set of industry leading technologies combined to enhance people's life style.

Monday, February 20, 2006

Capacitor life and failure of INTEL motherboards

DELL spent $300 million fixing PCes with failed capacitors (for those who don't know, DELL only uses old FSB based INTEL CPUs) . It was then reported that similar capacitor failures were common for HP's INTEL PCes. I have conjectured that the extreme heat from the INTEL CPUs are shortening the life of the capacitors on the motherboards.

According to this study by United Chemi-Con, "temperature (ambient temperature and internal heating due to ripple current) is the most critical to the life of aluminum electrolytic capacitors".


"Because a capacitor is essentially an electrochemical device, increased temperatures accelerate the chemical reaction rates within the capacitor (usually a 10°C rise in temperature will double the chemical reaction rate). Therefore, higher temperatures cause accelerated changes in decreasing capacitance and increasing tan d due to the gradual evaporation of the electrolyte through the capacitor seal".

So, in essence, higher temperature leads to faster evaporation of electrolyte inside the capacitor, and leads to dead capacitors. How much faster?

According to this study, when you increase the temperature by 10C, the weight loss of the electrolyte increases 1.9 times. In other words, 10C of temperature increase cuts the capacitor life by half. Needless to say, when the capacitor is dead, the motherboard is dead, and the system is dead.

Other factors, such as operating voltage and ripple-current also affects capacitor life.

As we can see from this test, INTEL's Pentium requires about 100 watts more than AMD64, thus generates more heat inside and outside the fragile capacitors, making INTEL systems less reliable.

According to some report, INTEL is switching to a different voltage regulator design, probably one without capacitors for the future Conroe and Merom systems.

Even though heat is not a problem for power efficient AMD64 CPUs, some Opteron boards already use digital VRMs.

A casual look at DELL's balance sheet

Every month, we look at our money in the bank and then look at the bills. Let's do this for DELL.

Cash + cash equiv: $7.042 billion
Short term investment: $2.016b
Other (who knows what?): $2.62b
Investment: $2.69 b

Total money in the bank: 14.368

Bills to pay: $15.927 b
Debt: 0.5 b
Other debt: 2.549 b

Total: 18.976

DELL is $4.6 billion short of paying its bills and debt, even if you add the $4.089 billion DELL can collect from its customers, it's still $500 million short.

Outdated INTEL technlogy doesn't sell

Philips cancelled a 700 million contract with DELL. SAP recommended Opteron as 2006 upgrade for better ERP peformance. More and more enterprises are realizing that INTEL technology is antiquated and has no future.

I had predicted that the impending mass extinction of Netburst based Pentium 4 and Xeon will slow customer purchase of INTEL technology in 2006. Few enterprises want to invest millions in CPUs that will be obsolete in just 6 months. Even those diehard INTEL fanboys will wait till 4Q06 for Conroe and Merom. Coupled with growing market recognition of AMD64, I expect 15-20% fall of INTEL's revenue in 1Q06.

Yet, even fewer enterprises will invest millions in CPUs that will be obsolete in just 3 months, therefore expect INTEL's revenue to fall even more in 2Q06, as we get nearer to the release of INTEL's next generation 32 bit technology that is incompatible with the AMD64 and HyperTransport industry standards.

Those who are uninformed and still pouring millions into the dying Netburst P4 and Xeon will regret deeply in just a few months, and will turn their backs to INTEL in the near future.

The whole INTEL architecture is in an unstable flux. We hear about all these INTEL vaporware which no one has seen yet touted as the next killer chip. Even if its engineering matches its marketing, INTEL means discontinuity. INTEL's outdated FSB technology often requires new chipsets and new buses when upgrading the CPU, this means you will have to get a new motherboard for a new INTEL CPU. You need a new motherboard when going from Paxville to Dempsey, then you need another new board when going from Dempsey to Woodcrest, when you choose Conroe for desktop, it's a whole new purchase yet again-- even the voltage regulator must be changed.

AMD64 provides a continuous upgradable path. You can upgrade from single core AMD to dual core then to quad-core. The beauty of HyperTransport even allows today's 3rd party chipset to work with future AMD CPUs -- AMD64 CPUs connects directly to memory and HyperTransport is the only communication between the CPU and the rest of the external world, and HyperTransport is backward compatible.

Software compatibility will be a major issue for INTEL. Microsoft Windows Vista is built for AMD64, you can see this when you use Microsoft's C++ compiler options. SUN Java officially only supports AMD64. Linux has work around code for INTEL's EM64T but there are major performance compromises. Solaris 10 doesn't support EM64T at all.

Sunday, February 19, 2006

AMD's strategy must based on killing both INTEL and DELL

Strategic situation:

1) Most of INTEL and AMD's revenue came from outside of Americas. Over 59% of INTEL's revenue comes from Asia (including Japan).

2) 64% of DELL's business is in Americas, only 12% of its revenue comes from Asia.

3) HP is winning the consumer market in US; HP and SUN are winning the global x86 server market; Acer is gaining in Europe; Lenovo is gaining in Asia.

4) AMD grows by takeing market from INTEL.

AMD's Strategy:

1) Winning the lower end of US server market, the slashing of Opteron 2xx prices was a decisive move that will drive the Xeons into extinction.

2) Stab at the beast's belly: Asia. For that, AMD must firm up its relationship with Lenovo and enforce JFTC recommendations with various Japanese vendors. Dell is irrelevant in that market.

3) Winning US corporate desktop/workstation market with HP and SUN. AMD should establish a programme to bundle Opteron servers and Athlon 64 PCes for corporate markets.

4) Develop a high end Turion 64 to push down INTEL's ASP on mobile, making all INTEL chips the ultra low end.

An integral part of the AMD strategy should be playing the DELL game for fun and profit. The aim of the game will be:

1) DELL getting bigger and bigger portion of the INTEL pie, right now, DELL accounts for about 25% of INTEL's units, eventually, DELL should account for 40% of INTEL's units, based on the assumption that INTEL's unit share drops to 60%, and ASP drops by 20%.

2) HP, Lenovo, Acer and Gateway becoming more AMD, with Lenovo, Acer and Gateway predominantly AMD.

3) HP and SUN ruling the x86 server market with unbeatable price performance.

4) DELL being pushed into the ultra low end of the PC market with outdated INTEL technology designed by middle-east amateurs.

Always remember:

DELL is just a screw driver company that can be replaced by another screw driver company overnite.

The myth of Centrino power efficiency

This is a real measurement of power consumption of Turion 64 and INTEL Centrino 32. Under load, the Turion 64 MT-40 2.2GHZ consumes 26.4 watt, while the Pentium 32 M 2.13 GHZ consumes 23.3 watt. However, keep in mind the Pentium 32 M needs a separate memory controller which also consumes power.

Saturday, February 18, 2006

If Jim Schneider and Kevin Rollins were telling the truth

If they were not lying and defrauding investors, and the last week of January 2006 did only generate 30% to 40% of revenue of a normal week (2-3% revenue from 7% of the days). Now, considering that DELL has the best prices for INTEL parts, this would indicate that INTEL PCes and servers were not selling, even at rock bottom prices.

RBC Capital Markets did channel checks and concluded that INTEL's "March quarter, as well as 2006 business trends, are significantly lower than current expectations."

How much lower? My estimate is 15% to 20% lower.

Informed computer users who want to buy a PC now will choose AMD64, even INTEL diehards not in a hurry will wait till 4Q06 for Conroe. The 3rd world is INTEL's only refuge, but 60% of DELL's business is in US. (100- 60 = 40, matches DELL's weekX revenue numbers).

DELL investors want DELL dude back

I am amazed by DELL stake holders' insight. The DELL dude who made DELL popular was arrested in 2003 for buying pot. Some DELL investors are now saying that things went downhill since he was fired.

The value of this truth telling Journal

I hate to see poor individual investors screwed by the rich. Enron is a good example. WorldCom is another. Many less informed, less intelligent and gullible people got financially ruined by high-IQ crooks.

DELL reported after Feb 16, 2006 market close, it claimed that Dell Delivers Strong Performance with Record Revenue and Earnings for Fourth Quarter, laiden with hyping messages such as "strong growth" and "strong profitability" "across all regions" and "across all segments". The media echoded. USAToday: Dell sets revenue record for 4Q; Reuters: Dell Q4 earnings rise, Dell sees industry-beating growth in Asia for '06; AP: Dell 4Q Profit Soars 52 Percent to $1.01B; CNN Money: Dell beats the Street....

DELL's stock jumped momentarily after hours, some DELL investors got really excited, INTEL investors were even more excited.

I knew it was impossible for DELL to beat, outdated INTEL 32 bit technology doesn't sell. No matter how much lipstick you apply to a pig, it's still a pig. So I took a look at DELL's report, and wrote "DELL: the real story". It analysed the core businesses of DELL, and showed that DELL's revenue was in sequential decline in the holiday season. I also pointed out that DELL and INTEL are in perfect sync, and DELL's weak January is a bad omen for INTEL's 1Q06.

Wall Street is no fool. DELL's stock plunged big time the next day, as people started to see the real picture.

INTEL endures more frags

Pentium XE 955 (65nm Presler dual die) fragged by Athlon 64 FX60 and Athlon 64 X2 4800+. INTEL Pentium M 760 fragged by AMD Turion 64 ML-44. Jump to the conclusions page if you don't care about the details. Now the world is waiting for the Conroe 4 issue core the middle-east amateurs touting about set to release in 4Q06. However, before INTEL makes that move, AMD's grand masters from DEC, IBM and SUN have readied more frags for the bragging INTEL amateurs -- Socket AM2 CPUs with 800MHZ DDRII support and enhanced HyperTransport. Some analysts are saying that INTEL's 2006 will be significantly below expectations.

Even INTEL engineers say their leadership is "brain damaged".

I think so too.

Friday, February 17, 2006

INTC stock gets whacked thrice per quarter

Observation and projection for INTC

1) first INTC reports bad results, whack #1

2) then AMD reports share gains, whacked again

3) then DELL reports bad results the next month, whacked yet again

Observation and projection for DELL

I smell something cooked in Jim Schneider's testimony.

DELL's price/book is at 17.5 right now, much higher than GOOG's price/book of 11.5.

At the current rate of decline, DELL's stock holder equity will become 0 in 18 months -- Enron time.

Dell's weekX

For the last 4th quarter report, DELL included 14 weeks of results, from Oct 29, 2005 to Feb 3, 2006 and compared it with the results from Oct 30, 2004, to January 28, 2005, 13 weeks of last year.

Specifically, 13 weeks were the same, the extra week was from Jan 30, 2006 to Feb 3, 2006, let's call this extra week the weekX (Jan 30, 2006 to Feb 3, 2006).

The obvious question is what kind of revenue contribution came from the weekX?

If we simply assign the average weight of 1/14 or 7.14% to weekX, we find DELL's revenue declined sequentially on all segments and declined year/year on desktop and server.

DELL dudes are no idiots, so they would try very hard to downplay the weekX.

During the DELL conference call, its CFO Jim Schneider stated that "we had an extra week in the quarter that contributed about 2 to 3 points of added growth, which was slightly more than we expected."

So according to DELL's CFO, weekX was only 30% to 40% of an average DELL week, and actually DELL was expecting less from weekX. Basically, DELL was saying that weekX should be and was pretty much a dead week. What's special about this week, which was one month after the Christmas week? What about weekX+1 and weekX -1?

The Wall Street analysts are not stupid, they sensed the problem there and pressed on for more details on revenue from the weekX, and Kevin Rollins, DELL CEO stated under oath:

"I would say, because the consumer business trends off in the January timeframe, and it has its kind of lowest in ebb in Q2, the gains would have most likely been made in the corporate and enterprise arena worldwide at the end of that week. That’s why we were a little bit guarded on the extra week, because corporations don’t buy per day. They have a budget in a quarter that their IT folks are able to spend or not spend and meet. So we did not think we would get a full week’s impact out of that extra week due to that."

This is a very wordy answer for a simple question. Every company has a determinative method of recognizing revenue, DELL is no exception. According to DELL's SEC filings, essentially, revenue is recognized when the products are delivered to customers. Therefore, there is a specific revenue number for weekX and DELL knows what that number is -- just account the products leaving DELL's shipping docks, and DELL can pull that information at a mouse click as its system is fully computerised.

I bet that one day, SEC will have to open DELL's books and compute the revenue for weekX.

SUN Java does not support EM64T

Go to the Java download page at java.sun.com, there is only Java runtime for AMD64, no support for EM64T. I think INTEL's engineers should work 10 times harder so they can get an AMD64 compatible CPU out.

SUN has a nice online presentation showing the advantage of AMD64 architecture and SUN's Galaxy design at here.

Thursday, February 16, 2006

Opteron 185, 285, 885 benchmarked, 35 watt dual core AM2 desktop

Unlike the 120 watt 285SE, 885SE (special edition) used by SUN, these are 95 watts.

Some more Socket AM2 info is here. Most interesting is that there will be a dual core Socket AM2 CPU rated at 35 watt max, the same as current Turion 64 ML.

DELL: the real story

DELL just reported results for the period of 14 weeks from Oct 29, 2005 to Feb 3, 2006 with revenue of $15.2 billion and EPS of $0.43, beating average estimate of $0.41. Sounds good?

First, wait a minute, in last year, the reporting period was from Oct 30, 2004, to January 28, 2005, 13 weeks. There is one extra week for the current period, there should be a 1/13 adjustment for a fair comparision. The burning question is: did all those analysts add an extra week when estimating DELL's results? Let's adjust DELL's revenue, and let's compare DELL's results year over year, and quarterly sequential.

Real growth was computed below as:

(Revenue_2005/14 - Revenue_2004/13) / Revenue_2004/13 - 1 = Revenue_2005/Revenue_2004* 13/14 -1

Segment2005 Q 2006 QPrev Q Adjusted 06 RevReported GrowthReal Y/Y
Real Q/Q Growth
Desktop$5.6 b$5.6 b$5.1 b$5.1 b1%-7.14%0%
Server$1.3 b$1.4 b$1.4b$1.3 b10%0%-7%
Mobile$3.1 b$3.8 b$3.6$3.5222%13.82%-2.2%

According to IDC, 4Q05 global PC market grew 17.1% year/year. DELL is substantially below market growth in every sector. In fact, its desktop revenue decreased 7% year/year, and server showed 0 growth year/year. When we compare Q/Q results, the situation is even worse, the fourth quarter was a holiday season with 8% seasonal growth, yet DELL showed flat or negative result Q/Q for all segements. DELL's sequential decline pattern matched that of INTEL's declining pattern perfectly.

Now, look at DELL's balance sheet: it has $9.06 billion cash and short term investments, but $9.84 billion of bills to pay. DELL's stock holder equity dropped from $4.8 billion to $4.1 billion in just three months.

$4.1 billion net asset and $75 billion market cap, with revenue dropping year/year on two its main segments.

I predicted that DELL will the next Enron, mark my word!

Wednesday, February 15, 2006

Dual Core Turion 64 in the wild

LaptopLogic has a screenshot showing Dual Core Turion 64 running at 1.8GHZ. What's amazing was the voltage was only 1.075V. In comparision, INTEL's Core Duo 32 bit CPU requires 1.2V.

Tuesday, February 14, 2006

INTEL's postmortem

Unlike my modern history of AMD-INTEL, which started from 1GHZ Athlon, Jakub Wojnarowicz at FiringSquad traced back from the K6-II and 3DNow!, and pointed out that INTEL started to merely reacting or responding to AMD's moves since 1998.

However, the seemingly frustrated author failed to discuss INTEL's intrinsic defects that led to its forseeable demise.

I wrote two articles on the root of problem: Intel's defective gene and the inbreeding that perpetuated the defects.

The last time INTEL brought in some new gene pool was Bob Colwell, who took some RISC ideas into the x86 and designed the Pentium Pro, which is the basis of all INTEL Cores today.

AMD may also have some defective genes, but it kept bringing in new blood: first new tech talent from NexGen, DEC, then management from Motorola (Hector) and more talents from IBM, SUN and INTEL. Opteron is Alpha EV7 for x86, and I am sure the IBM, SUN folks will bring fresh ideas. AMD is constantly evolving.

Dual core Opteron has 4X price/performance advantage over Xeon DC

Opteron 265 price has been slashed to $307, and Opteron 270 price has been cut to $441. INTEL's Paxville 2.8GHZ is selling at about $1200.

However, Opteron 265 is much much faster than Paxville. In Apache Benchmark, with 50,000 users, Opteron 265 handles 1510 requests/second, Paxville 2.8GHZ does 1199 requests/second.

Price/Performance for INTEL Xeon Paxville: 1199/$1200

Price/Performance for Opteron 265: 1510/$307

Opteron 265 has 4X price/performance advantage over dual core Xeon.

The three rules of the Street game

Rule 0: Stocks are eventually valued by forward P/E. This is the fundamental rule.

Rule 1: Analysts are very smart crooks employed by big money.

Rule 2: Your goal is trying to win money by buying and selling stocks.


*) Sometimes the crooks may even try to create a perception that Rule 0 is changed, for instance, they may try to redefine valuation formula to other seemingly justfiable schemes, as we have witnessed in the dot.com era.

*) Big money also has control of media.

*) Crooks may tell truth, but it all depends on timing.

*) To win, you must be smart.

AMD institutional holdings check

Downgrades, upgrades, buys, sells.... the game of the Street. Now Nasdaq.com has updated AMD holdings info showing some December 31, 2005 reports, so let's check it. Notably:

*) FMR Group increased it AMD holdings by 128.21% to 20.9 million shares, while dumping 13.92% of its INTEL holdings.

*) CALAMOS ADVISORS LLC went from 0 AMD to 10.249 million shares.

*) FRIESS ASSOCIATES LLC had a similar story went from 0 to 7.35 million AMD shares, it dumped all its INTC shares.

*) Goldman Sachs seemed to be a real loser, it sold 12.6 million AMD shares, but it also sold about 10% of its INTC holdings.

*) Morgan Stanley is true to its words, upgraded AMD and increased its AMD holdings by over 200%.

*) UBS crooks had sell rating on AMD, but silently increased its AMD shares by 123.70%. Seeming to be satisfied by the quantity of AMD shares it has now, it is upgrading AMD from sell to neurtal, saying INTEL's new CPUs in 2006 won't slow AMD's market share gain.

Microsoft.com standardized on AMD Opteron

There is an interesting Microsoft white paper (word file) on how Microsoft moved away from INTEL to AMD Opteron running Windows server 2003 x64. By moving from IA32 to AMD Opteron, Microsoft saw a 50% decrease in CPU load. The paper provided quite some insights and best practice suggestions.

"In addition to the 50 percent decrease in CPU utilization, the 32-bit servers experienced noticeable spikes in which the CPUs were utilized at 100 percent for sustained periods of time. The operations team determined that a spike occurred when one or more application pools ran out of virtual memory and recycled. On the x64-based servers, these spikes do not exist because the application pools are no longer running out of memory."

In this paper, Microsoft tested some applications on Opteron on Windows x64 and was seeing up to 17.7 times performance gain (1770%).

"Microsoft.com has standardized on the HP Proliant DL585 models for the x64-based Web servers with four 2.2-GHz AMD CPUs and 16 GB of RAM. These enterprise-class servers have many redundant components, including power supplies, network adapters, and cooling fans, and have the capability for redundant array of independent disks (RAID) disk storage and RAM configurations."

No surprise here, Opteron is the king of AMD64. We all know INTEL's EM64T runs slower in 64 bit mode than 32 bit mode. IA64 (Itanium) has also lost Microsoft's blessing.

Monday, February 13, 2006

Why DELL will go AMD sooner than later

because it has no choice!

0) Because of the competitive advantage of AMD64, DELL will enter the AMD market eventually. But DELL wants to choose the best timing to swicth to AMD --- when AMD64 has wide market recognition and capacity, DELL will just ride the wave. That's the perfect plan. But, INTEL is no idiot.

1) INTEL can't allow DELL to suck its blood any more--lower units and lower ASP leads to lower revenue, more DELL business means less money for INTEL.

2) So INTEL has to eliminate the extra kick backs to DELL.

3) If DELL decides to switch to AMD, then INTEL eliminates the kick backs, DELL will cry loud saying INTEL cut the candies because of AMD, and that's anti-trust violation.

4) So INTEL has to strike first, before DELL makes the move, INTEL must eliminate the rebates. This way DELL has no way to complain, this also weakens AMD's lawsuit a little bit.

5) DELL going AMD will only lead to marginal benefit for AMD at first, it will eat a lot of whitebox AMD business. However, once AMD has the capacity, DELL may go 100% AMD.

AMD's Dual core faster than INTEL's quad-core

AMD's Opteron 285 (2.6GHZ) will arrive for purchase on Feb 24, 2006. Unlike the 120 watt Opteron 285SE which has been used by SUN for months, this is a 95 watt CPU. It seems AMD can continue to derive efficiencies from its advanced SOI process.

In Apache Benchmark, a 2.6GHZ FX60 dual core is twice as fast as a INTEL 2.8GHZ dual core Xeon Paxville. This implies that a 2.6GHZ AMD dual core is faster than INTEL's 2.8GHZ quad-core expected in 2007.

No real surprise here, we know INTEL's outdated shared bus architecture can't scale beyond two cores.

The INTEL-Skype dirty plot and why FTC is sitting idle

In the previous article "INTEL to stop AMD with Skype exclusive deal", we analysed INTEL and Skype's dirty plot to exclude AMD from the VoIP market. Today, CNET.com made a further report on how the knee-capping attack was carried out by INTEL and Skype. Specifically, CNET found that Skype was using an operation called "Get CPU ID" to identify the type of processor running on the PC. Unless the CPUID matches "INTEL", it will prevent the user from doing 10-way conference calls.

Note this has nothing to do with CPU performance. You could slow an INTEL chip down to 1MHZ and Skype would still try 10-way call, and we all know AMD CPUs are vastly faster than anything INTEL can offer. In this CNET test, the slowest AMD CPU is faster than the fastest INTEL CPU. INTEL is losing market share in all segments, and INTEL's stock crashed to half price of AMD's.

INTEL has used the CPUID trick before, according to AMD's lawsuit, some INTEL software deliberately cripples AMD CPUs when detected using the CPUID operation. There are also reports that once you modify the INTEL software by removing the check for CPUID, then AMD CPUs run the software much faster.

What we see in the INTEL-Skype case is different. This time INTEL is not meddling with its own software, but the most popular VoIP software from a third party. This is clearly intended to block AMD from the VoIP market, and give INTEL a monopoly power in that market. I am quite sure that such plots are part of the VIIV programme and Eric Kim is behind it.

This kind of behaviour is criminal and must be stopped right now.

If the INTEL-Skype dirty plot is allowed to continue, Microsoft can legally add this code and kill anyone immediately:

if(this is something I don't like, such as CPU, Browser, Anti-Virus software...) {
then: sorry, you can only open 5 windows

}else {
open as many Windows as you want


INQ needs to write what it is good at

Charlie Demerjian at INQ might be good at assembling PCes, but he shouldn't pretend to be a PC market analyst. AMD capacity is the least of all concerns for DELL.

Most idiots have an overblown impression of DELL. DELL has no 30% market share, moron! DELL is only slightly higher than HP in unit share. DELL ships about 10 million PCes per quarter. According to IDC, DELL's global market share was 17.2% in 4Q05, HP's unit share was 15.7%, DELL is only 1.5% more than HP in units, HP is doing AMD all the way. You got to have some common sense.

Demerjian's analysis based on AMD's share in non-DELL market is flawed, the dynamics is much more complex than that--DELL entering AMD business will definitely cause AMD to slow its growth in HP and whitebox market. But even if we took his approach but with the right numbers, we have

1) AMD's share in non-DELL market = 21.4/(100-17.2) = 25.8%

2) Assume after DELL-AMD deal, AMD will take the same share off the DELL pie, AMD will get 25.8% * 17.2% = 4.43%

3) Total AMD share after DELL deal is thus 21.4%+ 4.43% ~ 26%

Now tell me if FAB36 and Chartered FAB7 can supply the chips.

I also told you why DELL will spit at INTEL and kneel to AMD soon.

Sunday, February 12, 2006

Socket F Opteron to be released in 2Q06

This page lists the spec for Socket F (1207) Opteron obtained from AMD's presentation at the ISSCC conference, die size 220 mm^2, frequency 2.6GHZ, max power 95 watts. AMD said there is a 7% frequency margin, indicating initial speed can go up to 2.8GHZ. AMD also claimed that by dropping voltage from 1.35v to 1.1v, there is a 58% reduction of leakage current, however, frequency drop is only 15%, this means AMD can make a very low power version of the new Opteron. You can see an image of the Opteron core here. Pay attention of the size of the cache. Now consider that INTEL has to keep doubling their cache sizes.

Saturday, February 11, 2006

INTEL next generation server CPU stuck at 32 bit

Recent reports showed that INTEL is readying its next generation Sossaman Xeon LV built on a 65 nm process. SuperMicro proudly announced support of this leading edge 32 bit IA32 technology based on 1995 Pentium Pro in an era of pervasive 64 bit computing.

Back in 2003, four INTEL teams had studied doing 64 bit on Pentium and concluded that was not possible. After the release of AMD 's 64 bit Opteron, reverse engineering the AMD's 64 bit instructions has been proven a tough job for INTEL engineers. INTEL's first attempt was to emulate AMD64 on Pentium 4, as Pentium 4 has 36 bit physical address and is easy to pretend to be more than 32 bit. AMD Opteron has 40 bit physical address and allows 1TB of memory, the CPUID instruction on the Opteron is set to report 40 to the OS, so the OS knows the capabilities of the CPU. Although Pentium 4 and Xeons have only 36 bit address, INTEL engineers blindly copied AMD's instructions, thus INTEL's clone also reported 40 bits, instead of 36. This caused 64 bit version of Windows and Linux to crash miserably on INTEL. As of today, INTEL's so called EM64T on Pentium 4 is still missing crucial features in AMD64 and performs slower in 64 bit mode than 32 bit mode.

Since INTEL is set to terminate all Pentium 4 based CPUs and will base everything on the Merom architecture. It has to start the cloning of AMD64 afresh. Since Merom is based on Pentium Pro core , and is physically 32 bit (unlike Pentium 4's 36 bit), emulating 64 bit on P3 is proven to be even harder than doing it on Pentium 4.

INTEL has to comfort customers and analysts by claiming that two 32 bit INTEL cores together is better than one 64 bit core, as if two 286s (16 bit) are better than one 486(32 bit). Although INTEL has promised 64 bit support for the Merom architecture set to release in 4Q06, its continuing deemphasis on 64 bit on the eve of Windows Vista raises concerns that INTEL's implmentation may have compatibility issues with the x86_64(or AMD64) standard.

In any case, running a 2P server with Sossaman32 CPUs inevitably creates a substantial imbalance between CPU power and memory capacity: 4 high speed 2GHZ Cores fighting for a maximum of 4GB memory is not a pretty sight in any server configuration. The situation is especially dire for Sossaman32 users who run memory demanding processes such as J2EE, JSP, .NET, ASP, PHP, LAMP web applications, database servers, concurrent transactional apps and other data intensive applications. The shared 667MHZ front side bus provides a meager 166MHZ average bandwidth to each Sossaman32 core for memory and I/O, which is another major bottleneck.

AMD exited 32 bit mainstream market long ago. Its 32 bit product line is only for embedded market. The AMD GeodeNX is actually a 32 bit K7.

AMD demoed quad-core Opteron in 2005

One old INTEL marketing strategy is to promise pie in the sky vaporware to keep customers interested, then keep pushing the delivery time back again and again and again. INTEL did this on Itanium, we heard all these marvelous stories about dual core 2GHZ Itanium 2 years ago, and the story repeated with delivery time changed over and over, now INTEL is pumping 1.6GHZ version. INTEL did this on the 4GHZ Pentium 4, until it was finally cancelled. INTEL did this on Merom, first it's 2H06, people was expecting it in mid 2006, then 3Q06, then 4Q06, I would not be surprised if INTEL pushes this again to 1Q07---INTEL's engineers simply don't have enough skill to copy AMD64 instruction set properly, as we have seen in the past. INTEL said two 32 bit Core Duo is better than 64 bit, which was like saying two 16 bit 286 was better than one 32 bit 486 -- a convenient excuse for their failure to copy AMD64.

AMD demoed quad-core Opteron to its customers back in October 2005. Major server vendors such as HP, Sun, IBM and Rackable had seen the quad, and motherboard makers such as Tyan and SuperMicro had seen it too.

Furthermore, the quad-core Opteron will run on existing Socket 940 platform for single core and dual core Opterons, offering a painless upgrading path for Opteron users today.

INTEL's so-called multi-core is actually a primitive multi-die kludge, which is basically multiple CPU dies tied to the same shared bus.

Friday, February 10, 2006

Why should AMD move to China 100%?

US government only likes INTEL, its procurement specifies INTEL brand, in other words, US government explicitly excludes AMD(and others). US FTC ignores AMD's complaints, DOJ silently endorses INTEL's monopolistic actions by doing nothing to stop it. AMD is fighting for its survival. This is the situation 3 years after the introduction of Opteron and AMD64.

AMD should sue US government for illegally excluding AMD.

US government has large deficit and huge debt, it's on the brink of bankruptcy.

China has an annual trade surplus of $200 billion with the US. China has a foreign reserve of $800 billion. Chinese may not be rich, but their government is well managed financially and has mega tons of money and they spend lavishly on hi-tech gear.

If AMD moves 100% to China, its revenue will jump $1 billion instantly. Chinese government will immediately specify AMD brand for all government purchases, INTEL will be excluded in China

*) The millions of Chinese government employees will buy 2P dual core Opteron workstations

*) The 3 million strength People's Liberation Army will only buy AMD64 and say no to INTEL 32 bit or EM64T copycat, millions of Opterons there. 10x more purchasing power than any Google.

*) All Chinese schools and universities will buy AMD64 only. That's huge.

*) All those countries which don't like America (such as most Europeans) will buy more AMD and spit at the American bully INTEL.

*) Once US found China's AMD performs better than INTEL, it will dump INTEL too.

*) If US government continues to exclude AMD, then a China based AMD can invoke WTO rules to demand fair trade.

*) ..................................

The world will be a better place, as we move to pervasive 64bit computing.

Thursday, February 09, 2006

Raid or visit?

There was a discussion on whether INTEL was raided or visited by Korean FTC. AMD said it was a raid. INTEL said it was an unsheduled visit during normal business hours.

Let me take out the dictionary,Websters' College Dictionary 4th Edition.

Raid: any sudden invasion of a place for discovering and dealing with violations of law.

Visit: to go or come to see

INTEL folks, what Korean FTC did fits the first definition perfectly. Officials stormed in, probably armed, and ordered you to turn in boxes of confidential documents for investigation of illegal dealings. That's a raid by definition of the English dictionary. There was no force used, but I bet the Korean officials were prepared to use force if your folks refused to cooperate.

SUN drives me crazy

Another JXXX. It was SeeBeyond, now it's JCAPS, yes, very creative, it's the catch all, J followed by all CAPS. If Scott can't remember all the JXXX trademarks, JCAPS summarizes them all. I think SUN can turn profit by just cutting the attorney fees for registering these JCAPS.

I challenged SUN to create a Solaris 10 control panel similar to Windows server 2003 control panel long long time ago. I still don't see it coming. Guess it is super hard. Now I know why Bill Gates is smart and Microsoft is making money. You get paid by selling products, you don't get paid by giving away technology. Microsoft is selling products, even though they are crappy; SUN is giving away technologies, even though they are marvelous. Windows Server 2003 with a GUI control panel is a complete and usable product; Solaris 10 without a control panel is a piece of cryptic technology that few can take advantage of.

Why hasn't US FTC raided INTEL yet?

Does INTEL own the US government? The Japanese had raided INTEL offices and uncovered a lot of incriminating evidence; the Euorpeans had raided INTEL and DELL offices. Now the Koreans have raided INTEL.

In contrast, the US government is paying $563 million more of tax payers' money by illegally specifying INTEL brand in its computer purchases. This is state sponsored iNTEL monopoly, pure and simple.

If you compare the DOJ's Microsoft case and the AMD's INTEL case, what INTEL has done is 10 times more severe. Why aren't FTC and DOJ doing anything, despite the continuing behaviour such as excluding AMD CPUs from Skype 10-way conference call?

INTEL is using its monopoly power to keep us in the dark age of 32 bit computing.

I think AMD better move to Europe or China and break free from the illegal monopolistic control of the CPU market.

Wednesday, February 08, 2006

INTEL to stop AMD with Skype exclusive deal

Just a few days ago, I cautioned that INTEL may try other tricks. To my surprise, INTEL has quickly come up with another idea to block AMD's road to fair market access, Intel has worked with Skype to programmatically limit the functionality of AMD64 CPUs.

According to CNET News:

Skype 2.0 software will allow 10-way conference calls only if it detects code specific to Intel's chips when the PC boots, said Rob Crooke, vice president of Intel's Business Client Group. Intel approached Skype with its plan to optimize code on its chips for Skype's software so users would have a good experience while hosting a multiperson conference call, Crooke said.

Specifically, if Skype 2.0 detects a 32 bit INTEL Core Duo CPU or an EM64T thermal throttling P4 D, then ten (10) way conference call is allowed; if an AMD CPU is detected, sorry, Skype only allows 5-way conference call. Put it simply, AMD is given only 50% of the Skype conference call quota, even though AMD64 CPUs are five generations ahead of any of INTEL's offering and AMD64 CPUs frag INTEL day and night.

In response to a ZDNet report that "Skype 2.0 looks like a virus", I made an educated guess that Skype developers were using obsolete INTEL CPUs incapable of NX bit virus protection found in AMD64. I now know I was right. Skype and INTEL are in bed.

Shame on you, cheap Skype Providers! Shame on you, INTEL Crookes!

Dutch providers, this is America, not the redlight streets of Amsterdam.

AMD should amend its lawsuit to include this.

EBAY should fire the Skype execs.

We all should use better VoIP solutions, Google's American technology is much better, even MSN Messenger does a better job.

Incidentally, Forbes wrote "Whitman's Skype, Intel Up Ante On Advanced VoIP", so now people know, Meg Whitman favours INTEL. But I heard that eBay's Xeon servers are running into heat problems.

Tuesday, February 07, 2006

INTEL Xscale still at 180nm process

So much for a self-claimed leader of semi process, or, maybe it's the beauty of "copy exact"?

SUN's Galaxy

SUN has made major price cut on its Galaxy servers. Looking at the highest configuration, it costs only $5895, and you get
*) 2x Opteron 275 2.2GHZ dual core CPU
*) 4x 1GB DDR ram
*) 2x 10K RPM 73GB SAS drives
*) 4 GbE ports
*) Dual redundant power
*) 1x Service processor for remote management
*) Integrated Lights Out Management
*) Solaris 10 + JES
*) 3 Year Warranty & Parts Exchange

Try go to dell.com and configure an INTEL dual core server, you pay over $8000.

Not to mention that the Opteron is about 2X the performance of Paxville Xeon in Apache HTTPD benchmarks.

Also according to the newest SUN announcement, it has already shipped tens of thousands of "Galaxy" Sun Fire X4100 and Sun Fire X4200 servers. However, according to this analysis based on SUN's last quarter results, it had only shipped 8000. So SUN must have sold a lot more Galaxies in the last month since the end of last quarter.

Beating DELL on price also? Now, that's ass kicking.

My hats off to SUN folks.

AMD may release quad-core Opteron HE soon

According to this MSI roadmap , thin & light dual core Turion 64 notebook will be available in 2Q06 with 25 watt TDP. This indicates that AMD should be able to release HE version of quad-core Opteron at 55 watts soon.

MSI is also releasing Socket F Opteron boards supporting DDRII up to 533/667/800MHZ. The MSI K9SD Master to be relased in June 2006 features 16 DDRII slots. In comparison, the INTEL Bensley/Dempsey/Blackford only support DDRII 533/667 FBDIMM.

Monday, February 06, 2006

INTEL: 2 x 32 > 64

INTEL had a technology conference call on the 32 bit Core Duo processor with revolutionary INTEL(R) Deeper Sleep Technology(DST). Admittedly, I almost fell into deep sleep while listening to the CC, but bursted into wild laughter when I heard Mooly Eden, Intel VP of the mobility group, said something I considered very funny.

What Eden said was basically that "Two 32 bit cores are better than one 64 bit core", apparently refering to Turion 64.

No, no, no, Mooly, 2x32bit is only 33 bit! You know that, right?

Then our smart Ben Lynch surprised us techies with a deep question on "relative deficiencies in INTEL interconnects and scalability for multi-processor systems or multi-core CPUs", and how INTEL is going to address that with the next generation Merom architecture. While Molly was fumbling for an answer or excuse, another INTEL guy cut in and said this kind of server question should be defered to next IDF. But Ben Lynch pressed on, saying interconnect efficiency is also pertinent to dual core mobile. Cornered, the INTEL folks danced around, talking about balance of battery life and performance.

One analyst asked why users would need dual core mobile. Mr. Eden spent a long time and came up with all sorts of multi-tasking mobile computing scenarios, such as playing doom3 and ripping mp3 at the same time, zapping a virus while editing a spreadsheet, etc, etc.

Sunday, February 05, 2006

Beware of INTEL's tricks

I was reading the book "Inside Intel" by Tim Jackson. It's an eye opener to see how INTEL got rich. For example, in the early days of 8086, to eliminate Motorola, which had the 68000 CPU, INTEL initiated a marketing compaign called "Operation Crush". The goal was, quoting the leading INTEL exec, "We have to kill Motorola, that's the name of the game. We have to crush the f**king1 bastards. We're gonna roll over Motorola and make sure they don't come back again." One key INTEL tactic was the release of cataglogs of next generation CPUs 5x the speed to convince customers to choose INTEL. Such parts did not even exist on the drawing board. Motorola was fooled, instead of dimissing INTEL's vaporware, it released an honest catalog of next generation CPUs which were far less impressive.

Most disturbing was how INTEL deceived the Court and jury during an INTEL lawsuit against AMD on the issue of 286 microcode. It was about the interpretation of the term "microcomputer" in an INTEL-AMD licensing agreement. INTEL altered the date of some crucial internal document and removed some incriminating text from it. The jury was deceived, and issued a verdict against AMD.

It was by some pure luck that AMD found the original INTEL document and the judge ordered a new trial four months after the jury verdict. With the untampered evidence, AMD won. But AMD lost precious many months marketing its own 486 processors due to an injunction issued previously as a result of INTEL's lawsuit.

The lessons AMD should learn:

*) INTEL is again promising pie in the sky 45nm and next generation chips no one has seen. AMD should devise a good strategy against this and seize the moment. Platform stability is a key AMD advantage.

*) Beware of potential tampering of evidence in the anti-trust lawsuit. AMD needs to beef up its computer forensic expertise in its legal team. I think Google folks should be very good at mining the massive amount of documents and emails.

Note: The book had the four letter word + ing, I removed it, so children don't learn from the bad example.

Saturday, February 04, 2006

We all live in the aftermath of 1GHZ Athlon

Condensed history of pervasive AMD64 computing

0) Feb 1999: INTEL Pentium 3 - with stuff super scalar and out of order execution borrowed from RISC

1) Jan 2000: 1GHZ Athlon - with the goodies

2) Aug 2000: INTEL Pentium III at 1.13GHZ rushed out but recalled, INTEL found it impossible to push P3

3) Nov 2000: INTEL Plan B, 22 stage hyperpipeline Pentium 4 jacked up frequency to 3GHZ later

4) 2001: AMD abandoned the frequency fight with Athlon XP, but worked on 64 bit Hammer day and night

5) 2002: Meanwhile, INTEL marketing almost drove AMD to extinction using exclusive deals, AMD stock fell to $3

6) Jan 2003: INTEL dudes got over zealous, aiming 10.2GHZ Nahelem in 2005, 20GHZ to 100GHZ later

7) Feb 2003: Four Intel teams failed to do 64 bit on x86, and they didn't believe AMD could do it

8) Apr 2003: AMD came out with the 64 bit Hammers (AMD64), with better performance and perf/watt , pulled down version for desktop the Athlon 64 was released a bit later

9) Nov 2003: AMD start building new FAB

10) Feb 2004: INTEL execs got desperate, P4 pipeline jacked up to 38 stages, but no more performance, only more heat

11) Oct 2004: INTEL found 4GHZ impossible, silicon was melting. INTEL CEO admitted blindly racing at 200mph without noticing the brick wall ahead.

12) June 2005: INTEL does more exclusive deals, AMD hauled INTEL to Court

13) Sept 2005: AMD64 sales jumped up instantly

14) Jan 2006: INTEL goes back to 32 bit P3, but changed the name to Core (Core Duo, Core Solo)

15) Today: AMD cranking out AMD64 chips like crazy

Friday, February 03, 2006

Fan boys ask world to wait for INTEL

Hans Mosesmann at Moors & Cabot are asking people to wait for INTEL's Woodcrest CPU, which is expected to be out in 3Q06. He claimed that Bensley (Dempsey CPU + Blackford chipset) is being shipped for testing in 1Q06 (now) and will be out in volume in 2Q06, but then the socket compatible Woodcrest CPU can replace the Dempsey CPU in 3Q06.

Does Hans Mosesmann think people are stupid? First, in Q1, they buy INTC Paxville, which is too dumb, too slow and too hot; then in Q2 they buy INTC Dempsey the Athlon MP imitator, which is slower than old Opteron 246 in some tests, and then 3 months later, they find Dempsey too hot and "pull in" INTC Woodcrest, which is supposed to be cooler than Dempsey. If people were indeed stupid as Moors & Cabot liked to be, then INTC would make the same money three times in a row, leading to some $1.7 EPS.

The good news for the shrewd server buyer (such as Marc Andreessen) is, AMD will release quad-core Opteron in mid-2006. The fact these quads can plug into existing Socket 940 boards means their power consumption is no more than 95 watts. There are no surprises here. The Turion MT-40 is a 2.2GHZ CPU with max power of 25 watts, 4 of these make 100 watts. AMD should have no problem making a quad-core 2.2GHZ Opteron at 95 watts.

The choices for Opteron buyers are clear. They can buy single or dual core Opterons now with power of 30 watts (Opteron EE), 50 watts (Opteron HE) or 95 watts. Then they can upgrade to quad-core Opterons later with a simple BIOS flash.

No wonder that 90% of Rackable's revenue comes from Opteron, and it's no surprise that SUN could snatch a $400 million workstation deal with Opteron at the last minute of the bidding process.

Hans Mosesmann's FUD on Socket F Opteron and HyperTransport is pitifully laughable. Let me assure you this, the Socket F Opterons will be so advanced and strike so hard, INTEL will be beaten into guacamole.

BTW: Look at this score card, zero stars on seven of of the eleven companies covered. A monkey could have done a better job by throwing darts.

Analyst Hans Mosesmann, with analytical accuracy seemingly less than that of a dart throwing monkey, wrote with a sense of desperation,

On 2006 outlook: "AMD Valuation. Our 12-month price target of $10 for AMD shares is based on a 20x multiple to our 2006 EPS estimate of $0.50. The 20x multiple is an 11% premium to the company's historical 18x P/E ratio and is actually the same multiple we are applying to INTC shares, which is justified by our view given that AMD is a significantly better managed company today, under CEO and Chairman Hector Ruiz."

On AMD lawsuit: "[T]he timing of the complaint and the company's tone of indignation (perhaps exasperation) give us a sense that AMD's traction in processors that they enjoyed last year is not meeting expectations"

"It strikes us that AMD's approach is based on throwing everything but the kitchen sink at Intel in the hopes of getting something to stick."

Angry man, how is your $12 AMD short doing?

Come on, SUN, can you beat Rackable?

Rackable reported yesterday that its 4Q06 revenue was $83.1 million, with over 90% of that revenue came from its Opteron server business, less than 10% from INTEL server business. During the conference call, Rackable commented that SUN does not pose a competitive threat, because SUN's Galaxy (also Opteron) product line is incomplete, but they expect SUN to release more Galaxy models.

SUN folks, wake up!

RACK is eating the Opteron pie, and DELL is eyeing it.

Thursday, February 02, 2006

Gateway should go 100% AMD

Gateway reported 4Q05 results with $0.06 profit per diluted share. The company sold 1.359 million PCes, of which 1.134 million were sold in the retail market. Gateway stock price is at less than $3.

My suggestion to Gateway: go with the leader, go 100% AMD.

AMD will give some nice discounts for sure, and Gateway can save the cost of dealing with the outdated INTEL technology, such as the costly BTX platform for extracting the 150 watts of heat from Pentiums.

Best PC processor can't run Windows x64?

Bill Gates touted his spanking new Windows x64 in 2005, but the best PC processor can't run it.

Sorry, Bill, your OS is for AMD64 only.

Wednesday, February 01, 2006

SGI CEO fired for buying INTEL

Silicon Graphics (SGI) used to be a big name in IT, it was the leader in computer graphics, movies such as Jurassic Park was made with SGI technology. I used to write 3D animations on IRIX workstations using GL. At some point later, SGI made the total switch to INTEL's Itanium. Betting the farm on something INTEL pumped billions into couldn't go wrong, right?

Nope. With Itanium struggling and falling far behind Opteron, SGI's business went down in a quick spiral. It was delisted from NYSE and traded on pink sheets.

With product revenue dropping 50% year over year, SGI finally fired its CEO for 6 years of failure. The blame was on INTEL not delivering a better Itanium chip, but who made the decision to buy INTEL took the hit.

I have a proposal to save SGI and I am giving it away for free.

OpenSolaris 10 on Opteron is the way to go.

AMD64 is five generations ahead of INTEL

It's impossible for me to cover this topic in great detail, so I will hit the key points only.

AMD64 Instruction set

In Feburary 2003, on the eve of AMD's launch of the AMD64 family CPUs, INTEL expressed its disblief. According to Richard Wirt, an INTEL senior fellow, four separate design teams at Intel had examined how the company could take one of its 32-bit chips and transform it into a 64-bit machine, all four Intel teams concluded that such a feat was not doable.

INTEL did try hard to do 64 bit on x86, but their engineering didn't know how.

But the grand masters at AMD did what INTEL thought was impossible. Opteron 64 hit the market in April 2003 and quickly won almost all performance benchmarks.

Seeing is believing, INTEL tried to reverse engineer AMD's instruction set onto Pentium IV and Pentium 4 based Xeon. Emulating AMD64 instruction set was easier on Pentium IV, because it had a 36 bit physical address. However, benchmarks show INTEL's EM64T runs slower under 64 bit mode than 32 bit mode. Moreover, INTEL used some old AMD PDF files, and did a bad job, some Microsoft and Linux code developed on AMD64 failed on run on INTEL's clone. As of today, INTEL's EM64T is still missing some crucial capabilities of AMD64.

But running AMD64 instructions on Pentium III proves to be much harder, as of today, INTEL hasn't yet figured out how to do 64 bit on Pentium M and Core Duo.

And AMD is not sitting idle, it's adding a new set of instructions to the AMD64. INTEL engineers will have more sleepless nights digesting AMD PDFs.

True Multi-core

AMD64 architecture was designed to be true multi-core from the ground up. A multi-core CPU is much like a multi processor system, the cores must communicate with each other to maintain consistency. Inside the AMD64 CPU, there is a crossbar switch that connects the multiple cores together, so they communicate internally and at extremely high speed. We see from benchmarks that dual core Opteron is almost twice as fast as a single core Opteron at the same clock speed.

In comparison, INTEL's dual core implementation is a kludge. In INTEL's design, the two cores share the same FSB, when they need to communicate, they first go out to FSB and come back again, without knowing they are sitting next to each other. The result? Poor performance .

This AnandTech article provides good explanation of the dual core designs.

The Embedded Memory Controller

Chip design gurus have long realized that a major bottleneck in system performance is memory latency. Just like memory is much faster than hard disk, the CPU is much faster than memory. When a CPU needs to access memory for instructions or data, it has to wait for the memory content to be retrieved, the time of waiting is the latency. During the waiting period, the CPU can't do anything.

In the old FSB based architecture (all INTEL's), the memory controller is in an external chip called the north bridge, while the CPUs run at 2-3GHZ, the conventional memory controller runs at about 200MHZ. Furthermore, in the old FSB design, the data have to make two hops, from memory to memory controller, then to the CPU. As we can see from this article, memory latency in a Pentium 4 design is between 300 to 400 clock cycles.

In AMD64 design, the memory controller is embedded in the CPU and runs at CPU frequency, the CPU connects directly to the memory without any intermediate. As we can see from this IBM test on single and dual core Opteron, memory latency on the Opteron is only about 50 nano second for local memory access.

Like the Opteron, all modern CPUs, such as Alpha EV7, IBM Power5, SUN UltraSparc T1, AMD Geode LX, Athlon 64, Sempron 64, Turion 64, have embedded memory controller(s).

From INTEL roadmap as far as 2009, we don't see an embedded memory controller design.

Cache Coherent HyperTransport (ccHT)

In a N processor AMD system, since each CPU has its own memory controller and associated banks of memory, there are N memory controllers which provide N times the memory bandwith. To have these N memory controllers act coherently, there are multiple ccHT links between AMD CPUs, which is used for fetching memory from another CPU. As we can see from the IBM document referenced above, in the case of remote memory access, the latency is also quite small.

INTEL is rumored to work on something similar to ccHT called CSI, however, since the cancelation of the Whitefield project, CSI is missing from INTEL's foreseeable roadmap.

Direct Connect Architecture

In FSB based architecture such as INTEL's, the CPU, Memory and I/O share the bandwith of a uni-directional bus, just like many folks share one phone line in a conference call --- only one guy can talk in either direction. In AMD64 architecture (Opteron, Athlon 64, Turion 64, Sempron), there are separate dedicated connections between CPU and Memory, between CPU and I/O, between CPU and CPU, between CPU core and CPU core. In AMD64, there is no crosstalk, and everything is bi-directional--traffic goes both ways the same time.

From INTEL's roadmap, it's stuck with FSB architecture until at least 2009.


INTEL is 5 generations behind AMD, and there are other major areas that INTEL is lacking, such as IOMMU for fast DMA. To match AMD in 2 core performance, INTEL will have to use very large cache size, which will negate its shrink to 65nm. At 4 core and up level, INTEL is simply hopless.