Wednesday, September 27, 2006

IDF shows Intel out of steam

Can Intel avoid bankruptcy? That's the question we all want to ask.

The answer from IDF is NO.

There is nothing new shown in the IDF on the eve of AMD's K8L flood. Those teraflop floating point device is not going to save Intel-- clearspeed has such chips, IBM's Cell is equally promising.

Intel touted that Woodcrest Xeon are now 50% of their 2P volume. This means 50% of their 2P server chips are sold below $200. Furthermore, since 2P Woodcrest is faster than 4P Xeon, it will be very interesting to see what Intel's share of the 4P mainstream market is.

Judging from Intel's IDF presentations, they are still trying to gain back market share--which is 100% impossible now that DELL is flooding the market with $399 dual core AMD PCes, replacing the dying Netburst.

Intel shipped 5 million Core 2 Duos in two months. This was after Intel allegedly stock piling production since March. The most optimistic estimate we can have is Intel will ship 10 million Conroes in the next three months, with the other 40 million being legacy 32 bit or Netburst chips. AMD will simply take the 40 million market as much as it can.

The result?

In the coming quarters, Intel will suffer massive losses due to lower unit share and drastically lower ASP.

49 Comments:

Blogger Scientia from AMDZone said...

Can Intel avoid bankruptcy? That's the question we all want to ask.

The question is if it would even be possible that Intel could go bankrupt while AMD is still unable to manufacture 50% of the chips. The answer from common sense is no, unless Intel for some reason is unable to reduce costs.

Those teraflop floating point device is not going to save Intel-- clearspeed has such chips, IBM's Cell is equally promising.

Well the difference is that Cray has a contract to build a PetaFlop Opteron machine and IBM has a contract to build a hybrid Cell/Opteron machine. It would be interesting to see if Intel could build an entire supercomputer out of lightweight processors. IBM is not doing this with Cell which is lighter weight than the embedded Power chip it used for Blue Gene. I don't believe that a processor that was lighter weight than Cell would work unless it were in a hybrid system.

Judging from Intel's IDF presentations, they are still trying to gain back market share--which is 100% impossible

It's 100% impossible because of simple math. Intel will only hit 25% volume with Conroe by year's end. The only way that Conroe could take share would be if Intel didn't lose any P4 share and that is extremely unlikely. It is much more likely that K8 will take P4 share.

Intel shipped 5 million Core 2 Duos in two months.

Intel makes 3.5 million processors a week. 8 x 3.5 = 28 million. It's actually 30 million due to a quarter's being 13 weeks (instead of 12). This is only 17% conroe by volume.

This was after Intel allegedly stock piling production since March.

They didn't stockpile Conroe.

In the coming quarters, Intel will suffer massive losses due to lower unit share and drastically lower ASP.

Well, not losses, exactly. Their revenues will drop from what they were in 2005 but they will still have profit. They could end up having to reduce stock buyback to balance the budget though.

11:04 AM, September 27, 2006  
Anonymous enumae said...

Sharikou said...

"Judging from Intel's IDF presentations, they are still trying to gain back market share--which is 100% impossible now that DELL is flooding the market with $399 dual core AMD PCes, replacing the dying Netburst."

Could you link to that $399 dual core system.

The cheapest dual core AMD system I found was $869.

Or this one $919.

But I did find this one, and at $990, it is way better than what AMD is offering it comes with a 19" monitor and a 250GB hard drive vs a 17" monitor and a 160GB hard drive.

$40 cheaper when equiped the same (monitor and hard drive). Also the E6300 is better than the AMD Athlon64 X2 Dual-Core 3800+.

11:14 AM, September 27, 2006  
Anonymous Anonymous said...

Yawn. Explain to me how $399 dual core AMD-based PCs are good for AMD?

Good for Dell- sure.
Good for customer- absolutely.
Good for Intel- well, no.
Good for AMD- definitely not.

Calculate the BOM on that $399PC, making sure you keep a profit margin in there for Dell. Then tell me that AMD is doing much more than breaking even at best. Shipping die wrapped in money is not the best business plan...

11:41 AM, September 27, 2006  
Anonymous Anonymous said...

You know I did have a response to post, but what's the point. WTF is wrong with you man. Get a hold of yourself, intel is not going bk and dell is not going to help amd much by buying $30 processors for their $300 crappers. Get off the pc and get some fresh air!!!

12:02 PM, September 27, 2006  
Anonymous Anonymous said...

E521 dell with no monitor, an AMD x2 3800+, 512MB RAM, 160GB HDD = $649

Thats the lowest price on a dual core from dell

12:13 PM, September 27, 2006  
Anonymous Anonymous said...

The legal proceedings, quad core, Geneseo - all in a 48 hour stretch. Dude, you are late to the game. Intel has been sleeping, but has been awoken.

Sorry, AMD even with all they have done is once again going to play second fiddle.

I am so glad they are still alive after all these years. Without them, we would have been stuck with crap processors forever.

Now, if you are such a smart lawyer, engineer, Ph.D etal, please find a way to marginalize Microsoft.

Now that would get your blog churning.

12:21 PM, September 27, 2006  
Anonymous Anonymous said...

normally I disagree with Sharikou on principle, but I do have to admit, Dell is cranking out some very cheap X2 computers, in fact he's right that it can be gotten for $399 before taxes (if applicable). The thing is, you have to look under Small Business computers, choose the X2 base system WITHOUT THE MONITOR. Now, take that as you will, but at least it's honestly sub $400, which can be substantiated. here, build it yourself:
http://configure.us.dell.com/dellstore/config.aspx?c=us&cs=04&kc=6W300&l=en&oc=bdcw5bz&s=bsd

12:26 PM, September 27, 2006  
Anonymous Dr. Yield, PhD, MBA said...

Back to the BOM of the $399 3800+ system. Estimates are from cheapest part on Pricewatch that meets the spec (generics included), and in some cases discounted slightly. Also, I would think that the system would benefit from something better than a single DDR2533 DIMM with what I assume to be poor timings:

Case + 300W PS: $30
MB with GF6150: $70
Cables, power cord: $ 5
512MB DDR2-533: $45
WinXP Home: $50
MS Works 8.0: $15
160GB SATA HD: $50
16X DVD-ROM: $15
USB mouse + KB: $15 (5+10)
Total so far: $295

Add in a 1 year warranty reserve as required in accounting at an estimated $20, we are now at $315. Shipping is free, $30- now $345. 10% profit: $35, $380.

Crap- only $19 left to buy my 3800+ Let's say Dell cuts their profit to $15, now you get $39 for that die. Let's hope AMD can really make die for less than $40 and turn a profit on the deal- cause based on this, Dell has AMD's shorts around their ankles. As I said above (sorry, anon that time), $399 dual core systems are not good for AMD. Because even if they can make 'em that cheap, there are all the distribution costs as well. Remember the mantra of the internet bubble- "If I lose $5 on every sale, but sell millions of them, I'll lose... never mind. Where's my IPO?"

Let's hope for AMD's sake that they are getting better pricing on their higher end parts that they are selling to Dell, because if they aren't, it won't be pretty at end of quarter time.

2:37 PM, September 27, 2006  
Anonymous Anonymous said...

Sharikou, can you comment on this thread.

"Conroe performance claim being busted"

http://sharikou.blogspot.com/2006/04/conroe-performance-claim-being-busted.html

Oh and how many more quarters until intel is bankrupt again??

3:03 PM, September 27, 2006  
Anonymous Randy said...

Huh, what do you know, $399.
Looks like Sharikou didn't just come up with that number out of his ass. Kinda makes you wonder that maybe he's not just talking out the side of his neck. Can't say the same for all his detractors here though.

By the way, I can build an x2 box with similar specs for similar price and no I don't get their cpu's for free, neither does Dell.

3:12 PM, September 27, 2006  
Blogger Sharikou, Ph. D said...

Case + 300W PS: $30
MB with GF6150: $70
Cables, power cord: $ 5
512MB DDR2-533: $45
WinXP Home: $50
MS Works 8.0: $15
160GB SATA HD: $50
16X DVD-ROM: $15
USB mouse + KB: $15 (5+10)
Total so far: $295



I bought a MB with ATI graphics and X2 3800+ combo at Frys for $159. The X2 3800+ was shown to be $109 and the MB was $50. The MB is better than DELL 521, it has 4 SATA ports and has i1394. I then bought a case+keyboard+optiocal mouse+power+speaker+power cord combo from newegg.com, total $30. Then I bought a Seagate SATA 300GB drive for $79.

You have to remember, the vendors above were still making profit. DELL has a lot of bargain power.

In any case, when AMD sells a chip at $100, it laughs to the bank with $30 profit. When Intel sells at $100, it loses $25.

3:48 PM, September 27, 2006  
Anonymous Anonymous said...

"In any case, when AMD sells a chip at $100, it laughs to the bank with $30 profit. When Intel sells at $100, it loses $25."

So now we are up to $125 for Intel to produce a CPU? That # keeps going up to fit your analysis.

What's also funny is you are now eben increaseing your previous estimates of AMD cost - didn't they used to be $40 according to you?

5:26 PM, September 27, 2006  
Anonymous Anonymous said...

"I bought a MB with ATI graphics and X2 3800+ combo at Frys for $159. The X2 3800+ was shown to be $109 and the MB was $50. The MB is better than DELL 521, it has 4 SATA ports and has i1394. I then bought a case+keyboard+optiocal mouse+power+speaker+power cord combo from newegg.com, total $30. Then I bought a Seagate SATA 300GB drive for $79."

OK so your total above is $268...what is your point?

Throw in a Windows XP, MS Works, a DVD ROM and some DDR2 memory. Win XP alone is how much?

For kicks maybe, I don't know, say throw in some labor costs, service, overhead at Dell, warranty and let's just suppose Dell is out to make some money too! (Also might want to assume while they are getting better prices on components they are not getting them at cost...)

While I doubt Dell is losing money, it's not like the $399system is a huge profit generator and I'm sure they've haggled down AMD on chip price below there standard published prices - as you stated so elegantly Dell "has a lot of bargain[ing] power"

Much better for AMD to get this product out from Dell before say Opteron 1P, 2P servers which clearly don't make as much money for AMD! :) Don't want to be wasting that limited CPU capacity on those high margin server chips!

5:37 PM, September 27, 2006  
Anonymous Anonymous said...

"By the way, I can build an x2 box with similar specs for similar price and no I don't get their cpu's for free, neither does Dell."

Randy - does your system include SW, support, warranty and are you charging yoursself anything for labor? If so maybe you should go into competition against Dell....

5:39 PM, September 27, 2006  
Anonymous Anonymous said...

Dr PhD pretender

Who is out of steam?

Lets see what is there to say..

INTEL released a revoluationary generation of chips; the win every benchmark against AMD by a longshot, are 40% or more faster then the previous generation

Ramping 65nm faster then ever, got 3 fabs on line and crossed over in volume from 90nm.

Nah INTEL got nothing.

For the first time AMD got no free side show as they got nothing to show, NO chips, no die shot even, nothing but a 50% price drop that priced them out of offering limo rides and free snacks..

The question is more why there was NO AMD freeloading at IDF... Its because they got NOTHING and can't afford ANYTHING.

7:45 PM, September 27, 2006  
Anonymous Anonymous said...

Now you all know why Intel had too many employees and huge R&D budgets. They kept bring out all kind of stop-gap, half-bake chips with cheap-tricks.

The engineering cost to bring these chips are expensive and their product life cycle is so short. Intel must be so desperate to resort to this.

Intel CEO boasted the non-native approach has a 20% higher yield thus bringing down the fix cost. Sure thing. But he did not mention the veriable cost (Engineering cost / number of chips sold over product life cycle).

I went to college so I know ok?

-Longan-

9:58 PM, September 27, 2006  
Anonymous Anonymous said...

OK so your total above is $268...what is your point?

Throw in a Windows XP, MS Works, a DVD ROM and some DDR2 memory. Win XP alone is how much?


dell, hp and others get substantially cheaper prices because they order in bulk. And OEM parts are not packaged like retail units which cuts-off a few more dollars. Also have a look at a hard disk brought off from retails stores and another from inside a dell (or hp, sony) m/c and note the diff in finishing quality,

S/w makers like Norton, google and others pay Dell some money to bundle their products (trial) on a dell m/c,

Add all these up and it makes up a substantial amount for Dell.

10:26 PM, September 27, 2006  
Blogger 180 Sharikou said...

IDF is not another nail in Intel's coffin. It is the second nail in AMD's following the launch of Core 2 architecture because Intel is now executing.

Now that your theory that Intel can't even sell 400k Conroes is washed out, you're resorting to vague and generic statements with no analysis.

AMD is heading to a GAAP loss by Q207. Intel will regain market share in Q406 if they haven't already gained some in Q3.

It costs Intel 40-50 bucks to make a Conroe. Starting at 183$ that's some sweet profit on 5 million units in 7 weeks even if you assume an ASP of 250$ on core 2 parts in Q3. AMD will be massively in debt once the ATI acquisition comes through in Q4. Can they service the debt if they are losing money. I don't think so. GAAP loss by Q207 my friends.


The pendulum is swinging the other way. Buy Intel - short AMD. You heard it first from me.

http://sharikou180.blogspot.com/
(A more balanced point of view)

10:41 PM, September 27, 2006  
Blogger duploxxx said...

180 sharikou....

it costs Intel 40-50 bucks to make a Conroe. ?????? did you forget the x0000 employees? oh yeah i forgot because of the huge stock they have they are payed with netburst cpu's.....since nobody else want's them anymore

by the end of this year k8 65nm is that cheap that they can sell a 5000 at a price of the e6300 and still make money, if intel doesn't go to 45nm fast they are doomed (not bankrupt.....)

saw the new x2 3600 already in benches its a killer for all the current netburst and even the new 915 high volume budget market (dual core do more crap)

you my friend are a joke, it looks like people aren't reading your blog so you drop al the crap here?

1:21 AM, September 28, 2006  
Anonymous Anonymous said...

It's 100% impossible because of simple math. Intel will only hit 25% volume with Conroe by year's end. The only way that Conroe could take share would be if Intel didn't lose any P4 share and that is extremely unlikely. It is much more likely that K8 will take P4 share.

Here's a thought:
40%-50% of both Intel and AMDs volume are low-cost processors (Celeron and Sempron). So if Core 2 is 25% of the production then only 25-35% of the volume will be P4 and PentiumD. Intel will have no problem 'getting rid of' roughly half as many Pentiums than they did last year IMO.

4:41 AM, September 28, 2006  
Anonymous Graham said...

Interesting... I am starting to see a trend in your analysis. If I want to make money on the stock market, I will just do the opposite of what you say Sharikou. While you are predicting bankruptcy, two investment firms UPGRADED Intel's 1-year stock target and the judge presiding over AMDs frivolous lawsuit thows out the majority of the case (thus showing just how frivolous the lawsuit is). You're track record is impeccably wrong. I expect your bankruptcy prediction to miserably wrong as well. Thanks though for being a beacon for what NOT to do.

8:50 AM, September 28, 2006  
Anonymous Graham said...

Oh and how many more quarters until intel is bankrupt again??

Well, Sharikou has been all over the map on this prediction. I believe he said somewhere near the beginning of Q2'06 that Intel had 6-8 quarters until bankruptcy. Sot that puts it at around Q3'07 or Q1'08.

I've tried to make his prediction interesting by making a wager on it. However, I realize a monetary wager is not that interesting plus I don't need money (made a lot by buying loads of Intel stock at the 17's)

So how about this for a wager... If Sharikou loses, he wears a pink bunny suit like Charlie from Inquirer @ IDF. The only twist is that he has to wear the bunny suit doused in gasoline while smoking a big cuban cigar. I can't imagine he'll go for it ... I wonder why?

No need to discuss what happens if Sharikou wins the wager. Why? Because he'll be wrong as always.

9:13 AM, September 28, 2006  
Anonymous Graham said...

snipped out some long-winded rant by Longan but had to keep this:
I went to college so I know ok?

-Longan-


Longan, I went to college also and here is what I know... Intel has sold MILLIONS of 65nm chips, AMD zero. Intel has 3 65nm fabs churning out chips today. AMD has 1 and they are trickling their chips. Intel has already demonstrated first silicon on a 45nm process. Intel has the best desktop processor in the world that squashes AMD in all benchmarks. Intel has the best mobile processors in the world. They run longer and cooler than AMDs dismal Turion mobile (still waiting for steamroller? keep waiting.) Intel has highly competitive 2P and 2S server solutions and will be shipping Quad core chips a quarter ahead of AMD.

My mommy loves me so I know ok?

Graham

9:34 AM, September 28, 2006  
Anonymous enumae said...

duploxxx said...

"oh yeah i forgot because of the huge stock they have they are payed with netburst cpu's.....since nobody else want's them anymore"

You are not looking at the general public, they are not gamers, or heavy photoshop users or even viedo editors. They are bargain shopers who need a computer to get online and read email.

Are you trying to say that a Intel Pentium D 915 can not do that, or even Celeron?

"By the end of this year k8 65nm is that cheap that they can sell a 5000 at a price of the e6300 and still make money, if intel doesn't go to 45nm fast they are doomed (not bankrupt.....)"

By the time Intel goes to 45nm all Celeron and Pentium chips will be using Core technology.

How are they doomed?

"Saw the new x2 3600 already in benches its a killer for all the current netburst and even the new 915 high volume budget market (dual core do more crap)"

I am wondering, if you are an extreme user, is your first choice a X2 3600+, or even a Intel Pentium D 915?

Again, those are mainstream, I do not know, but maybe you do, does the X2 3600+ open email faster than the 915, or browse the web faster?

9:55 AM, September 28, 2006  
Anonymous Anonymous said...

" Graham said...

snipped out some long-winded rant by Longan but had to keep this:
I went to college so I know ok?

-Longan-

Longan, I went to college also and here is what I know... Intel has sold MILLIONS of 65nm chips, AMD zero. Intel has 3 65nm fabs churning out chips today. AMD has 1 and they are trickling their chips. Intel has already demonstrated first silicon on a 45nm process. Intel has the best desktop processor in the world that squashes AMD in all benchmarks. Intel has the best mobile processors in the world. They run longer and cooler than AMDs dismal Turion mobile (still waiting for steamroller? keep waiting.) Intel has highly competitive 2P and 2S server solutions and will be shipping Quad core chips a quarter ahead of AMD.

My mommy loves me so I know ok?

Graham

9:34 AM, September 28, 2006 "

what kind of chips are these 65nm sir?
OHH RIGHT, ALMOST ALL ARE OLD NETBURST PENTIUMS BASED IN 65NM CORES!

10:13 AM, September 28, 2006  
Blogger 180 Sharikou said...

My dear Duploxx - now now, you sound like a petulant child who can't form an argument so decided to just shout instead. I can't make any sense of your comment. I've read it thrice and this is what I hear in my head:

"I hate you Sharikou 180...blah blah blah - did I mention I hate you!"

But let me try and respond. First on the Conroe cost. Using Sharikou's own over-simplified formula that derived Intel's manufacturing cost of a Netburst CPU at 100$ - I simply reduced that by half since the Conroe die size is 50-60% smaller than Pentium D. Now I will be the first to agree that this is not the right way to calculate this but if Sharikou can do it...I guess so can I.


No, I didn't forget the x000 workers at Intel. I also didn't forget Intel's 35 bln $ of revenue and their 800 million $ profit in a bad Q2.

I'm so happy for you the 3600+ benchmarks are good. However, my opinion remains that momentum will swing back to Intel starting Q4. For all your ranting and raving, let's remember that in spite of Netburst being a crappy product, Intel has sold millions of them so far. Because they have a brand called Pentium among other things. That brand is going to help them continue to sell at the low end. BTW - to prove my point the single core Conroes will still be branded Pentium.

I'm glad I amuse you. Because you my friend do nothing for me. Come on - say something interesting. I'd ask you to say something intelligent but I think it's better if you take baby steps first.

10:29 AM, September 28, 2006  
Anonymous Edward said...

"Intel has the best desktop processor in the world that squashes AMD in all benchmarks. Intel has the best mobile processors in the world. They run longer and cooler than AMDs dismal Turion mobile"

I really don't understand how people could use the word "best" so liberally if they had mildly capable intelligence.

At least one should say, Intel has the fastest desktop CPU assuming price is not a factor, on a 32-bit operating systems with limited memory size and possibly no RAID configuration.

Also, while there is ULV Core Duo but nothing like that for T64 X2, those ULV ones are slow & expensive. For the mainstream (i.e. 25~35 TDP), Turion and Core Duo have about the same power consumption, with Turion having better Cool & Quiet. (I haven't seen one website comparing T64 X2 and Core Duo with the same screen size, graphics chip, and battery capacity - please let me know if I missed it.)

11:33 AM, September 28, 2006  
Anonymous enumae said...

Edward said...

"I really don't understand how people could use the word "best" so liberally if they had mildly capable intelligence.

At least one should say, Intel has the fastest desktop CPU assuming price is not a factor, on a 32-bit operating systems with limited memory size and possibly no RAID configuration."

Edward I do not want to argue with you.

Best desktop, meaning overall performance, we are not talking about 2P and up architectural performance, we are talking single socket "Desktop".

Can you show me a link in which there is an FX62 beating an Intel X6800 on average in any configuration, 32bit or 64bit?

The last time I looked it was about 20-30% on average, that the X6800 was beating the FX62.

If not, it is simple to say that Intel has the best "desktop" processor.

1:05 PM, September 28, 2006  
Anonymous Anonymous said...

The $300 E6600 Core 2 beats the FX-62 across the board at a fraction of the price with much better power consumption. AMD has nothing that can touch the E6600, E6700 and X6800.

6:10 PM, September 28, 2006  
Anonymous Anonymous said...

http://www.theinquirer.net/default.aspx?article=34742

AMD finally is on pre-production stages, Id say the tapeout was sucesfully, heres some pictures of a QUAD core 65nm waffer

8:57 PM, September 28, 2006  
Anonymous Anonymous said...

" Anonymous said...

The $300 E6600 Core 2 beats the FX-62 across the board at a fraction of the price with much better power consumption. AMD has nothing that can touch the E6600, E6700 and X6800.

6:10 PM, September 28, 2006 "

in some tests, not in all of them...
want to go superpihappy again?

9:00 PM, September 28, 2006  
Anonymous Anonymous said...

>My mommy loves me so I know ok?

>Graham

Mmm....I am sure she does. Lemme guess, you bought her a two-in-one appliance, a computer with a netbust computer also functions as a room heater. Ok, that will great for this fall and winter.

But wait till next summer. Hahaha.

-Longan-

P.S. I am so surprised to see AMD will able to produce Athon-64 XP6000 using 90nm process.

As an AMD share holder, I demand AMD corporation not to waste the expensive capital equipment. Continue to milk out existing equipment while equiping the new fab with advanced equipment. Keep the expense down, down, down, and more return of investment for meh.

Use supperior and elegant designs to beat brute-force competitor.

Ok, it seems to me, AMD management is meeting my demand. Ha! :P

9:16 PM, September 28, 2006  
Anonymous Edward said...

"The $300 E6600 Core 2 beats the FX-62 across the board at a fraction of the price with much better power consumption."

There's this famous "joke" about a patient seeing a doctor:

Patient: "Doctor, I don't know why but it really hurts when I knock on my knee right here..."

Doctor: "Then don't do that!"

So my advice to you is this: don't buy FX-62!

OTOH, I run both an X2 4200 (overclocked to 2.7Ghz) and a E6600 (a Dell which I cannot overclock). For benchmarks, yes, E6600 is in average 5% faster (amid the 10% clock advantage of the X2). For real-world workstation tasks, such as compilation, simulations, and mathematical calculations, the E6600 isn't much faster, if not slower at times, especially when the dataset is very large - which happens commonly for workstation tasks.

Note I'm only using 32-bit Linux and Windows on both. For 64-bit, I'm sure X2 would fair even better.

OTOH, if I was running ogg or wmv encoding, then E6600 would be on top. What this tells you, at the minimal, is that the simplistic sense of "best" is at best misleading and often totally wrong.

Just to let you know, for CPU alone, the X2 4200 is 2/3rd the price of the E6600. For the whole system, the price of the X2 4200 is about 1/2 that of the E6600.

Of course, you can overclock E6600 to 3Ghz suppose you get a good motherboard & heatsink. For absolute highest-end desktop processor, Conroe wins and there is no point to buy or compare an FX-62. But for $300 or less processors (which BTW is less pricy than E6600), that is, for a good mainstream PC CPU, the 3-year-old X2's are still very competitive. Expect to plug in native quad-core 6 months later in your AM2 machine, too.

10:13 PM, September 28, 2006  
Anonymous Anonymous said...

Put your money where your mouth is Phd Pretender

I assume you are long on AMD and shorting INTEL...

Put it in writing if you are so sure. Make the options play on INTEL and you could then write the blog sipping a drink on some fine white sand beach.

but unfortunantly you got no money and couldn't make money on your forecast if your life depended on it.

That is the bottom line guys. Sharikou PhD is one poor un-employed idiot. If he knew a thing he'd be rich... but unfortunatnly he has predicted NOTHING.

10:19 PM, September 28, 2006  
Anonymous Anonymous said...

Graham wrote:
>Longan, I went to college also and here is what I know...

No, you don't. Haha...

You must have been a Pentium-D-grade college student, while I was a Athon-A-grade student. Ha!

I was talking about the high variable cost per chip for the quad-core glue-on offer from Intel.
Do you agree or disagree with what I said? YES or NO?

If you have any DIRECT refute about that then you talk. Or else go talk to your mommy. (She loves ya, I don't.)

-Longan-

P.S. I got an impression that the AMD-fanbois are generally smarter than the Intel-fanbois.

11:06 PM, September 28, 2006  
Anonymous Anonymous said...

"OHH RIGHT, ALMOST ALL ARE OLD NETBURST PENTIUMS BASED IN 65NM CORES!"

Actually the technology started it's ramp on Yonah - that would be mobile (READ - NOT A NETBURST) for the knowledge challenged.

Also all Core 2 is 65nm (granted that is still a relatively small amount.)

So fdo the maath crossover means >50% production on 65nm subtract off percentage of Core 2 (I think in 10-15% neighborhood if you count Woodcrest and Merom in addition to Conroe). Add in Core (1) mobile product and probably >1/2 of Intel's 65nm production is netburst.

Netburst is a mix of 90nm and 65nm processes.

11:36 PM, September 28, 2006  
Anonymous Anonymous said...

"At least one should say, Intel has the fastest desktop CPU assuming price is not a factor, on a 32-bit operating systems with limited memory size and possibly no RAID configuration."


So you could point us to the 64bit benchmarks where K8 outperforms Core2 (other than cryptography?)

And the $500 and 300 Core 2 models are not price/performance competitive or are you only looking at extreme edition? Last I saw the 6700 matched an FX62 at lower price.

And no RAID?!?!? Are you still referencing that old inquirer article on one chipset (which has been addressed) or are you a little more current on the info now?

11:41 PM, September 28, 2006  
Anonymous Anonymous said...

Eve of K8L? Are you sure you're informed. K8L won't be released until the first half of 2008 (Jan-June 2008) and is codenamed Greyhound.

Next year we will see Barcelona and Brisbane.. both of which are NOT K8L.

8:15 AM, September 29, 2006  
Anonymous Dr. Yield, PhD, MBA said...

Longan, you wrote earlier:
Intel CEO boasted the non-native approach has a 20% higher yield thus bringing down the fix cost. Sure thing. But he did not mention the veriable cost (Engineering cost / number of chips sold over product life cycle).

I hate to break it to you college grad, but what you call the variable cost is actually fixed. The engineering cost is fixed, and after the product is shipping, would be referred to as a "sunk cost"- the money is gone, no getting back if you sell zero or 10 billion of the parts.

The variable cost is how much it costs to make 1 additional part. And in this case, 20% less is a fair estimate- the yield of 2 individual parts packaged in a MCM is substantially higher than 1 part of 2x the die area, due to defect density. I have done an extensive treatment of this in other threads (mid-August I think).

So to answer your question to Graham, NO, I disagree with you. The fixed costs of designing the chip don't matter once its done. Variable cost, non-native quad core is cheaper. The real question- is 45nm native cheaper than 65nm non-native?

As to your your impression that AMD fanboys are generally smarter than those of the Intel ilk would be incorrect in your case. And that's speaking as neither an Intel or AMD fan-"boi".

9:03 AM, September 29, 2006  
Anonymous Anonymous said...

"Dr. Yield, PhD, MBA said...
I hate to break it to you college grad, but what you call the variable cost is actually fixed. The engineering cost is fixed, and after the product is shipping, would be referred to as a "sunk cost"- the money is gone, no getting back if you sell zero or 10 billion of the parts."

Omg! You are so uneducated. When Pat say 20% yield improvement, that is the per unit scope. The fix cost for that aspect is the cost of the wafer material.

Per unit cost, I've already state, (cost of Engineering / number of unit) = variable cost.

Gezz, please don't make me go back to my old book and cite you the precise page and chapter for this definition. I went to a class on this subject and did my home work.

Go get a book Engineering Finance for dummie to read. You got this totally wrong:

"The variable cost is how much it costs to make 1 additional part."

That is the fix cost, you got it reversed. The cost of blank wafer is fixed per die. The more dies you produce, the engineering cost can be spread out, thus, it's called variable cost.

-Longan-

P.S. May I have one of your self-given title, Phd Yield or the MBA? Now that I am right.

10:08 AM, September 29, 2006  
Blogger pointer said...

Edward said: There's this famous "joke" about a patient seeing a doctor:

Patient: "Doctor, I don't know why but it really hurts when I knock on my knee right here..."

Doctor: "Then don't do that!"

So my advice to you is this: don't buy FX-62!

OTOH, I run both an X2 4200 (overclocked to 2.7Ghz) and a E6600 (a Dell which I cannot overclock). For benchmarks, yes, E6600 is in average 5% faster (amid the 10% clock advantage of the X2).


wow, Edward, you really need to seek doctor advice - you are 'improving'. From last time attept to compare E6400 to 4200 (which most of the benchmarks show E6400 equivalent is 4800), now try to use E6600 (which actually beat Fx2). Nice try but it can't save your AMD in the performance desktop/laptop segment.

Don't be so desperate as AMD has lost this round. AMD can always come back, if not next chip, then next next chip if you have faith on it. I was always under the impression you might have some insider information on the future AMD chip. Just be patient if you are confindent with them. (my gut feeling is that AMD will still be lagging behind, even with the next 'native' quadcore when compare to the Intel chip at the same time frame though ... and your reaction seems to make me to believe my gut feeling :))

10:22 AM, September 29, 2006  
Anonymous Dr. Yield, PhD, MBA said...

Longan, you are confusing amortizing a fixed cost across parts with variable cost. Throw all the insults you like- I could care less. So see if you can follow along...

Variable cost: If I want to make more processors, I need to: buy more materials (wafers, gases, metals), and possibly incremental technicians to cover the additional WIP. Every wafer started incurs an incremental cost to the company. That is what drives cost/wafer and ultimately cost/die.

Fixed cost: NRE (non-recurring engineering, aka design costs, tapeout costs, photomasks, etc), share of depreciation on PPE (Property, Plant, Equipment).

The fixed cost is fixed- it doesn't change. If you perform more design spins, then you incur more of these costs, and in that sense they are variable. But there is no additional expense if you sell 1 wafer or 100,000 wafers.

If you sell 100,000 wafers, you get to amortize the fixed costs, and they are relatively less per part sold. The cost to the company hasn't changed- they still spent the same amount.

If you designed a product $10M and it never sold a single unit, your definition would mean the company incurred infinite variable cost. In fact, it cost the same $10M to kill the product as it would if they sold 1 to every person in the country.

Hopefully this is simple enough for you. I've wasted enough time correcting the fact that you were sleeping in class and then whining to your TA about your grades.

1:29 PM, September 29, 2006  
Anonymous Edward said...

"From last time attept to compare E6400 to 4200 (which most of the benchmarks show E6400 equivalent is 4800)"

Ah, I missed this part... well, I'm just too lazy to dig back, but I've always thought/said E6400 comparable to X2 4600, with E6300 to X2 4200. I think you're just trying to make an argument by faking/distorting my words.

Or maybe I mistyped the numbers - and you can backup your claims by providing a link to it?

In any rate, nothing is truer than running both systems yourself, like what I did. No, I don't and won't buy FX-62, because it's not worth it. No, I didn't overclock the E6600, because it's a Dell, plus I have no authority to do it.

Before you have any first-hand experience like me (java web services, database, and communication channel simulation), I don't know what you are trying to argue for. Sorry, it's not even arguing, but barking, IMO.

5:33 PM, September 29, 2006  
Blogger pointer said...

well, I'm just too lazy to dig back, but I've always thought/said E6400 comparable to X2 4600, with E6300 to X2 4200. I think you're just trying to make an argument by faking/distorting my words.

me too was lazy ... well, u did compare the E6400 to X24600 ... but you still 'imrpoved' and i still suggest you to visit a doctor ... some expert in illusion

Before you have any first-hand experience like me (java web services, database, and communication channel simulation), I don't know what you are trying to argue for. Sorry, it's not even arguing, but barking, IMO.

wow, you can understand barking ... good edward.. come .. sit ... good boy ...

if you have hard time to catch what i mean (inclusive of not able to understand what viral marketing means that a lot of ppl were trying to explain to you) ... may be if i throw an object you can chase and catch?

9:15 PM, September 29, 2006  
Anonymous Anonymous said...

Alright, I got the terms "fixed-cost" and "variable-cost" swapped and wrong according to the economy books. I stand corrected and appologize. You can keep you titles.

However, my definition for variable-cost-per-unit is still correct. The standard term for it is the engineering-cost-per-unit.

-Longan-

7:59 AM, September 30, 2006  
Anonymous Edward said...

"if you have hard time to catch what i mean (inclusive of not able to understand what viral marketing means that a lot of ppl were trying to explain to you) ... "

Oh well, you (inclusive of those "a lot of ppl") are just helpless. When I told you things of material significance, experiments that I performed myself, facts that I see with my eyes, you just divert to clueless chanting of totally related stuff (which again you have no clue of). It seems to me, that you have no good training of computer architecture, you have no experience in both (K8 and Conroe) systems, and you only follow those amateur analysts and sites for "news" on computing technologies - then plese, go bark at someone else, or go get some good education first. It's not my duty to convince you otherwise if you so want to beleive in your beloved Intel marketing; it's just waste of my time.

9:24 PM, September 30, 2006  
Blogger pointer said...

Oh well, you (inclusive of those "a lot of ppl") are just helpless. When I told you things of material significance, experiments that I performed myself, facts that I see with my eyes, you just divert to clueless chanting of totally related stuff (which again you have no clue of).


oh well, you should go see those forum with your 'own' eyes and see how AMD spammed :) (indirectly, whatever u said .. it is viral marketing ... if still cannot understand ... i'll give up, it's not my responsibility to make you understand ... and same with those benchmarks ... for people like you, there were a lot of good benchmark sites before C2D days ... and out of a sudden all are amateur (guess what, there are NONE now for you all because all praising C2D :) ... inclusing one AMD fan site hahaha )


It seems to me, that you have no good training of computer architecture, you have no experience in both (K8 and Conroe) systems, and you only follow those amateur analysts and sites for "news" on computing technologies

well, i have to admit my job does not require me to test intel CPU or AMD cpu (i guess, ur job is) and i'm not that rich to buy few systems and my E6600 is just on its way to my home. nonethelss, i depends on computer to earn a living (without too specific in such leaving a clear digital trace here ... i need to understand communication protocal, internal and external bus protocol, etc).

it seems to me you are really desperate with AMD current situation though :) ... as i said, be patient (old advice), don't bark (newly added advice). AMD still has the upper hand with server 4p and above ... and it also invaded a lot of new OEM product lines: AMD will stay and will not BK.

- then plese, go bark at someone else, or go get some good education first

btw, which police trained you? from movie i know that the scotland has pretty good police dog training ... you must have passed the training with flying color ...

oh, plz don't bite too.

9:24 AM, October 01, 2006  
Anonymous Anonymous said...

"In the coming quarters, Intel will suffer massive losses due to lower unit share and drastically lower ASP."

Sharikou - please define massive losses now, so that you cannot adapt this subjective statement at a later date. Is >$500Mil, $500K or $500 considered massive?

While you're at it I've notive you've nerver provided a specific mathematical formula to determine market share runrate (I suppose so you can wiggle out of that one too?). As an example please show us how you calculate AMD's market share (runrate) exiting Q2'06? (or any other past quarter for that matter)

I doubt I'll get a response on this as many have asked and you seem to refuse to answer this. For fear that you will be held accountable to you predictions?

7:44 PM, October 02, 2006  
Anonymous Anonymous said...

From German Computerbase:

http://www.computerbase.de/news/hardware/prozessoren/intel/2006/oktober/idf_fertigung_quad-core_71_dollar/

70-80 $ per Quad-Die, this means only material and production...

3:50 PM, October 07, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home