Conroe volume is negligible
Intel will ship 1 million Conroe CPUs in 7 weeks, or 0.14 million per week. Since the world's weekly PC consumption is 4 million, Conroe represents less than 4% of the market. It's non zero, but close. 550 PC makers will produce Conroe desktops, 1,000,000/550/7/7 = 37. On average, each PC maker will get 37 Conroe CPUs per day. Now you know another reason why Michael Dell has to go AMD. Only AMD can save Dell from going down like Enron.
Looking at Intel's new price list starting from $30 for a Celeron, I seriously question the IQ of Intel management. There is only so much demand, once you sell a chip at $30 and at a loss, the market shrinks by 1. Intel's inventory has piled up to the neck - $8.66 billion worth of legacy chips and more are produced every day. At the end, a lot of these legacy chips will be sent to the landfill. A rational move should be selling the higher priced CPUs first, and give up the low end crap -- they will be written off later.
AMD's Sempron starts at about $50.
I reiterate my projection that Intel will post operating loss starting 3Q06, and BK in five to seven quarters.
posted by Sharikou, Ph. D. @ 7/27/2006 05:36:00 PM
52 Comments:
Intel fanboy says, which landfill?
Does anyone know how much it costs Intel to produce 1 low end CPU such as a Celeron?
Dr., I thought the original plan was to ship Conroe by the end of July. Now it is going to be August. So they are 2 or 3 weeks late. Also, Woodcrest was released on June 26. If you believe the Dell server site over the last few weeks, they are beginning to ship now except the most powerful Woody 5160 which delays shipment another 4 days. Maybe Conroe systems will first show up at the end of September based on the Woodcrest delay. So much for Intel's Q3.
http://configure.us.dell.com/dellstore/config.aspx?c=us&cs=04&kc=6W300&l=en&oc=pe2900-min&s=bsd
I doubt they are going to ship as many Conroes as soon & in quanities they are currently predicting.
Rahul said the following on April 27 concering WCM
"It seems to me that Intel’s new processors are heavily cache dependant. That being said, it would normally take a great deal of silicon to produce a processor with 4 megabytes of cache. You would think this could limit their production output – but then there’s another issue. If they are able to produce the processors there’s the issue of yield. When you start building such a complex processor with a large amount of cache the yields will probably come down significantly....Though this doesn’t seem to bode well if they are trying to eat up share from their competitor, it may not be their goal. Even if they win in performance they may not be able to produce enough to keep the entire market satisfied."
http://voodoopc.blogspot.com/2006_04_01_voodoopc_archive.html
Intel reminds me of Where's Waldo.
WhoisJ_Galt
TigerDirect.com is selling Intel Core 2 Extreme X6800 CPU's right now. They are available and shipping. You can get either the cpu itself, or combo with a motherboard. It seems they are available now if you want one.
How many of the 4 mil/week are desktops and how many laptops? Fair to say 50/50? Then Conroes will account for 8% of desktop market which seems perfectly adequate for a ramp-up destined to fill the high-end segment demand at first.
Sharikou, what's your reference on the 8.66b Netburst inventory? The Intel slide you have on your blog shows 4.33b total inventory and it's obviously some mix of laptop parts, Netburst and Conroe/Woodcrest.
This may sound a tad negative but I don’t believe Intel when they say 1 million Conroe in 7 weeks because Woody is still missing in action after the launch 4 weeks ago.
Sharikou, what's your reference on the 8.66b Netburst inventory? The Intel slide you have on your blog shows 4.33b total inventory and it's obviously some mix of laptop parts, Netburst and Conroe/Woodcrest.
$4.33 billion inventory, 49% margin leads to $8.66 billion. To be precise, I should say "legacy" instead of Netburt. Only 0.2 billion was Conroe, the rest are Netburst and Core Duo32 crap which will worth 0 in 5 months.
Where are the 4x4.. how many are shipping?
Where is the K8L samples?
Where are the 65nm chips.
You all are idiots... COnroe will take the top and midstream. They don't represent 50% of the market. PentiumD at the bottom will make AMD eat losses on their low yielding 65nm process.
get off your sorry AMD bandwagon. There more you crow now the more a fool you look like. Wait till 2008 for a glimmer of hope. The window will be short as INTEL's 45nm will crush AMD again.
The Doctor
Has Shari Poop ever been right about anything on this blog?
What ever the actual amount or value of the legacy chips there’re in the warehouse, they are definitely not worth much today. The write off number may stay large for tax reasons, but I’m not an accountant. We should ask “The Doctor” he knows everything.
Conroe starts to cripple under 64 bit eviroment running 32 bit applications.
http://www.xbitlabs.com/articles/cpu/display/core2duo-64bit.html
I wonder would it crap all over the place had the benchmaking was carriedout using true 64 bit apps.
"it would normally take a great deal of silicon to produce a processor with 4 megabytes of cache."
Might want to check out actual die size before these stupid comments - Core 2 is currently smaller than an Athlon processor.
"When you start building such a complex processor with a large amount of cache the yields"
As many have pointed out (including Sharikou) the cache is now the majority of the area. Cache is not the "complex" part of the chip, it is a 6 transistor cell that is repeated over and over... The only major issue with cache yield is random defects and given that Core 2 die is (I believe) slightly smaller than P4, yield will not be worse. Any yiedl issues will be on the logic part of the processor which is comparable, if not smaller than previous Intel CPU designs.
Sharikou, you will probably not publish this again, but for the fith or 6th time you are WRONG again on the inventory.
The entire inventory is not simply Netburst. Putting aside the 0.2Bil Core2; there is still a chunk of the inventory that is mobile chips ("core" or as you call it "core 32" architecture which is not netburst!)
Also the 49% margin is across all product lines including servers which generates more margin than desktop so I think converting the inventory with a straight 49% margin is overestimating potential revenue.
Of course these little facts get in the way of your 90% of production is Netburst and 8.66Bil inventory crap that you continue to spew incorrectly.
If you have ang guts you'll publish this and let readers draw their own conclusions!
"And Intel says 550 PC makers plan to come out with models with Core 2 Duos."
While I like your little calculation in your blogs, I don't think all 550 models are coming out in next 7 weeks. I could be wrong but I also suspect that Intel doesn't equally allocate new product along all vendors (Again, I could of course be wrong on this though!)
To the poster who mentioned 4mil CPU's being desktop/laptop combined - good point. There's probably a small amount of servers in there too?
How are the low power athlons coming along anyway - have they been ofifcially lauched yet? What % do those make up of AMD's product shipment on desktop? Those seem to have ramped rather quickly.
Intel will ship 1 million Conroe CPUs in 7 weeks, or 0.14 million per week. Since the world's weekly PC consumption is 4 million, Conroe represents less than 4% of the market.
No offense, but you're being kind of contradictory. First you say that Conroe is not making up a large amount of sales while at the same time you say that Netburst processors would never sell. I would think that if Conroe doesn't make up a significant percentage of Intel sales, that means it's the Netbursts that are selling.
In regards to the article you linked, did you miss these lines?
And he said the Core 2 Duo production ramp-up will perhaps be the fastest in company history.
"We'll ship 1 million Core 2 Duos in less than seven weeks," Otellini said. It took a year to ship that many Pentium 4s, Intel's previous major advance in microprocessors.
Conroe is going to be their fastest ramp. Even if your 4% figure is accurate, you can't really expect them to have a massive 20% switchover in less than 2 months.
Sharikou said: "$4.33 billion inventory, 49% margin leads to $8.66 billion."
That's not how inventory is reported. Inventory level is based on sale price not cost. Here
You keep talking about reasons why Dell switchs to AMD, but I still haven't seem any AMD Dell systems. They haven't even demoed them like they demoed Conroe systems before. It seems strange to me that the "news" a few days ago was that Dell saids they are expanding their AMD relationship beyond servers. However, we've known that and they've been saying that for months now.
It seems to me that perhaps Dell feels that the new Core 2 Duos are sufficiently impressive so they're now dragging their feet in offering AMD systems.
"Looking at Intel's new price list starting from $30 for a Celeron, I seriously question the IQ of Intel management. There is only so much demand, once you sell a chip at $30 and at a loss"
I know you have previously calculated chip cost for Intel (which is of course pure speculation on your part). But you also ignore different die size of Celeron vs Conroe. Cost/wafer is ~equivalent for a given technology node, so smaller die (Celeron's don't come with 4MB of cache do they?)= more die per wafer = lower cost/die. Add in the fact that these are lower performing parts so yields and bin splits should be better.
So to add this up for you, cost of a celeron chip will be lower than your calculation of an average Intel CPU that you have previously done.
ALthough you may have factored all of these variables already into your cost/die analysis....
The Doctor seems to have lost patience (sorry...couldn't resist the pun) with Intel's lagging behind AMD's draconian 90nm chips that have been beating them for 3-4 yrs.
But alas, all is right again. Intel finally has a chip that can compete with a half-decade old µ-arch.
That is until AMD does 65nm's right, proving its not what you do, its how you do it.
A) Shipping 1 million Conroe over next 7 weeks.
C) 200Mil of Conoroe inventory (that 200million is cost not sale value)
How can Intel have 200Mil in Conroe inventory already and yet only ship 1 million chips over next week.
In your statements on Conroe availability are you factoring in inventory that Intel has already been providing to OEM's? 200Mil is more than 1 million in chips today yet alone 7 weeks from now.
Was the 200Mil in inventory Conroe or Core 2? (meaning servers and/or mobile as well). Your 2 statements aren't consistent with each other.
Was the 200Mil in inventory Conroe or Core 2? (meaning servers and/or mobile as well). Your 2 statements aren't consistent with each other.
Dude, those were Paul Otellini's statements, not mine.
Bigbadwolf "Conroe starts to cripple under 64 bit eviroment running 32 bit applications.
http://www.xbitlabs.com/articles/cpu/display/core2duo-64bit.html"
You did actually look at the article right? (And not just the stupid headline on the Inq?)
EVERY BENCHMARK BUT ONE showed the Conroe outperformed FX62 in 64bit mode. You might want to actually try LOOKING at the graphs (you know the things with the numbers and the pretty colors).
You said Intel was going to start posting losses in Q2. Wha happened?
I 'spose in Q3, when Intel makes a couple billion in profit, you will forcast an Intel loss in Q4.
I'll bet you cry a little while you type, the more you try to look smart the more stupid you look.
Trying to see the math here:
Intel has ~75% of the CPU market. Conroe shipments will only count for ~4% of the market, yet the rest of Intel's inventory is obsolete...
It will be obsolete evenetually but if they can pare the iventory down before Core 2 say hits 50%-75% production (which you claim is somewhere after Q1'07, that gives Intel about at least 3 quarters to reduce desktop chip inventory, keeping in mind that not all of that inventory is desktop.
Now how would you pare down the inventory...
Dr. Sharikou
* Conroe volume is negligible = True statement for a year
* Dell needs AMD = Definitely True statement
* Intel Celeron $30 is stupid = Absolutely True statement
* Inventory $8.66 B = Most likely close to True with small margin for error
* Intel BK 5 to 7 Qtrs. = I have issues with this one (I’ve read all your past post) your numbers add up on paper and your logic is sound, but in the real world businesses adjust, then execute there plans be it marketing, sales, production. Make multiple changes. Intel is not a static business model it’s a moving target to the competition. Intel’s most potent tool is lying a lot. But BK in 5 to 7 Q’s, I don’t think so.
I promise to apologize formally on this blog if this one comes true. Hell, I’ll do a small act of worship if you nail this one.
I have programmed my Palm and Outlook to start beeping Feb 1st 2008
"Where are the 4x4.. how many are shipping?
Where is the K8L samples?
Where are the 65nm chips."
There is a BIG difference!
AMD is a planned and reliable release of a future product.
Intel is a paper/propaganda release just to try and stop AMD sales. What people are critisizing is the fact that Intel is all talk no action.
Note: Use a better comparission next time!
"Dude, those were Paul Otellini's statements, not mine."
My question, dude, was this Conroe or Core 2 - I believe it was Core 2, if it was Conroe like you stated than Conroe would represnt more thana potential 4% of CPU sales. Either that or the 1 million over 7 weeks quoted by Intel did not also count Q2 inventory that they were building up prior to launch.
(Which would incicate more initial supply than you calculate)
"First of all - the calculations are just self serving and wrong. Not all 550 OEMs will have Conroe SKUs at launch and do you think Intel provides supply equally to all OEMs @ of 37 per day."
Its a figure of speach moron! Its a statistical analysis just to give a general impression/idea.
People, Come on! Take what is written in here with a grain of salt! We don't need to wast our time reading about how you disagree with Sharikou and not backup up your statement/point!
Don't write just to say that you disagree or that Sharikou is stupid.
Say something intelligent... back up your point... give a fact, etc. Yes I know that is very hard for Intel fans.
"Might want to check out actual die size before these stupid comments - Core 2 is currently smaller than an Athlon processor."
If you look at it that way it might seem reasonable. But take a look at it this way:
If you paint your initials on a dime, you will get 100% yeild. Now paint your initials on the tip of a pencil. Next paint it on the tip of a pin. My point (and I'm guessing) is the smaller you get, the harder it should be to get good yeilds, especially at the start of production.
Next, if you compare die size regardless of their manufacturing size, statistically your yeilds will drop if you have a larger die. As you said... "random defects" increased by my point above.
Given the two, this will affect yeilds. Don't just look at with physical size.
First it was lies about the demos at IDF, don't believe them numbers intel fixed them.
Then came the independent benchmarks, don't believe them INTEL fixed them. Wait until we get independent testing
Then came independent testing, don't believe it INTEL will slip the launch.
Then came the launch. don't believe them INTEL will go bankrupt. You can't buy any.
What do you think the 3 300mm 65nm factories are making you idiots.
They are making Conroes and Meroms and they are going to be the hottest thing this Christmas. What you won't be able to find is a PS3. Why.. because yields on that IBm 65nm process suck.
By the way I do know how much it cost to make a low end Celeron. Its safe to say.. INTEL isn't losing money.
The bottom line.. INTEL's core2 crushes anything AMD has. AMD has nothing that can compete at the high end nor main stream for another year.
AMD will scrape by using their top end process that is expensive and low yielding to compete with INTEL's bottom end high yielding 90nm process while it gets crushed by the high yielding 65nm process at the top.
It sucks when you can't change reality... you can blog all you want the bottom line is INTEL has the capacity, technology, benchmarks and the MONEY.
"Say something intelligent... back up your point... give a fact, etc. Yes I know that is very hard for Intel fans."
So why are you posting at all?
I've been seeing all over the net about how impossible it is for all of the companies selling these chips to get any. I wonder why it is so difficult for such an amazingly large company like Intel to begin producing them at similar numbers as they do there other chips.
CyberSurge
"If you paint your initials on a dime, you will get 100% yeild. Now paint your initials on the tip of a pencil. Next paint it on the tip of a pin. My point (and I'm guessing) is the smaller you get, the harder it should be to get good yeilds, especially at the start of production."
So, since you seem to have little to no knowledge about Si processing technology from the above statement; transistor sizes are ~ same between P4 and celeron; they're are just FEWER of them (since there is less cache); thus smaller size. So while your metaphor is nice it is completely not applicable to comparing dies size!
I assume you realize to use your metaphor, the transistors for P4 and celeron are "painted on" ~the same size on the wafer; just the wafer is diced up into smaller pieces AFTER you are done painting.
What you described above is mor e like technology node scaling not the difference between a small and large die!
As for random defects; virtually all process steps are the same between P4 and Core or P4 and celeron for a given technology node, thus random defects/wafer is relatively constant. So assuming all these random defects are die killers; you will get HIGHER yield with smaller die as a smaller area means a smaller probability of a defect showing up in that area (assuming random defect mode).
If you drew a big circle and a small circle on the ground and then threw a quarter in the air, is the probabbility of the quarter landing in the small circle greater or smaller than landing in the bug circle?
"I wonder why it is so difficult for such an amazingly large company like Intel to begin producing them at similar numbers as they do there other chips."
Any idea how long it takes to go from a raw silicon wafer to a package chip for a typical fab throughput? (Typically >3 months) As at an early stage of a product goes through multiple "steppings" to improve performance, address critical speed paths, improve overall bin splits at higher frequency - you do not want to "instantly" convert all of your product line over until you have a "good" stepping. This is a bit of an iterative process - you don;t want to crank out a large volume of chips on an early stepping if there is a decent chance it will change.
Even if Intel (or AMD when they release a new product)was able to converting much CPU mix much faster there are other factors like a simulataneouly equivalent conversion of supporting products like chipsets, motherboards, etc.
This is similar to Turion X2 or the low power version of Athlons or Intel's launch of the Yonah (Core) products. I don't understand why people are getting all ecxited about Conroe volume being negligible ONE DAY into its launch and expect chips to convert over "instantly" or at a ridciculous rate. A new product launch, at the very early stage, is generally targeted at early adopters where you can command a higher premium than the average consumer.
"The window will be short as INTEL's 45nm will crush AMD again."
ehmmm Intel will crush AMD again??? AMD crushed Intel for several years, now with Conroe they maybe can hold the crown for a few month but then they will again get crushed by K8L. And just switching to 45nm will not bring the crown back.
Some of you think Rahul Sood is an idiot? You Intel Fanbois make me laugh!
"You keep talking about reasons why Dell switchs to AMD, but I still haven't seem any AMD Dell systems."
Kevin Rollins (CEO Dell):
"On the corporate side, Dell started using processors from Advanced Micro Devices Inc. in its higher-end servers, breaking an exclusive reliance on Intel Corp. chips. The Dell-AMD relationship will expand, Rollins said, though he declined to say exactly how. Intel has new products on the horizon, as well."
LENOVO Thinkcentre:
"ThinkCentre A60 models offer leading-edge performance so you can do more in less time with true multi-tasking. AMD Athlon 64 X2 processors put the power of dual-core technology on your desktop — delivering performance and increased efficiency. Built to handle multiple programs running simultaneously, dual-core technology means optimized office productivity for your ever-growing application needs"
And those benchmark programs are all crap. Conroes realworld application benchmarks such as in gaming show it's very close to an FX, so you can't say Core architecture is that much superior to the 4 year old K8, Intel has only managed to catch up.
Intel (Netburst) CPUs have always been faster in MP3 decoding, zipping and video, and also Core is. But they are competing against an old architecture.
Looks like another paperlaunch! No Conroe available in whole Europe and the US
Dell XPS with Core 2 Duo ship date:
"10/9/2006"
And the note: "Please check preliminary ship date, as shipment will be delayed."
http://configure.us.dell.com/dellstore/config.aspx?c=us&cs=19&l=en&oc=DXPS700B2&s=dhs
AlienWare ship date:
"Monday, September 04, 2006"
http://www.alienware.com/configurator_pages/Area-51_7500_r3.aspx?syscode=PC-AREA51-7500-R3&subcode=SKU-DEFAULT
anonymous said...
"You keep talking about reasons why Dell switchs to AMD, but I still haven't seem any AMD Dell systems. They haven't even demoed them like they demoed Conroe systems before. It seems strange to me that the "news" a few days ago was that Dell saids they are expanding their AMD relationship beyond servers. However, we've known that and they've been saying that for months now.
It seems to me that perhaps Dell feels that the new Core 2 Duos are sufficiently impressive so they're now dragging their feet in offering AMD systems."
The Dell AMD deal currently stands only in the MP server segment, and I don't believe any of these systems are expected to be available until late 2006 or early 2007. That being said, it is common knowledge that Dell is planning on expanding their relationship with AMD. My guess is that they will not do this until AMD is capable of providing solutions in the form of chipsets and integrated graphics (which makes the ATI-AMD deal make a little more sense, but not much). I would expect any expansion announcements to be made later in 2006, with products being available in mid to late 2007.
Keep in mind that Dell works with margins between 6%-10%. They have very little room for error. They expect the expansion of their AMD offerings to somehow save the company, but things are not that simple. In order for them to guarentee they keep making money, they have to ensure that any product designs they make will sell, and sell in high volume. Why would they make NEW designs for an Athlon desktop that so obviously lags Core 2 when they already have the Core 2 designs done? Why would they throw AMD mobile offerings (which are currently crap even compared to older Core Duo - not based on the new Core uArch, or Core 2 Duo - parts) into their selection? It just doesn't make any sense at the moment.
This is why you see them sticking with the MP server segment for their AMD offerings - the only segment in which AMD is still competetive (and not because of their processors, mind you, but because of the HyperTransport protocal). Don't expect Dell to announce a grand switch-eroo until AMD has something competative to offer int he desktop and mobile space.
"
I assume you realize to use your metaphor, the transistors for P4 and celeron are "painted on" ~the same size on the wafer; just the wafer is diced up into smaller pieces AFTER you are done painting."
You idiot. He isn't describing die size or wafer size, he is describing the process 65nm is smaller than 90nm, therefore it is harder to get that process to produce perfect results. The "dime" and "pencil tip" aren't describing bigger wafers to smaller wafers, it is describing 90nm to 65nm. Intel's yields on their 4MB cache have got to be terrible. But they can disable the bad cells and knock some of them down to 2MB cache. But I'd be surprised if Intel is happy with their yeild results right now.
P4 die size = Celeron die size.
Celeron = P4 with cache disabled.
P4 manufacturing cost = Celeron Cost.
All this speculation is making me drool for the cheap prices on older Intel processors. I think I'll go out and buy an upgrade today!
Sharikou, thank you for your blog. I can always read it and know which technology to buy into by doing the opposite of what you say.
You're like the magic 8-ball:
"Intel is going bankrupt next year" = Buy Intel stock!
"Intel processors are junk" = buy an Intel system
"Intel did something evil" = Intel did the right thing
"AMD blows Intel out of the water" = on one or two out of 20 benchmarks
"Something will happen" = It won't.
Thanks again!
Since you've been saying that intel will be bankrupt in 5-7 quarters since last quarter shouldn't you at least update that comment to 4 - 6 quarters? Or will it perpetually be 5 -7 quarters?
Mojo, great comments. I think I will save them and see who is more accurate. Mr Sharifraud? or Mojo? I am banking on the Mojo prediction of Intel Q3 profitable and AMD not.
I'll bet you cry a little while you type, the more you try to look smart the more stupid you look.
Amen!
AMD October price cut:
http://www.digitimes.com/news/a20060720PR212.html
Keep in mind that Dell works with margins between 6%-10%. They have very little room for error. They expect the expansion of their AMD offerings to somehow save the company, but things are not that simple. In order for them to guarentee they keep making money, they have to ensure that any product designs they make will sell, and sell in high volume. Why would they make NEW designs for an Athlon desktop that so obviously lags Core 2 when they already have the Core 2 designs done? Why would they throw AMD mobile offerings (which are currently crap even compared to older Core Duo - not based on the new Core uArch, or Core 2 Duo - parts) into their selection? It just doesn't make any sense at the moment.
You have to be kidding right? By your logic Dell has to stop producing PD and Celeron systems since Core 2 is 70-90% faster.
With the prices of Core 2 mobos 2.5X Am2 SLI mobos, teh price performance crown has switched back to AMD. They are not far enough ahead that the price drop won't level it back out.
If they put 5000+ on 4x4 (they better if they want my money) they will have it locked up for at least a year.
80% increase on multithreaded with the first iterations is incredible. they even used ECC RAM so expect at least 10% more from nonECC.
If the leave the prices teh same then the mobo should come in under $200 which means $850 for 2 5000+ and the mobo.
it maybe cheating but that will take Core 2 out back and "poke it in the eye and beat it with a stick" - Rahul Sood
Admit that there is no more "good old boy" PC market. there's a new sheriff in town and he is serious.
"If the leave the prices teh same then the mobo should come in under $200 which means $850 for 2 5000+ and the mobo."
I have to disagree, 4x4 will be FX only acording to AMD.
Some have speculated they will release new processors marked FX to reduce prices but they have already stated the cheapest price for the two processors would be around $1000.
Enthusiaist motherboard will most likely be at least $250.
Now your looking at about $1250 not $850.
Link to the article.
MadModMike says that 4x4 will be for X2's and FX's but i think i will get two semprons in 4x4 (Just Kidding)
Regarding yield on large caches, there are methods for this that have been in use for many years.
http://developer.intel.com/technology/itj/q41997/pdf/manufacturing.pdf
"80% increase on multithreaded with the first iterations is incredible. they even used ECC RAM so expect at least 10% more from nonECC."
Why only an extra 10%, why not 20-30%, after all it is sreasonable to think you will get >100% performance by doubling chips.
By the way have these 80% benchmarks ever been published? The only thing I've seen is a 2 sentence Inquirer article with no mention of who did the testing or what the tesing was done on, or what benchmarks?
Let's try to be balanced here - if this article said the same thing about an Intel chip wouldn't you be crying out for the actual benchmark and demanding to know who did the testing. Why don't people do as they say - Wait for a retail product to com out, not an engineering sample or engineering version of a board.
"P4 die size = Celeron die size.
Celeron = P4 with cache disabled.
P4 manufacturing cost = Celeron Cost."
So, just spitballing here, do you think yield is also a variable here. Folks keep saying how crappy 4MB cache yields are (let's assume this is correct although noone has published any data to suggest it is). By being able to disable cacce in non-yielding areas you would get a better over yield on Celeron.
This would put chip cost at a lower level than P4.
Your P4 manufacturing cost = Celeron Cost applies to wafer level cost, not die level cost. Die level cost is only the same if die size is the same AND yield rate is the same. Also don't Celerons generally operate at lower speed (thus bin splits would be better as well, no?)
Post a Comment
<< Home