SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Politics : Formerly About Advanced Micro Devices
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: A. A. LaFountain III who wrote (58611)5/18/1999 2:22:00 PM
From: Kevin K. Spurway  Read Replies (1) of 1580820
 
Let's say a 500 MHz Xeon with 1024k L2 goes for $2000. Using your highend estimate of 80% margins that is $1600 in profit/unit. It's also $400 in cost/unit. You're saying that the margins on the Celeron could be as high as 55% (so for a $95 Celeron 400, that means about $50 in costs). Let's try to verify this $50 number. We know:

1) Mendocino Celeron has a larger die size than the Xeon core, making it more expensive to manufacture.

2) Celeron has (in some cases) no SECC cartridge/no slot making it cheaper to manufacture. Let's say these materials cost $10/chip.

3) Celeron has no large off-chip L2 cache. Anyone know the current spot market price of 1024k of full speed SRAM? Deduct 40% margin and you get Intel's cost on the Xeon SRAM. Let's use $200 as a cost here ($280 spot market)--this is a WAG but I expect it's way on the high side. Someone give me a good number here?

4) According to Ten, Celeron has no desperate need to climb the speed curve (less validation necessary), no need to validate on the 100 MHz bus. I'm not sure how good an argument this is because Celeron overclocks at least as high as Xeon does, if not higher (because of the lack of L2). In other words, Intel could sell a 550 MHz Celeron today if it wanted to. Since it is the same process that Intel is using on all of these chips, process development costs should be spread evently. That being said, Xeons should be tested to higher reliability standards than Celerons, have to go through multiprocessor validation, 100 MHz bus testing (more sophisticated electricals), etc. so I'll throw in a few bucks of extra cost here--say $20.

So to put together a "virtual Menocino" using the Xeon:
$400 Xeon cost
-$10 slot 2 cartridge
-$200 cache
-$20 testing, etc.
=$170 total cost.

Add another 10% for the 128k of L2 on die, and you get $177 of cost for a 400 Mhz Celeron (downbinned).

Are Intel's margins for the Xeon higher than 80%? Maybe.
Are the expenses I'm stripping out of the Xeon cost reasonably accurate? Maybe not.
Is Intel selling below marginal cost? Unlikely.
Is Intel selling below average unit cost? Possible.

Kevin
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  

HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2026 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News