The High Cost of Customization

22

Comments

+ Add a Comment
avatar

Drew7

I have what is considered an "iPhone 5 killer", and I spent $150 for it NEW... Shipped to me for FREE. It's an HTC Evo V 4G. And it has a dual core Snapdragon chipset running at 1.2 Ghz. And because it's on the Android platform, it not only performs on par with the iPhone 5 (some reviews even say better), but it's WIDE OPEN; Change the battery? Sure. Add more memory? Sure. Bigger screen, Gorilla Glass, and some things that the iPhone 5 doesn't have... Like FM radio. Point is, W-H-Y would I spend 3X the money for something that's 3X as LIMITED??? Because of some "custom tailored chip"? One where I truly can't tell the difference between it and my own? Pay money out the ass for something less... Be my guest. Glad I don't have your I.Q..

avatar

igoka

This is very strange article . Yea right poor Apple , so much investment , lets them charge whatever they want. Right ? How about other companies ? This is a business you have to invest to gain something and as I know they got a good profits . What so special about Apple ? Like Gordon once said " This is a cult ".

avatar

wcj786

The thing about this article is that the writer tries to imply that all custom chips are extremely expensive, which is probably true. But, he uses an example of Apple's A6 and tries to justify his opinion with the example.

The problem arises when he imples that it cost $508 million just to design and produce the first test fabs. THAT is untrue. Apple bought PA Semi and Intrinsity for approximately $398 million. That is a sunk cost outside of the design and manufacture of the A6. Sure, those new entities did design the chips, but the cost of purchasing those companies would end up being spread over EVERY chip they design, not just the A6. When they come out with an A7 or whateveer they decide to name their next chip, the cost for that new chip will not be $508 million also, as the cost of the companies will not be added in.

The point is that the example is erroneous and should not have been used to attempt to justify such an outrageous number for a single chip design. If the author had used the true cost numbers, based on his own statements, the cost would have been $110 million. Remember, that is for design and manufacture of the test fabs, not all the chips Apple used. That would have justified his statement of the custom chip without attempting to obfuscate the truth that the purchase of those companies is a sunk cost, not associated with the actual cost of the chip itself.

avatar

Renegade Knight

I disagree.

Apple to design a chip at all - had to invest in a team that could do the job. They bought a company. That's not a sunk cost because it's what they had to spend going in. Provided the goal was an Apple chip (which it was) and not a investment in a design firm with great upside potential.

It's a smart move if Apple can both make money marketing Intrinsity's services and patents to recoop some of their costs to develop their own chip. This appears to be at least partly the case (per Wikipedia)but their website isn't working for me to see if they do offer chip design services to others.

Reality is both your point and Tom's point are partly right. Apple could have just paid a lot less for the firm to design the chip rather than buy the firm. But Apple is proprietary. I suspect they wanted full control and no question of who's IP is in the design.

avatar

wcj786

You would be correct ONLY if the companies (and their employees) had no input into ANY other chip designed by Apple. If they do, then the price of those companies is INFRASTRUCTURE, or sunk cost, not part of the product cost. That is simple economics. They work on more than one product, their price can not be accounted towards any product, only infrastructure.

avatar

graydiggy

Obsolete in 2 years? Are you kidding me? They are already obsolete.

avatar

azuza001

I have to say when I read the articles title and the teaser part on the main page I was expecting a thought provoking or at least lengthy discussion on this. Thats not what I received. In reality all this is is a "FYI, it costs Apple probably about 500 million dollars to make it's processor for the Iphone 5, processor designs can be expensive."

My retort - "No $hit. Blue skies make the average person happier more often then grey skies do as well." In other words, this isn't even really an article, more like what you'd see in a forum post or one of those "Hey, you ever notice...." comedy routines.

I mean, 5 of the paragraphs in the article explain the other 2 paragraphs. I don't think I need to continue.

avatar

AFDozerman

I personally like it. Being short makes it a discussion magnet. I like discussion.

avatar

Eoraptor

Ditto. I opened the link expecting a treatsie on pros and cons of one-off chips versus fabber units. Instead I got... well maybe not apple fanboyism, but certainly nothing more thought provoking than cupertino sales copy about how this one company chooses to spend for production.

was there a point to this article somewhere I missed?

avatar

Bullwinkle J Moose

you want a thought provoking article?

Don't say it Bullwinkle

DON'T SAY IT!

ummm........

k, nevermind

avatar

AETAaAS

Don't know how 'custom' you can call a tweaked ARM chip mated to a PowerVR graphic controller. And with its competitors (Samsung, Qualcomm, TI and nVidia to name a few) also playing the SoC game, will Apple chips go the way of PowerPC when they are seen as being too costly and not competitive enough?

avatar

Hey.That_Dude

Apple's improvements to the ARM core IP where more impressive in the past.
As for PowerPC, IBM is kicking ass in the server's department (btw, you should look in your home router, you'll find a PowerPC chip in that). I'm sure there are plenty of people who would like an all mac ecosystem, down to the server's that store their iStuff data. SO them going into servers would make someone happy.

avatar

AETAaAS

I'm on a Linksys WRT54GS which uses a Broadcom processor. But that is besides the point, I mean Apple might shift from using custom processors to mainstream ones like they did with PowerPC to Intel, and not the loss of PowerPC as a whole.

avatar

AFDozerman

An interesting thought crossed my mind while reading this. Wouldn't it be great if chips were made like the suits that he mentioned? Just ring up some 3rd party contractor: "Hi, I'd like an X86 SOC with two ivy bridge cores, three piledriver modules, and a unit of six hundred GK110 stream processors with GPU compute specific drivers all tied together by twelve megabytes of L3? Oh, and while you're at it, can you do a PCI card based on GCN 2.0 with about sixteen hundred Radeon cores and six gigs of gddr5? Oh, and can you do it on your new 24nm node?" "Sure, sir. That'll be one thousand forty two dollars." "Wow. Too much... Let's see, hmmm. Well know what? I don't need so much X 86, can I drop the ivy bridge cores down to sandy and cut out an extra module?" "Sure, that'll be seven hundred eighteen dollars." "Cool, my shipping address is..."

Too bad it's much more intensive than that to design a chip and licensing and patents would get in the way of combining architectures like that.

avatar

Hey.That_Dude

LoL cakes. It take over a year to just develop that new decoder for h.265 (not test, just implement, test means waiting 3 months to get it back from the fab, then testing...). Not to mention everyone uses a different set of logic with different timing constraints. This stuff is really complex. Mashing it all on a single chip sounds awful.
Plus they all use completely different architectures. Ivy Bridge, Piledriver, and NVidia's GK110 all have different timing methods. Are you going to devote three layers to three different clocks or are you going to have one layer sectioned off into three parts and hope that the increased resistance doesn't screw your timing?
The engineers make it look so simple but it's really REALLY not. Much better to have a known bus that can act as a median... hopefully you don't overload your bus with that stuff you want.

avatar

AFDozerman

Oh, no. I understand. It was just one of those pipe dream thoughts. You know, those pothead "what if's"

avatar

Hey.That_Dude

Lay off them brownies man... people start dreaming up ideas and companies will bend over backwards to bring it to market(even if it makes no sense *cough* Windows 8 *cough*).

avatar

AFDozerman

Wachoo talkin bout willis? Lots of good things were created my gluing together two opposing things together; Jumbo shrimp, the el camino, that show catdog! All great things. XD

avatar

Hey.That_Dude

ARM chips must all be fabed at the creator's expense. ARM is fabless. Plus customization isn't that difficult since you already have to send out the design in veralog/VHDL.

avatar

Zstreek

It is hard to believe that there is sufficient benefit, in a custom chip over a stock chip,to justify the expenditure of $500B. My wife has the iphone 5 and I have the Galaxy Note II. I can't say that there is a real speed or utility difference between the 2.

I completely support this kind of development if it yields results. Wouldn't we all love to see a quantum leap in processor technology? But as the world stands today, apple products are still behind the curve. Not by very much, but from a hardware standpoint at any given price they are lacking in comparison to the competition.

Best of luck to them, I hope it works out. I will always support the best product and I reject brand loyalty.

avatar

Renegade Knight

The only thing that would justify the custom chip is if Apple designed in some custom features that made their phone work better than everyone elses or do something that nothing else could do. Otherwise it's just hubris.

The only other reason I can think of is that maybe it keeps iOS from being ported to other hardware like is done with Hackintoshes.

avatar

Zstreek

I agree. And I do not believe in the protectionist philosophy of business. Hardware and software should be open to modification and use by any who own them. Otherwise you stifle new and exciting ideas.

Log in to MaximumPC directly or log in using Facebook

Forgot your username or password?
Click here for help.

Login with Facebook
Log in using Facebook to share comments and articles easily with your Facebook feed.