Will Microprocessors be able to compete with chips?

Unknown

I saw this quote:

"Intel says that the upcoming Atom processor innovations will actually outpace Moore’s Law, accelerating from 32nm through 22nm to 14nm within three successive years."

Generally, the Atom processor is being placed in smartphones, notebooks and tablets. It is now becoming super thin and is already fanless. The thought occurred to me that markets might not be able to compete. I'm thinking about the $100 calculators that nobody wanted because of the transistor based calculators. Can chip companies compete against billions poured into research by the big boys?


from:
Intel is Stepping on the Gas for Atom CPU Plans

http://www.tomshardware.com/news/atom-cpu-plans-medfield-moorestown,12829.html

kwinn

Intel is one of the "chip" companies and there are lots of applications for less complex computer chips, microcontrollers, memories, programmable logic chips, analog, and interface devices etc. Plenty of room in the marketplace for the smaller companies. Parallax is a perfect example. There are also fab companies that produce chips designed by and for other companies and universities.

Heater.

Not sure I understand the question really.
What is a "chip company". In the computer world we find AMD does not make it's chips any more they have no "fab" plant. Similarly ARM do not make their own chips. These guys only design them.

How many microprocessor companies still design AND manufacture their devices? Intel and IBM are the only ones that come to mind.

So, x86 is coming to the smart phone and tablet. Is there no escape?!

P.S. I just found that perhaps the fastest microprocessor in the world is the IBM z196 with 4 cores and 6 execution units per core running at 5.2GHz. Never heard of it before.

Cluso99

So Intel expects to be producing 14nm. How long before they get to 2nm? Recent article says researches have a 2nm transistor. WOW.

But how long will these chips last? On the recent UPEW video, someone said that the metal flows and the new chips are only guaranteed to work for 10 years, even if never used. Soon we will have planned electronics obsolescence that only lasts for the 2 years warranty and a few days, just like car batteries!

Beau Schwabe

Cluso99,

"On the recent UPEW video, someone said that the metal flows and the new chips are only guaranteed to work for 10 years, even if never used." - Not exactly... DRC rules are designed so that based on metal migration 'during use' that it would be 10 years before you started experiencing problems. That is also based on 125 Deg C operating junction temperature. During the layout design you can certainly place the DRC rules further apart to increase that 10 year window.

The problem that I see with going smaller and smaller, is that the rules change as far as how the silicon responds. Characteristics that were once recessive are now dominant and become a real design issue the smaller you go and behave very differently than their larger so called equivalents.


Edit: taken directly from from the data sheet .... "These rules for aluminum alloy guarantee 10 years use life, with less than 0.01% cumulative failure (t0.01>= 10 years). For longer use life multiply the maximum current by the square root of (10 years / use life)"

The above translates in a way that based on the amount of current you require in your design the thickness and separation distance of the wires scale by a value determined using the above formula.

Note: the use life is determined by the fabrication process that is used and based upon their rules for the process.

HollyMinkowski

Heater. wrote: »

So, x86 is coming to the smart phone and tablet. Is there no escape?!

I've had to work with x86 before, I can dust off the notebooks and pick up where
I left off. If I can survive ARM coding I can suffer through x86 some more.

I don't like the sound of a 10yr chip lifespan at all :-(
I guess it's from the heat? I suppose it really doesn't matter since everything
is obsolete a few months after you get it...but a limited cpu lifespan just isn't
kosher. If I had a choice between two cpu's for a project and one was only good
for a decade while the other was good for centuries like a cpu should be, I'd reject
the short lifespan one even if it was cheaper and even if the equipment only needed
to function for a few years. The 10yr lifespan makes me look at the cpu as broken,
and broken just goes against the grain for me.

Beau Schwabe

"I don't like the sound of a 10yr chip lifespan at all :-(" .... Folks, I >really< don't want to confuse this issue here, this is an industry standard. When a process is defined the DRC rules are set to 10 years at 125 Deg C ... since >we< are no where close to running our chip at 125 Deg C, we are likely to see 25 years or more, perhaps even 50 years service from our chips.

Assuming everyone understands that glass is always a liquid, at normal temps it takes years to move (or migrate), the Metal inside a chip is the same way... when current passes through the wire in any amount, it causes the metal to flow or migrate. If the DRC rules are not designed so that they are spaced far enough apart, eventually the metal will migrate in such a way that it can cause a short with it's neighbour. Obviously current and time play an important role, since the current is directly proportional to the temperature. Increased temperature equals increased metal migration and ultimately a shorter lifespan for the chip.

localroger

One of the things Chip pointed out at UPEW is tha power consumption goes through the ceiling as the insulation barriers get smaller. Sure they can cram a 80186 on a chip the size of a 74LS00 die, but how much power will it need? One thing the industry is really only starting to grok is the usefulness of battery life. People still use TRS-80 model 100's for real work even though they are hopelessly obsolete because of 40 hours on 1 set of AA batteries. Until recently absolutely nobody made laptops that could go more than 2 hours or so on a charge (or netbooks at all). Now there are several models that can go 10 hours. But lots of us would forego the backlight and graphics to get back to 40 and the ability to use disposable batteries in a pinch.

HollyMinkowski

@Beau

I understand now, so the metal can eventually flow and short out.
Kind of like the problem with tin whiskers :-)

I suppose that explains the 500+ year likely lifespan for old x86
chips that I read about in an older tech book...those chips probably
had lots of space between the metal connections inside the chips.

We just have to move to optical data handling inside processors that
are formed from vertically stacked cores since making the wiring smaller
is causing heat, power and longevity issues now. And spreading cores
out horizontally in 2 dimensions is obviously not going to carry things much longer.

Beau Schwabe

HollyMinkowski,

"...the metal can eventually flow and short out...." - that's correct :-) ... it can also flow and cause an 'open' which is sometimes the cause of seemingly random intermittent behavior.

"...that explains the 500+ year likely lifespan...those chips probably had lots of space between the metal connections inside the chips..." - exactly

There are metals other than Aluminum that 'hold' their shape better against metal migration, but you pay for the price ... an Aluminum alloy mixed with Copper can greatly improve the resistance against metal migration. Copper alone is also an option, but it's difficult to process making it more costly. Other alloys with Copper can be used to increase the resistance further but at yet an even higher processing price. The biggest challenge that we face in the future isn't so much moore's law... it's the Quantum level creeping ever closer to our back door. At this level characteristics that were once recessive (we were aware of them but we could design around them) are now becoming the dominant feature where what once was a dominant characteristic is now becoming further and further from our reach.