AMD helped mark a milestone for a new era in which the venerable x86 and even microprocessor architecture no longer reign supreme.
That leaves Intel as the last man standing -- the last chip maker with its fortunes staked to the x86 processor architecture.
It’s not a bad place to be left standing, really. The market is still the second largest in non-memory semiconductors with annual unit sales measured in hundreds of millions. Its 80 percent-plus share of it makes Intel easily the largest semiconductor company in the world. Unlike the largest non-memory market of smartphone processors, it has average selling prices that range from about $100 for client CPUs to more than $1,000 for server chips.
But those prices are eroding like bad California beachfront property. ARM is on the rise with a penny- and power-pinching architecture that enables chips for clients and servers both selling in the low-rent neighborhood of about $20. Soon they will even include 64-bit varieties with performance nearly on par with members of the x86 family.
At this inflection point when we are watching a slow motion shift of monarchies from x86 to ARM, it’s worth remembering we are also watching the decline of microprocessor architecture. Increasingly system design is all about the software and the user experience.
I asked a long time Lenovo and former IBM tech executive about the big semiconductor design issues for mobile systems. There really aren’t many, he responded. The system issues are all about sensors, industrial design and software, he said.
Once upon a time, engineers made their livings designing the nuances of CPU pipelines the way great architects made skyscrapers and suspension bridges. I attended industry conferences where they explained the refinements of their works amid great competitive pressures. They were feted like Rembrandts and Da Vincis.
Those days are as much part of the technology past as the giant mainframe cabinets and consoles with their built-in ash trays in the Computer History Museum just down the street from the headquarters of Intel Corp.
But that doesn’t mean processors, cores and SoCs are dead. Far from it.
Brilliant engineers will be needed for as far as anyone can see into the horizon to design the chips that drive tomorrow’s personal and cloud computers. But microprocessor architecture and the x86 specifically no longer reign supreme.
So the x86 is dead. Long live the x86.
The Sage mainframe console sports a built-in ash tray (lower left).
Well done Rick - thank you.
An interesting angle, including for Intel, will soon be - a standalone processor versus SoC...
With all the accompanying implications - including packaging, multiple die (SiP) packaging, etc., etc.
Once all of the competing CPUs reach a certain point of diminishing returns, the differences that had been so important in years past lose relevance. At that point, things like support of the installed base of software become more important.
Still, the typical x86 powered machine is vastly overpowered for the needs of many users, hence the success of tablets. There will always be extreme users, such as gamers, analysts and engineers that will need that additional power. It remains to be seen if ARM vendors will be able make any inroads into that set.
I'm not a uP designer, but it seems to me that all we are seeing here is more of the same. Did not the x86 essentially become a risc processor back in the original Pentium days, with the complex functions added peripherally?
The risc processors (ARM) started to focus more on saving power rather than increasing performance.
Can't x86 morph again?
An ATOM 2.0 with x86 instruction set? That's basically where I'm coming from.
Also, Duane, I don't think the x86 is vastly overpowered compared with what the average user needs. The reason being, it is the up-to-date productivity apps such as Office or Adobe Acrobat, never mind the games, that demand the power. It's not that the user is writing CPU-intensive programs. At least, not in most cases.
Rick, you continue to amaze me with your dumbness and stupidity and also, your crappy headlines. You really think $50B revenue generating architecture is dead! I guess you got a new doze of ARM Juice from the ARMTimes sponsored conference. You have no clue and cant believe UBM has bozos like you writing articles. You bring shame to EE that he has to read your article
It is quite right that technology never dies, and the x86 is that much penetrated in the computing environment it will go on getting used on the consumer PCs, Yes it may happen that Intel will have to reduce the prices of their processor compared to at what price they are selling present days.
The ARM is not necessary power efficient then x86. I think people just over believed in arm, and kind of thinking, anything from arm must be low power processor.
The newest Atom "medfield" is actually comparable with newest ARM "krait" in power efficiency (if not better than ARM). The almost identical design motorola Raza i and m, are compatible at real life usage, yet the power consumption of atom is much lower than the krait. Consider this, that the atom is 5 years old architecture on a 32 nm silicon node, the krait is like 1 year old on a 28 nm silicon. I really don't know why most people still has so many confident in arm about "low power".
The ARM a15 from Samsung, is considerable faster than the medfield, but it's a competition between 5+ w soc and 1+ w soc.I am wander why both apple and qualcomm use their own arm design, if the a15 is really as good as arm declared. Reminder me just the declaration of ARM last year that "the x86 is impossible for mobilephone", "the x86 is much more power hungry than arm". which lead me to further concern their reputation.
I get what u mean. But, in terms of Medfield's single core with hyperthreading, how much performance can they squeeze out from a single core??
We have the Moore's law for a reason, in order to overcome the power wall, more cores have to be implemented. All i can say is Intel is just "adopt and modifying" design from their desktop counterparts which to me is a lot of work that needs to be done.
People have very short memories.
From early 1990's
"Death of x86 due to MIPS...."
From mid 1990's
"Death of x86 due to PowerPC...."
From late 1990's
"Death of x86 due to DEC Alpha...."
"Death of x86 due to Itanium...."
"Death of x86 due to ARM..."
And the pattern is...
I totally agree. And that was my point as well.
"The pattern is," as you so aptly put it, that x86 soon beats the newcomers at their game.
Hasn't happened yet in the mobile field, but there's no reason to leap up and claim they're finished.
Some would say its an ecosystem game, not an architecture play.
By the time Intel has a fully fleshed out line of mobile SoCs running all the cool mobile software, ARM will have the same in servers.
Duopoly dead ahead.
Rick, Get your head out of the gutter...i mean ARM world. Razr I beats Razr M (Qualcomm) on battery life by 4 hours. Check out the web to get 3rd party reviews. This is just the beginning from Intel. Keep on dreaming about ARM. Its just a matter of days