VMware CEO: Intel chip design too complex

In a recent and so-far relatively obscure video, VMware CEO Paul Maritz offers, at times, a sharp critique of the Intel chip architecture and the challenges of getting it into cell phones.

VMware CEO Paul Maritz

(Credit: EMC)

At the TiEcon 2009 conference in mid-May, Maritz gave a brief oral history of the Intel x86 chip architecture. He noted its shortcomings and the challenges presented by ARM, the chip design that powers most of the world's cell phones and will power Netbooks running on Google's just-announced Chrome operating system.

The video has been made available by TechPulse360.

"Consumer devices came along and there was one problem with the x86 instruction set. All of that complexity in there, accumulated over the years, meant it's a power hog. It loves electricity," he said in the video, referring to Intel's x86 architecture, which virtually all PCs use today.

Maritz--who worked for five years as a software and tools developer at Intel before spending 14 years as a top-level executive at Microsoft and then joining VMWare a year ago--continued: "In consumer devices like phones you can't have that. Battery life becomes paramount," he said.

Maritz described how Intel experimented with the ARM processor architecture and bought a license "for a much simpler microprocessor," referring to Intel's development of the StrongARM architecture, which eventually became a designed called XScale.

Subsequently, Intel decided to get out of the business, according to Maritz's depiction, because the devices were "low-end, low-power, low-profit." Here Maritz is referring to Intel's XScale business, which was sold to Marvell Technology in 2006.

Maritz continued, describing how Intel wanted to get back to its roots: "high performance, complex microprocessors." Then, Intel realized, according to Maritz, that it had to get back into that market: "This ARM thing is a real problem, we're going to have to go back into that space."

In response to the video, an Intel representative said: "Paul Maritz is not privy to all of Intel's future product plans."

Note: Though the video is from May, it did not come to my attention until very recently. I think the topic is important enough to bring up now because Maritz is a high-profile CEO at a large company that builds software that runs on Intel processors and because he's speaking about one of Intel's greatest challenges.

[Lerianis3]

The fact is that these 'high-complexity' processors are the mean because they are the ONLY things that will do everything that a normal user throws at them.

Let's get real here: processors in phones are doing MAYBE 1 or 2 things at a time, that are very similar, outside the iPhone.

[uhpl508]

You mean because the iPhone can only do 1 thing at a time?

[ikramerica--2008]

The iPhone can do many select things at a time. Run iPod while doing email as the phone updates contacts in the background, for example. Just can't run two third-party apps at once for now. And not because of the ARM, but a choice by Apple.

[ewsachse]

If the x86 architecture was so complex and inefficient, then why did it out last every RISC design from the 1990's and 2000's?

[cosuna]

Ever heard of "vendor-lock"? Microsoft didn't commit fully to any RISC. Most of NT's implementations on RISC were PoC-like (PowerPC, Alpha, even Itanium).

Microsoft has no idea how to handle multiplatform. Just ask HP on the problems it has had on its laptops that use 64-bit Vista.

[zyxxy]

Power is still be shipped by IBM, Sparc by Sun and Fujitsu. ARM is everywhere you look. No one uses x86 for engine/chassis management in cars. XBOX 360, PS3, and Wii are all PowerPC based. MIPS is embedded in a lot of networking gear. Hmmm, all RISC based. So what was your question again?

Oh, why does Intel own the **desktop**? Windows. No other reason.

I think there is every possibility that ARM based netbooks will out number Intel based netbooks five years from now. And Intel will still own the desktop.

[getwired]

Hardware and software commoditization. Same reason why we all use unleaded every day. cosuna isn't quite spot on - the NT ports to PPC, MIPS, and IA64 were all hard-core ports, and were very painful. Alpha, which SHOULD have been a great port, was a horrible one. The lack of true 64-bit support (it used DEC hardware emulation to run x86 code on Alpha CPU's - which meant little performance gain) killed NT on Alpha. That was, IMHO PoC-like.

None of the other architectures had the backing that x86 has had from the get-go. Only Apple and IBM got behind PPC, only HP and Intel were ever in, and are still in, the IA64 game (an architecture that needs to die).

I'd actually argue that HP's problems with 64-bit are pretty huge on their own - they can't decide where x64 and IA64 belong, respectively. They need to pick a horse and go with it.

[plbyrd]

I find it rich in irony that the company that bases its entire existence on emulating a technology it CHOSE TO EMULATE is complaining about the complexity of the technology.

Give me a break. It's the same old Wintel is evil argument packaged up to look like a pro-phone piece. What exactly is VMWare doing in phones and why do they matter in phones? Are they going to try to run Windows 7 on a cell phone with a 400mhz ARM or something?

[ikramerica--2008]

Absolutely, not only is his claim that the legacy x86 architecture makes it inefficient bogus, but his time spend at Intel and MS is also worthless, as is his experience as CEO of a major company that deals with this everyday.

Sarcasm off.

The man knows what he is talking about.

[getwired]

They emulated it not because it was efficient, and not because it was a good design. They emulated it because it was a commodity architecture that was killing customers with the entire hardware complexity "package".

The ARM processor is a remarkable architecture - but it has the advantage of not having nearly 30 years of backwards compatibility engineered into it. Honestly, I feel like the Mac OS has much more simplicity and elegance in it's innards than Windows does today - but it does so at the cost of backwards compatibility.

[libertyforall1776]

Funny, Intel seems to only try to be an innovator when it suits them, or competitive pressure pushes them. Remember the "iTanic"?

http://www.theregister.co.uk/2004/02/17/who_sank_itanic/

[getwired]

Regrettably, the iTanic still isn't under the water yet.

[EmbSysPro]

Intel has great technology and Itanium is an impressive design. If Microsoft had invested in Itanium early in its initial marketing cycle there may have been a different outcome.

Intel is in the business of making money first and foremost. Any innovations it creates are ONLY as a side effect of the pursuit of wealth. And that why they are in business right? To make money for their share holders?

I'm a die hard embedded MIPS and PPC guy. But if Intel can successfully execute on its 32nm/22nm x86 strategy and produce a true sub 1W SOC @ 1GHz or better then things will change.

[BethInAllen]

The ARM architecture is limited to 32-bit addressing i.e. 4 GB. Current Netbooks already come with 2GB. Unless ARM adds 64-bit addressing, it will be stillborn in the Netbook or Nettop marketplace.

[monkeyfun14]

With ARM chips having less processing power then a P3 why do they need 64bit addressing?