20 Comments

Summary:

Nvidia is demonstrating a quad-core mobile device chip at the Mobile World Congress, and by one benchmarking standard, the new chip is faster than Intel’s 2-GHz Core2Duo computer processor. As impressive as the new Tegra 2 is, devices with Tegra 3 could arrive this year.

coremark-kal-el-featured

Nvidia is currently demonstrating a quad-core mobile device chip at the Mobile World Congress (MWC), and by one benchmarking standard, the new chip is faster than an Intel 2-GHz Core2Duo computer processor. Currently dubbed Kal-El, the chip will likely be known as the Tegra 3 processor when it begins shipping to device manufacturers in August of this year, which is ahead of competing chips from Qualcomm and others. Regardless of which mobile chip designer wins the race to offer a mobile chip with four application processors, this news is another blow to Intel, which just lost a fair amount of software support from Nokia’s decision to partner with Microsoft.

While such a benchmark test is only one artificial indicator of chip performance — device speed is also influenced by memory, operating system and other factors — this news from Nvidia has to rile Intel. The company sold off its Xscale mobile chip division just prior to the smartphone revolution, which is now looking like a strategic mis-step. And in fairness to Intel, the new Nvidia chip is being compared to an Intel processor that launched in 2006; Intel has made positive strides to optimize its silicon for mobile devices since then but is still shut out of the mobile handset space. Even when you remove company names from the scenario, the benchmark test illustrates just how powerful mobile processors are becoming when they can offer performance similar to that of a laptop from just a handful of years ago.

Indeed, that theme sets the stage for mobile devices with even greater capabilities than what’s being shown at the current MWC event. Earlier this week, my colleague Stacey noted what advances we can expect from multi-core processors in mobiles: greater processing prowess for apps, yes, but also support for more cameras, both high-definition and 3-D, and improved gesture interfaces. More cores also helps speed up browsing, as Nvidia demonstrates in this video by enabling all four cores in Kal-El on an Android reference device:

Even as the current Tegra 2 impresses and is powering numerous new smartphones and tablets, Nvidia is packing their next new chip with 12 graphics cores. The company’s roadmap suggests new chips annually, with up to a 75x boost in performance over the current Tegra 2 by 2014. Indeed, the new Kal-El chip demonstrated 1440p video playback on a 2560×1600 resolution display, continuing my thought that mobile devices will slowly become the home set-top box for video entertainment and possibly even for gaming.

Related content from GigaOM Pro (subscription req’d):

  1. This is why Intel has no chance of entering the mobile/tablet markets. By the time they manage to optimize a watered down Atom chip for energy efficiency, Nvidia and other ARM chip makers will surpass it in performance. So now they actually have to get worried about making their i3-i7 line-up as efficient as ARM chips, because you can bet Nvidia is going after them with Tegra 4 in 2012 and Tegra 5 in 2013, especially once Windows for ARM is ready and will start shipping in laptops with ARM chips.

    Tegra 3 should already be enough for most people as far as performance goes. After all, the Core 2 Duo is the chip inside MacBook Air, right?

    Intel made the same fatal mistake that Nokia did. They didn’t realize, or refused to believe that ARM (or iPhone/Android in Nvidia’s case) are disruptive innovations, which basically means there’s no way you can fight against it with your old technology, just by improving on it.

    Intel’s only chance to remain relevant in the next decade is to start making ARM chips, which of course, is something they don’t want to hear, because that would mean huge conflict of interest for them.

    Their StrongARM division failed because of the incompatible business model of ARM chips with their cost structure. Intel was used to making a lot of money on their chips, while ARM chips had to be sold for a lot less. Even Atom is not exactly the type of chip they love because it doesn’t bring them that much money, and it’s why they haven’t improved it much for the past few years. They just didn’t want to make it powerful enough to be used in laptops. They make a lot more on their i3 low-end chips for laptops. It helped that they also locked in the netbook market, so they didn’t have anyone to compete with.

    So for them it seems “easier” to try to improve their x86 chips, even though in the end they have no chance of succeeding, just like Symbian never had a chance against iOS and Android, and it proved that they just wated 4 years for nothing. The same will happen with Intel. Two or three years from, we’ll look back and see how much time Intel wasted trying to improve Atom to compete with ARM chips.

    I say it’s a good thing Intel will no longer be relevant in the next decade, because it looks like the mobile chip market is doing fine without them, and competition in the ARM market is strong because of the ARM licensing model. Chips seem to be doubling or tripling in power every 12 months, which is almost twice as fast as in the PC market. I think this is happening because of higher competition, but also because the industry doesn’t have to relearn making chips that powerful. They mostly need to use old technology but in a smaller package in and a more efficient way.

    I remember 2 years ago I was thinking the phone is as powerful as a PC 10 years ago, last year I thought it’s as powerful as a PC 7 years ago, and now with Tegra 3 they will get as powerful as a PC 5 years ago. The gap is closing fast on Intel.

    Share
    1. Spot on!

      Share
    2. Very well put. Intel has completely missed the boat on Smartphones. It’s a classic dilemma faced by many established companies (read Microsoft) 0 do you continue on the path that made you successful for many years or take a risk in moving to newer path.

      Intel still may have a chance to succeed – not on its own; but buying into this market. What do you think about them buying Nvidia. It not only gives them entry into ARM based chips; but also nice graphics functionality – two areas that will see more and more action in the coming years.

      Share
  2. you do realize that we speak about a 5 year difference here, right?

    the tegra chip you are speaking about which nvidia “claims” is faster then a chip produced 4 years ago = 5 years difference….

    my money is on intel: http://www.youtube.com/watch?v=iv45xll5iBw

    Share
    1. Core 2 Duo T7200 is produced 3 (yes.. three) years ago, has 34 TDP envelope while Tegra 3 will have at max 4 watts TDP for tablet/smartbook systems and a much lower TDP for smartphones.

      Share
  3. From my post: “And in fairness to Intel, the new Nvidia chip is being compared to an Intel processor that launched in 2006.”

    So, yes, I do realize we’re speaking about a 5 year difference. ;)

    I’ve also been following Intel’s progress to pare down the power requirements of its chip, which it’s actively been doing since the first Atom in 2008 but still has the underlying chip in zero shipping phones. The vast majority of smartphones (which now outsell PCs, mind you) use ARM architecture, not Intel’s x86 architecture. In other words: there’s far more data to consider than Intel’s marketing pitch at MWC.

    Share
    1. Kevin,
      thanks for your fast reply. Indeed you did point it out and I also had read it.

      I needed to mention it though as 5 years is like an eternity in mobile technology.

      What phone did you use 5 years ago?

      What I am looking forward to is the possibility to access the same content seamlessly on my tablet, my PC, my Smart Phone, my TV and my Car. Still a while to go but since we quoted 5 years above – I don’t think we will have to wait that long :)

      ARM.L had 500.000.000 USD of annual revenue (last available financial report) while Intel spent 6 Billion on R&D. I truly believe Intel will catch up :) and will be a viable competitor.

      Share
      1. Five years ago, I let my blog readers choose my phone for me and at the time, they picked the XV6700 on Verizon Wireless. Ironically, that ran a on 416 MHz Intel PXA 270 XScale processor! ;) No argument that 5 years is several mobile cycles, but the ARM chips from 5 years ago weren’t even close to competing with x86 chips – now they are as the gap continues to shrink at a faster rate.

        Share
      2. This five year difference is wrt to performance of the ship and one should not confuse it with physical years. As someone pointed out earlier, just a couple of years ago (around iphone 1), the mobile processors were comparable to the desktop processors of the late 90′s. But in a couple of years, that has already shrunk to 5 years. In another couple of years, if nvidia keeps its promise (100x tegra 2 or 20x tegra 3 or ~20x core2duo), Intel should be really worried.

        I would love to have a laptop as light as my macbook air that is only 20 times more powerful.

        Share
  4. The most impressive thing about NVIDIA’s ARM chips are their ability to handle graphic. I’m using the 2GHZ Intel chip they tested against in my laptop and no way can it handle full 1080p. So, while Intel’s chip is older NVIDIA is making a great GPU/CPU combination that will give the user a better overall experience.

    12 GPU cores in this thing!

    Share
  5. From what I’ve seen these chips are also OpenCL 1.1 compatible. Does that mean they could compile OSX to run on them? I don’t know much about different hardware architectures. It would be an interesting CPU option for the Macbook Air :) (I guess Apple will make their own. But…)

    Share
  6. BTW, anybody can claim any performance number on stupid benchmark. I am sad that articles like this doesn’t spend enough lines on explaining how valid the benchmark is, before explaining the “Scores”. Nvidia always over promises and under-delivers. This is probably the same. Why didn’t nvidia show Spec scores?

    Share
    1. Agreed that benchmarks aren’t the “end-all, be-all” of performance claims – I noted in the post that there are many other factors involved. I consider benchmarks to be a high-level indicator of what to expect, but real-world usage is ultimately what matters. As far as the validity of the benchmark, it was created by the Embedded Microprocessor Benchmark Consortium and is considered by some to be a standard tool to compare core performances of chips.

      Share
      1. Have you looked into the fact that the so called Coremark benchmark from the said consortium is so great that, anybody can post any scores on any platforms? You can make up Coremark scores! Let me say it right, “You can simulate the benchmark test in your head and come up with the score and post it on the Coremark website!”. Based on my mood today, I can make tegra a shitty platform today or put it on par with Supercomputers and post the scores on Coremark website! No authentification needed! Have fun!

        Share
        1. I’d imagine any benchmark test can be simulated and faked, although as a public company, Nvidia has an obligation to its shareholders that certainly should help prevent that. Regardless, instead of focusing on the benchmarks — which I’ve already said are one of many performance indicators — why not turn the conversation into something constructive with opinions on the actual demonstration results shown in the video or the ability for the reference design to pipe 1440p to an external monitor and integrated display at the same time?

          Share
  7. Intel stayed ahead in the semiconductor market because Andy Grove always paid attention to market trends before they were apparent to the rest of the chip makers. Under his leadership, the company made numerous bold moves that were seen as high risk yet resulted in major steps forward.

    Since Grove left, Intel has been led by people who don’t have the vision or insight that Grove always provided. The bland leadership has made Intel just another semiconductor company that will bounce around in the market looking for a place to recoup its huge investment in advanced manufacturing.

    It is this change in the company dynamic that has resulted in a complete fail on the mobile market and will continue to dog it until there is new leadership. That’s not to say that any of the other processor manufacturers have the kind of visionary leadership that is necessary to be more than a commodity manufacturer – different companies will look strong at different times with nobody sitting on top.

    Share
    1. wow – seeing AG’s name after a long time gave me a (happy) pause. he was the greatest – true visionary and an execution machine. rare combo of everything. my hat tip to him.

      Share
  8. Kevin,

    Intel has to addaress 3 problems to address.

    1) PRoblem #1: The smartphones and tablets value communication and media, processor is an afterthought. That puts value to QCOM who is innovating on the ‘connector’ and Nvidia who is innovating on the ‘media’.
    2) Problem #2: Leadership. Paul would retire anyday. Sean got into a medical problem. Rest of the leadership is not strong to take Intel to the next level. Moreover. Intel has always grown leadership form within and that style is no longer viable in the new era. Need aggressive, fast turn around SoC with innovatin on both hardware and software.
    3) Problem #3: Smartphones and tablets are eco-system plays. Not just Processor, or operating system. Like what happened to Alpha (2X over others), with SPARC (2X over other for throughput), the architectural wind has moved to eco-system and Intel is not just on the wrong side of architecture, but on the platform and eco-system.

    How to address all 3 of them?

    Buy Nvidia

    1) Buying Nvidia addresses the Hardware (ARM/SOC, media processing).
    2) Buying Nvidia addresses leadership. The culture of Nvidia bring the right CEO (Jensun Huang – hard knocking driven given) and a culture of aggressively beating out competition. Nvidia came from #27 in graphics to #1. That is its culture. Intel’s culture is win around its defensible turf (x86 and process). Intel needs an outsider CEO. The inside leadership don’t cut it
    3) Anand Chandrasekhar despite having had success with Centrino, failed on a number of things with Atom, Smartphones, tablets. Bet on Mobilin, on Meego. Too much NIH. Intel has all the assets with Nvidia acquisition including leadership and everything non-apple (Windows eco-system and Android). They can take a pole position in tablet outside of iPAD and work their way.

    The ARM vs x86 is a pointless arguement. In my mind its a religious argument. There is no hard evidence to say that ARM is anything greater than 1.25X over any x86 w.r.t power/performance. Intel has lost the marketing war on this war. ANycase, its no longer the ISA wars. Its media, connectors and eco-system. If you don’t have it, you loose. QCOM lacks the media and both Intel and QCOM are even keel on eco-system.

    An Nvidia acquisition will make this a 2 horse race with QCOM. Otherwise expect QCOM to overtake INTC market cap within 18 months. Time is now. I am sure Jensen will take the offer if he is the CEO and that is the right thing for both companies. It also bolsters Intel’s Server/HPC/graphics. Its no longer an FTC issue. (AMD has ATI)

    Share
  9. [...] iPad has the same 10-hour battery life. Which is good, because already, Marvell, Qualcomm, Nvidia and others have announced quad-core ARM-based cores for inside tablets and other devices with parts [...]

    Share
  10. [...] early to predict what the tablet market will look like several device iterations from now due to powerful new processors on the way, faster mobile broadband in wider coverage areas and improvements in mobile software and [...]

    Share

Comments have been disabled for this post