NVIDIA’s Kal-El processor will be a 5-core low power chip

vSMP

NVIDIA has been talking up the company’s upcoming Kal-El processor for months, promising better performance and lower power consumption in its next-generation chip for mobile phones, tablets, and possibly even Windows 8 computers. Up until now NVIDIA had suggested that the Kal-El processor was a quad-core chip, but it turns out there’s actually a fifth processing core tucked away in there as well.

The company revealed the presence of the extra core in a white paper on multi-core computing released today.

Here’s how it works: there are four ARM Cortex-A9 processing cores that can be run at high speeds when necessary. But there’s also a new, lower power processing core which NVIDIA says uses vSMP (PDF link), or Variable Symmetric Multiprocessing tech.

This processing core can take over when your device is in standby or when you’re just listening to music or performing certain other tasks that don’t require much power. It will help prolong the battery life of your mobile devices.

While you might think that a quad-core, or five-core processor would use more power than a single-core chip, the opposite is actually true. That’s because you can run all four chip cores at 300 MHz and use less power than you would get running a single-core processor at 1.2 GHz while achieving similar performance. Work can also be divided between cores so that a single core doesn’t have to handle every job, and you can also have the chip to shut off processing cores when they’re not in use.

We could start seeing devices with Kal-El chips later this year or early next year as NVIDIA pushes out the successor to the popular NVIDIA Tegra 2 dual core chip which dominates the Android tablet space right now.

via Android Central

  • http://twitter.com/V_U vēer

    Thats awesome, Im waiting to see multi core ARM CPU’s invading full powered laptop space, now that is going to be interesting, given ARM size and power consumption we could see some remarkable new designs in fully featured laptop designs. Intel will have run for its money, thats for sure ;)!

    • Anonymous

      There are still some hurdles for ARM to get around. 

      1) They need to adopt a good desktop OS to enter the laptop market.  Aside from Chrome, they can’t get serious in competing until Windows 8 comes out. 

      2) ARM processors are still 32bit, limiting their upper performance and what features they can offer.  While despite all these improvements they’re still only rivaling low end Intel ATOM for processor performance.

      Though like AMD they can potentially offer better GPU performance than Intel is presently providing at that range as a side consideration.  Though most are using GPU’s not designed to support a desktop OS and that has to change as well.

      3) ARM is plagued by hardware fragmentation, since each manufacturer can freely customize their own version of a given ARM system.  While many hardware manufacturers only provide closed drivers that makes making software work on all the different hardware a real pain.

      So harder to get a OS to work on all hardware without either not bothering to be optimized for that hardware or picking and choosing which hardware they will support and locking out all other users.

      Though MS has imposed specifications that ARM manufacturers hoping to offer Windows 8 systems have to comply with to help limit that fragmentation.

      Nvidia uses closed drivers and the Tegra Zone is an example of either use them or lose out on these features type business model they are using.

      Remember, the downside to SoC is you can’t just replace a part that you can’t get working the way you like.  It’s all or nothing deal and close drivers means you have to rely on a 3rd party to get things fixed and it’s a lot harder to optimize everything.

      While Intel is also moving to SoC for their mobile offerings, they’re still one company with one common design to deal with versus the zoo of offerings from ARM.

      Basically, the ARM business model works great for the mobile market.  Where people don’t mind getting a new product every single year.  They can even be completely niche and still sell well. But the laptop market though doesn’t change that fast, needs to meet the needs of a wider range of consumers, and people expect more value for what they pay.

      Though there are other markets ARM is likely to do much better and they do plan to get ARM in just about anything you can imagine.

  • aftermath

    This is not new.  This is not news.

    I’m not an expert on NIVIDIA stuff because it is driver encumbered garbage (isn’t it “interesting” that of the tree major players in x86 GPUs, Intel, AMD, and NVIDIA, only the one who has purely proprietary drivers, NVIDIA, is heavily pushing ARM-based, proprietary garbage?), but…

    …I’m pretty sure that of the 8 cores in Tegra 2, there are three CPUs cores: a dual core A9-cortex ARM and a lowered A7 for handling things like (low power) interrupts and idle state.

    • Anonymous

      It *is* news.  Check out the diagram…that A7 is still there as well.  The A7 in Tegras is purely a behind-the-scenes SoC management chip, and doesn’t act as a “normal” CPU the way that the companion core can.

      • Bob

        It’s not an A7 there’s no such thing …. its a power optimised A9.

      • Anonymous

        It’s not an A9.  It’s an ARM7.  Calling it an A7 is probably semantically wrong, but it’s not another Cortex-A9.  Look at the layout diagram, just left of Core 1.  There are *five* A9s, and one ARM7.

  • Mr. T

    “That’s because you can run all four chips at 300 MHz and use less power than you would get running a single-core processor at 1.2 GHz while achieving similar performance.”

    I know this is what all the marketing for multi-core processors claims, but it’s not strictly true. If my application expects serial processing, four cores at 300 MHz is exactly the same as a single core at 300 MHz as far as my application is concerned. Depending on the number of other processes running, my application’s scheduling priority, and the OS’s scheduling algorithm, a single core at 1.2 GHz may appear anywhere between slightly slower and four times faster to my application than four cores at 300 MHz. Multiple cores are great for some problems, but lousy for others.