Chip analyst Mike Feibus: 'If you buy a tablet because of the number of cores it has, that’s a mistake'

September 10 2013

Companies hype quad-core now and we'll probably see eight-core tablets, but that's not a good measure of performance, says chip analyst Mike Feibus
Companies hype quad-core now and we'll probably see eight-core tablets, but that's not a good measure of performance, says chip analyst Mike Feibus

Mike Feibus, principal analyst at TechKnowledge Strategies, was one of the thousands of attendees at Intel’s big IDF conference this week. In an interview with TabTimes, the chip expert talked about what’s wrong with tablet specs, competitors to the iPad and where the tablet industry is headed. 

TabTimes: How much should tablet buyers compare chip performance specs when trying to decide which device to buy?  

Mike Feibus: The specs companies are using to describe the processors inside tablets don’t really help much. You can have two tablets, and one has a 1.4 GHz chip by Intel and the other a 1.6 Ghz chip from Qualcomm and the Intel tablet will be faster - or vice versa. There are a lot of other factors that determine performance.

Now we’re seeing more tablets with quad-core processors. It might be logical to think they are going to perform better than dual-core models, but something tells me you are going to disagree ... 
If you buy a tablet because of the number of cores it has, that’s a mistake. It’s not a good measure of performance. We really need better benchmarks for tablets. 

So why more cores at all?
In some ways it’s just the processor companies yielding to the conventional wisdom that a higher number is better. I’m sure we’ll see tablets with 8-core processors at some point. 

The companies will hype more cores, but privately they’ll tell you it’s hard to increase performance that way because it comes down to how many threads the OS or the applications can address and in a lot of cases the software hasn’t been written to take advantage of those extra cores. A lot of things on the Web are single-threaded and a powerful single-core processor can handle those just fine. 

Intel seems to be in a perpetual state of trying to catch the performance and power efficiency of ARM chips that dominate smartphones and tablets. Does Intel’s latest Atom processor, Bay Trail, finally give them a leg up?  
Intel has systematically been closing the gap. There is nothing inherently mobile about ARM just as there is nothing inherently performance-oriented about x86, it’s just the direction those two camps chose to focus on. Intel’s done a complete 180 the past few years to offer the kind of low power efficiency features that have been ARM’s trademark. 

But ARM keeps its lead?
I think 2014 could be the year Intel eclipses ARM. By this time next year I think you’re going to see a big jump in the number of design wins for Intel in smartphones and tablets. 

How important was Apple’s decision years ago to start designing its own chips for mobile devices? 
It was huge. Apple knew what it needed to do to create something new, the iPhone and then the iPad, and it went out and invested in the technology it needed. Now the other vendors understand the workloads and how to compete so I think Apple’s edge there, and even Samsung’s for that matter, is a lot less than it was. Qualcomm, Intel and Nvidia are all doing a great job to address the mobile device market.

Share with: Comments (1)    v

Free newsletters for more tablet news, insights, apps and tips



Comments

 
  • John Swain
    10 months 3 weeks ago

    Intel was fixated on clock speeds for decades as a means of improving performance. It hit an unseen brick wall when Moore's Law failed them above 3 ghz. Heat and power consumption made it nearly impossible to create an inventory of "never-ending" faster chips. Some gamers compensated by building water towers and resorted to over-clocking their machines in order to gain a slight advantage within custom built systems.

    Binary code went from 4 bit words in 1980 to 8 bit words for the purpose of sending code in larger chunks through the operating system with each beat of the clock. Soon, 16 bit, 32 bit, and now 64 bit operating systems are the norm, but they are still in the 2 to 3 ghz range.

    Moore's Law was still in good stead, but the means started to change. The future indicates that 128 bit and 256 bit words will undoubtedly come to fruition but the cost and size of the chip is in it's infancy and, as of yet, it is not ready to be marketable.

    So, what's the next trick? A drum roll, please...... Answer......more cores on the same chip. Multi-cores and multi-threading of 64 bit words at the same clock speed keeps Moore Law marching on.

    So what's the hang-up?.....Software, it's not written to take advantage of all those cores and threads.....yet. But why would it be? Why write software for nonexistent hardware? The quintessential question of which comes first applies here. Multi-core processors will eventually be supported by software.

    Epilogue:

    Each step forward requires the planning and movement of the prior step. The capacity to foresee the path is the key to success, even if no clear path exist.

    Intel, for all of it's successes, missed the direction and magnitude of mobile computing. ARM chips have gained a market advantage.

    Microsoft is stuck in a cumbersome operating system.

    Apple is ignoring phablets, as if it was an aberration emanating out of the far east, or could it be.......Samsung might sue Apple for patent infringement, what a twist of fate.

Latest in tablet business / productivity