A quad-core iPad 3? Not so fast

Will Apple's next iPad reach performance nirvana with a monster quad-core chip? Nirvana, maybe, but not necessarily via quad-core.

But let's put aside whether Apple's latest chip will be quad-core for a minute and look at what others are doing. (See a post at The Verge on this topic too.)

One of the leading lights among ARM processor suppliers is Texas Instruments. They've made it clear that they're not going quad-core this year and not even next year (necessarily) with their first next-gen OMAP5 chips.

In a CES demo, TI showed off the next-gen OMAP5 silicon running Google's Android 4.0 Ice Cream Sandwich. OMAP5, due by 2013, will have two next-generation ARM Cortex A15 processor cores (count them, 2).

Here's the interesting part: next to the main processor cores are two smaller M4 cores. ARM describes them as microcontrollers for digital signal control. Very different creatures from the big, brawny A15 cores.

TI went on to explain to me that OMAP5 is their version of ARM's Big Little concept. Two muscular general-purpose cores next to two smaller specialized cores.

OMAP5 will also have a dual-core Imagination PowerVR SGX544 graphics processing unit. (But I digress. Let's stay focused on the CPU.)

And Intel? Its Medfield chip for smartphones and tablets is single-core but benchmarks nicely against dual-core rivals.

Then there's Qualcomm. That smartphone chip heavyweight has been less specific but it's not likely they will ship quad-core chips commercially this year into smartphones and tablets.

Which brings us to Apple's next-gen chip, which we'll call the A6. I queried two analysts about the prospects of a quad-core A6.

Gus Richard, Piper Jaffray: "It depends what they do with the graphics core. If they go hog wild with the graphics core, the chip could be quite large" leaving little available chip real estate for four homogeneous general-purpose cores, said Richard.

"You may or may not go from dual to quad. To me, it's really a moot point, though. It's about how you change the performance with the transistors you got," he said. "They may have enhanced the cores and that may be more than enough to offset for not having four cores."

Anand Shimpi, Anandtech: "If people are thinking quad-core in the [Nvidia] Tegra 3 sense, it seems very unlikely."

"Having four [Cortex] A9s (analogous to Nvidia's Tegra 3), that's just not the right move."

Shimpi continued. "My main reason for quad-core not being the right next move is that I think single-threaded performance still has a lot of room to improve. And improving that doesn't mean going with more of the same cores."

"Another option would be to have a dual-core A15 based part. That would be the most interesting from an architecture standpoint. But it's a riskier route. But if I were in Apple's shoes that's the way to guarantee that you don't have to worry about updating iPad silicon for another 12 months."

He continued. "Another option is that you have two A15s and two A7s. That's one way for them to be a quad-core but not really be a quad-core [because] you wouldn't run four cores at the same time."

And about the iPad 3's GPU, Shimpi said it's more of an issue of memory bandwidth than raw GPU compute power. "How do you deliver that bandwidth. Is it through additional memory channels? Is it through higher-clocked DRAM?"

And how to best summarize all of this? Whatever Apple decides to do with the iPad 3, it will most likely be fast yet relatively power efficient. And that's all that matters to consumers.