Earlier in the month I tentatively made the call that AMD's (NYSE:AMD) Jaguar core-based Kabini/Temash APU's were good enough to give AMD hope. Now that we've seen Temash for real in reference devices as were shown off yesterday at Mobile World Congress I have to say that AMD has a lot more than hope. AMD showed off three very impressive tablet designs from each of the major Taiwanese ODMs, Compal (OTC:CMPCY), Quanta (OTC:QUCCF) and Wistron. The A6-1450 will give tablet/hybrid manufacturers a lot of options they do not currently have.
The key here is AMD's TurboDock technology that will allow the chip to scale up and down in Thermal Design Power (TDP) depending on what the user wants. So, take the tablet on the road and it runs at 5.9W TDP -- which will dial back even further to 1W for normal usage like browsing, etc. - dock it to the base station and it can scale up to a 15W TDP with fan-based cooling . Check out the video of a 13 tablet running the Fishbowl app from Microsoft (NASDAQ:MSFT) dialing up and down Temash in-situ versus a Z-2760 Clover Trail from Intel (NASDAQ:INTC).
The latest benchmarks that we've seen have the A6-1450 pushing Cinebench R11.5 tests slightly better than a 17W Core i3 Sandy Bridge, a test that Intel's chips do very well in. If this is 5.9W performance it's a ridiculous jump. But, even if this is 15W performance let's think this through. Temash has the south bridge on die so a 15W TDP is really 15W, versus the 19-20W the i3-3217 plus chip set is pushing. It's producing nearly 4% better results from 21% less power and 70% of the die size. All of these things together point to real design advantages. Add in OEM fatigue over Intel's lack of competence at the low end of the mobile market and I'm seeing a number of defections in the near future.
As I suspected when I saw the early benchmarks, Jaguar simply outclasses everything in the sub-1.6GHz arena when everything is factored in: die size (and therefore yield), performance/watt and cost. These are chips that are supposed to be competing with the next generation Atom SoC, Bay Trail, and they are more than capable of competing with the low end of Intel's big core processors. Even if Bay Trail is a huge leap over Clover Trail, where would that put it, performance-wise? Near 40nm Bobcat?
It is no wonder that Jaguar is powering not only the next Sony (NYSE:SNE) Playstation but also the next Xbox. Both of those high volume design wins should allow AMD to satisfy its take-or-pay wafer supply agreement with Global Foundries and grab a little extra margin out of every Temash it sells into the channel. As we have seen with the Surface Pro, the demand for a good, full Windows 8 tablet is out there; even it is mated to the wrong CPU/GPU solution for the form factor a 17W Core i5.
Last year when the rumors were flying that Apple (NASDAQ:AAPL) nearly ditched Intel for AMD in its MacBooks it was obvious that Apple could do no more than kick the tires because AMD was in no position to supply them. Trinity was too big, not fast enough, and the yields too low to ensure the numbers that would satisfy Apple's needs. But, today is completely different. Kabini/Temash is being built on TMSC's 28nm process which is now solid and the 110mm2 die size for the A6-1450 ensures high volume is achievable.
Am I suggesting that Apple ditch Intel for Temash in its MacBook Air? Not really, but a Turbo-Dock enabled Kabini that can scale up and down from 15 to 25W TDP is not beyond the realm of possibility and would, frankly, bury Ultrabooks. Not that we've heard neither hide nor hair of such a product but until a week ago we hadn't heard about Turbo Dock at all. Honestly, it would be the only product in Apple's product refresh this year that wouldn't suffer from margin compression.
The more we hear about Temash/Kabini and the performance of these Jaguar cores the more we have to consider that AMD has literally blown up the expectations of the mobile computing market for x86 CPU's in the same way the Qualcomm's (NASDAQ:QCOM) Snapdragon S-4's did last year and into the first part of this year. Intel doesn't have anything that competes on price, performance or size and ARM core SoC's - despite the best efforts of Nvidia's (NASDAQ:NVDA) marketing arm - have not bridged the performance gap enough.
I will remain skeptical of Nvidia's Tegra 4 until I see it working in a tablet from a credible OEM and at this point Hewlett-Packard (NYSE:HPQ) is the antithesis of that. They may have finally solved the memory bandwidth problems of the past Tegras but Nvidia may also have burned a few too many bridges by over promising and under-delivering.
In technology you have to go for where there is no product and fill that void aggressively. A product like Jaguar mated with even the rudimentary ability to be utilized by the customer in multiple ways creates a kind of best of both worlds for everyone.