I hate the word hope. Not because it's ugly or unpleasant to say, but because it is simply that which one has when one has nothing else. Four months ago I had no hope for AMD given how badly it had botched not only its product roll-outs in 2012 - I had high hopes for the Trinity APUs -- but also how management has allowed the extremely important Kaveri APU to slip to 2014.
As SemiAccurate reported previously, if Kaveri is going to be that late because Intel's (INTC) Haswell is going to be that good, then why even release the chip at all and move forward with the next planned release, which hasn't been announced yet? But, one has to assume it will be a hybrid x86-ARM (ARMH) SoC designed at capturing a specific market segment, and for AMD that market segment is gaming.
Console sales are nearly as uninspiring at this point as PC sales are. These will be nice wins for AMD but they won't save the company. One could make the argument that the margins on those chips are so small that Intel couldn't really be bothered to challenge seriously for them, whereas AMD needs all the revenue it can get. They will keep the lights on and people fed but not much more.
So, what has changed? What has me thinking AMD might now actually survive long enough to see its longer-term plans make it to fruition? In a word, Kabini.
At CES last month, AMD pretty much shocked everyone when it informed them that the new versions of its low-power, entry-level APUs were not only ready but shipping to OEMs. Now I say shocked people because no one was expecting anything from AMD, but it was pretty obvious that AMD was that far along in this part of its roadmap.
Both Kabini - and its tablet-focused ultra-low-power envelope cousin Temash - represent, I believe, the home run AMD needed to hit with Trinity and missed. Trinity suffered from a number of issues, most obviously, that it wasn't enough better than Ivy Bridge to entice OEMs to cross Intel.
The same cannot and will not be said between Kabini and Clover Trail, Intel's latest Atom abortion. Kabini, by the look of the design, will be capable of serving everything from tablets to ultrathins /low-end laptops and all-in-one desktops. But the kicker that no one at AMD will confirm yet but is the obvious opportunity-- cloud micro-servers.
Kabini is a full SoC with both integrated south bridge and GPU and the first AMD APU to use its Graphics Core Next [GCN] technology. The result is a massive increase in graphics performance over the current Brazos 2.0 E-1800, as well being a true quad-core CPU all built on a 28nm process that, according to TMSC, is yielding well from the start. Clover Trail cannot touch this. Temash is simply Kabini with parts stripped, and the TDP dropped to the 4-5 Watt range. Kabini SKUs have been uncovered covering 8 to 25 Watt TDP so far.
There is a religious war being fought at Intel over the death of the PC, and slowly the company is coming around to understanding that tomorrow's computing world is not about raw computing power but raw computing power per watt. That war will leave Chipzilla in a rough way as it battles through at the management level to come to grips with the current changes. If Intel wanted to rule the low power market, it could. Killing off its motherboard division was a good first step. There are usually eleven more to come after that.
That's why ARM SoC's are winning and that's where Kabini and the Jaguar core have a huge advantage over Intel. Haswell will be good for general purpose devices, and I'm sure it will scale down to a point and put even more pressure on AMD's ability to sell Trinity/Richland APUs in mid-end devices.
But, Kabini has the opportunity to literally own the low cost/low power x86 market for nearly an entire year before Bay Trail is ready. Given how fundamentally terrible Clover Trail is - the real reason why the Surface Pro is both so very late and so very over-priced - there is no guarantee that Bay Trail will be any better than Clover Trail. Intel is still in show-me mode with its low-power per watt solutions. Kabini could swoop in and literally save this entire market segment from irrelevance while iterations of it can be put everywhere.
In this context then, AMD's only real competition is that of the Cortex A-15 derived chips powering Google (GOOG) Chromebooks and the like. If Microsoft cannot get developers to embrace Windows RT, however, there will be little room for the usual suspects like Samsung and Qualcomm (QCOM) to strip mine more users from the x86 word into the ARM one.
When I saw the reports that the story around Kaveri was being written by Samuel Beckett, I gave AMD up for dead. But that decision turned out to be the right one because, right now, AMD has to use judo to survive against Intel and ARM and not confront them head-on. It doesn't have the resources for that. Delaying Kaveri, re-designing it to compete with Broadwell while truly focusing on creating an SoC that can credibly power the growing hybrid laptop/tablets that the current crop cannot, was the right decision. And it just might make AMD enough money to get Kaveri to the table on time and, more importantly, in time for AMD's shareholders to have something more than hope.