Quad core CPU? Advanced Micro Devices (AMD) recently released their new line of Jaguar based APUs. In a recent article, I made the case for AMD's future in scalability. By focusing on repeatable blocks of IP, it can very specifically focus R&D efforts. This is important for two reasons. The first is because it allows customers flexibility and choice. The second is because AMD is in the process of reducing expenses. By focusing very specifically in only a few areas, it can more efficiently control spending.
When Could We See These Custom Designed Chips?
A few days ago, I would have probably said when the Xbox 1 or PS4 launch. Now I'm changing my guess to the end of this month (source: PCWorld.com). According to PC World, Samsung's (OTC:SSNLF) Ativ 9 Lite will start reaching customers on "July 28th." Based on conflicting reports between many tech sites, my initial thought was that Samsung would simply launch different flavors of SKUs. Now, according to articles such as the PC World article I am quoting, it is now described as a custom designed chip.
Why Is This a Big Deal?
In my mind, this is on the scale of the console announcements. AMD has alluded to additional design wins before. But I short-sightedly wrote this off as the unannounced Xbox One win. If this is indeed a custom chip, this has big implications. The biggest is that this means this deal had been in the works for a while. According to a quote from AMD's Mr. Feldman, it takes about "$400M and three years to build an x86 server chip. But only $30M and 18 months are required to make an ARM chip" (source:SemiAccurate). Given that an ARM designed chip takes 18 months, and an x86 chip takes longer, this goes to show how long ago this deal was signed.
What I am curious about is who is building it? At this point, I'll pull a page from SA contributor Russ Fischer's playbook and think outside the box and say Samsung (my 100% speculation). The Wafer Supply Agreement was amended last year to break exclusivity from Global Foundries in March -- roughly 18 months ago, which happens to be the same time Mr. Feldman said takes to design a chip with ARM Holdings (ARMH). Review my last AMD article for an explanation of the WSA.
The significance of this cannot be understated. This means AMD could have more impending custom design wins. For instance, (again my pure speculation), a Samsung tablet powered by AMD. As an AMD long, now I am looking for ambiguously described processors of AMD origin in higher profile product announcements.
The following quote is taken from the PC World article linked to prior:
It's something we wanted that was outside [AMD's] roadmap, Samsung PC Product manager David Ng told me in a briefing on Friday. Asked why Samsung chose to partner with AMD for this custom CPU, Ng replied "A huge part of it was the superior integrated GPU in AMD's parts.
The most important implication in my mind would be if we learn that Samsung is making these chips. First, it demonstrates portability of AMD's architecture between different foundries. Second, but more importantly, it would show the WSA amendment made last year allows Samsung to manufacture 28nm APUs. This could allow Samsung to launch an AMD tablet. This is all based on my speculation of Samsung making these chips in-house. This would explain why Samsung using a custom chip remained a surprise to at least some analysts (source: technewsworld.com).
The second implication is that it demonstrates the point I made earlier about AMD's scalability. For the past several years, AMD has been relegated to cheap designs. The Samsung Ativ 9 Lite is the kind of design win I like to see. The launch of the Ativ 9 Lite at the end of the month will be the first time we see a product release based on AMD's commitment "to an ambidextrous approach and providing choice to our customers across both ARM and x86 ecosystems." And although no official word has been released, the recent leaks of Beema appear to point to this trend continuing.
Additional disclosure: I actively trade my AMD position. I may add to it or take profits at any time.