Rumors broke today, sparked by a Korean business site, that Samsung may be considering making a bid for AMD. It’s not unusual for these kinds of rumors to make the rounds, though it’s been awhile since I heard Samsung’s name floated. Any such acquisition would mark a sea change for the Sunnyvale-based company, and for the enthusiasts that have been its long-term customers.
Why would Samsung buy AMD?
There are several reasons why Samsung might buy AMD. Advanced Micro Devices is one of the few companies with experience in designing both the CPU and GPU halves of a custom SoC. Imagination Technologies and Nvidia both have much more experience building smartphone and tablet hardware, but these companies are on better financial footing and would likely cost significantly more — assuming either was interested in a sale to start with.
AMD’s stock has been battered by short selling, tough competition, and the company’s long, slow road back to a viable CPU architecture. Its stock currently trades at $2.66, down from $9.81 five years ago. After the disastrous Bulldozer debut of 2011, AMD has been forced to execute a slow roadmap of iterative improvements and cost-optimized product transitions. A massive injection of Samsung’s capital could give the chip designer some breathing room and let it put its expertise to work again.
AMD already has a mobile GPU that could be ported to tablet form factors, even if smartphones are currently out of reach, and the Radeon brand still has broad recognition. The company’s plans for an ARM-based server business may be in their infancy, but AMD has built at least one bog-standard ARM core, and is working on its own custom architecture to debut in 2016. Having wrested foundry leadership away from TSMC, at least for the moment, Samsung might believe that acquiring a chip firm is the next logical step in its long-term competition with companies like Qualcomm and Intel.
What would a Samsung-AMD acquisition mean for enthusiasts?
Enormous change. When I wrote up AMD’s roadmap plans last fall, I discussed the fact that AMD is transitioning away from being a company that competes directly with Intel in the core PC market. Chimpzilla wants to engage in different arenas, including embedded semi-custom architectures, high-end GPUs, and ARM-based dense servers. That doesn’t mean AMD has given up the x86 fight, but it does mean the company’s priorities aren’t focused on the bleeding edge of x86 CPU design.
The hard truth is this: The single-thread performance gap between AMD and Intel is larger now than it was in the K6 days. It’s larger than it was when Hyper-Threaded Northwood blew past 3GHz and left the single-threaded 130nm Barton struggling in its wake. AMD has made huge improvements in performance-per-watt — I wouldn’t have bet a nickel on a Bulldozer-based SoC ever hitting the 17W power envelope four years ago — but the company’s CPU division has been fighting an uphill battle for the past four years on shoestring budgets and grudging implementation timetables.
AMD’s HSA and high-performance on-die integrated graphics have pushed the technological envelope over the past few years, but APU prices and non-game benchmarks both remain largely chained to conventional CPU performance. Developers and programmers agree that features like heterogeneous compute and many-core architectures are the long-term future of computing, but AMD needs a new, higher-performing CPU architecture to better match the innovative work it has been doing on the GPU side.
If AMD agrees to be acquired by Samsung and jettisons its x86 business, it would be a tacit admission that the company didn’t think its next-generation design would deliver what it needed to better match Intel or emerging ARM vendors.
With all of that said, keep in mind that this is just a rumor. Until we hear otherwise from a more reliable source, everything is business as usual.
No comments:
Post a Comment