For storage service advantage of ARM is well known; good enough for that application. Real time from flash memory array is an into future time story including acceleration.
Success for ARM in web tier has been, and will continue to be, software systems application optimization of the raw x86 ports for customer applications. As the software service community begins to understand this opportunity, which a leading few understand today, including waiting for applications and refined tool's, whole platform development will recognize the intellectual property advantage of commercial reuse. That, of course, is a first mover advantage requiring action over all the potential compliment parties looking at each other waiting to see who starts first.
On open processor modules, has been one keystone of Intel's ARM crush. The other a wave of surplus three generations removed. Open Hug, meaning too asphyxiate, and for industrial embedded, Q7 and MXM3, are lures Goliath relies to place head to head. On system application prototype advantage of the Intel plug compatible replacement two lithography generations ahead, ARM has been stymied at 32 nm, where anyone who does that ARM specific test except for fan less contained chassis chooses Intel or AMD x86. Prototype software development begins and Goliath knows that is that.
At 28 nm the question is open, on performance, but more so price performance on planar layout that is wide and slim v the compact nature of a costly tightly coupled stack. Where footprint does not matter ARM may win and AMD certainly takes advantage today for cost : price. For performance that is low power for price, and design power mm^2, for dense server the jury is still out.
Within the Intel server sales paradigm any power reduction, within the established infrastructure, is evolutionary. When no one is fired for buying the incumbent over upstart. A credo only the best Intel competitive strategists have learned, by doing, to overcome. So do or do not.
The volume opportunity for ARM Intel displacement, over one architectural generation that is nine quarters, is currently one Intel production long run or approximately 60 million units.
By bringing ARM architecture to the forefront in desiging CPUs for servers, I think AMD sees a development of heterogenous architecture at the data center level, similar to bigLITTLE processing by ARM. For this, the software ecosystem has to be tweaked and enterprise architecture is an environment that can see wide fluctuations in compute demands.
On audit ARM licensees, and compliment see this acceleration opportunity that can also be described as coprocessing. There are multiple types of ARM +. As there are multiple types of x86 +
A raw green field of programming guides, and the required application notes that are GPU/APU, DSP, FPGA specific and the industrial embedded vertical software system applications, where some laugh moving for commercialization too server, relational, batch processing. Absolutely tweaked for every vertical and how too move to horizontal? Intel is addressing as x86 core + Phi on chip. The product and tools more universal yet can that ease of solution be as application specific?
AMD is looking to prepare itself for a variety of compute demands and a great deal of fluctuation by creating a heterogeneous architecture. ARM is going to experience a wave of products based on its solution that will disrupt the traditional x86 stranglehold on the market. But with both X86 and ARM architectures, AMD is equipped to provide the kind of variety they will need to remain competitive in 2014 and beyond.
Intel PE is 13. It is low comparing with other Semiconductor Company.
You might be right about me to be pessimistic. I am not sure server market will gown, in term of revenue, significantly in next 6/7 years for 3 reasons:
1. I am seeing those traditional servers buyers, such as Google, Facebook, are migrating to their in-house solution slowly. Those are not always x86 solution.
2. With more companys moving to clould and virtualization, less servers are required due to high server utilization;
3. x86 servers are mostly built for general purpose server. The power efficiency and cost efficiency are horrible. I'm seeing more purposefully built servers which could be accelerated by ASIC, FPGA, or even GPU. For power perspective, ASIC is 1000x more power efficient than x86, even FPGA is 100x better in lots of cases. Most of those purposefully built servers are likely ARM based SOC product, with lots of purposeful built acceleration engine direct target to its applications.
A Book For All Reasons Bernard Cole1 Comment Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...