That is true. Sadly anybody with insight into how much data center customers pay is under NDA.
That being said, we can do some back of the envelope math to estimate Intel's ASPs for server platforms.
In 2015, Intel reported DCG revenue of $16 billion with ~$14.8 billion coming from CPU/chipset sales.
If we estimate the total x86 server market at around 20 million units, this would imply CPU + chipset average selling prices of around $740 each.
Xeon D is certainly at the far low end of that given how many Xeon E5/E7 chips Intel ships to cloud/enterprise customers, so I wouldn't be surprised if Xeon D CPU + mobo in high quantities were more around $200-$250.
Qualcomm Shows How Intel Could Lose The Data Center
* Qualcomm's server joint venture in China lends considerable credibility to its ARM server initiative.
* The joint venture may serve as a model for ARM adoption by other large scale cloud service providers such as Amazon and Google.
* The cloud providers are best-positioned to overcome the key stumbling block to ARM server adoption: software.
Qualcomm (NASDAQ:QCOM) announced yesterday a joint venture with the Chinese provincial government of Guizhou to build a data center based on ARM server technology. Guizhou Province will build a cluster of green data centers using more than 2.5 million servers with chips based on Qualcomm's ARM server processors. The joint venture will also focus on research and development of server chipsets for sale throughout China.
When Qualcomm first announced its Server Development Platform back in October 2015, I was very skeptical that it would make any inroads in the Intel (NASDAQ:INTC) dominated server business. Servers have become commoditized, and Intel dominates this commodity market.
Following Intel's 2015 Q4 earnings report, I began to wonder if Intel is really as secure in the data center as it seems. Intel appears to have given up on smartphones and lightweight tablets, unable to compete with ARM processors on cost and performance.
Intel's failure in mobile is not merely a failure of a technology (x86 processors) but a failure of a business model.
[...] The custom SOC paradigm offers technical and business advantages. The key technical advantage is that mobile device makers can optimize the SOC design for the mobile device and potentially for the operating system (and vice versa), in the case of Apple. The business advantage is the classic one of vertical integration: profit that might have gone to the commodity processor vendor is retained within the mobile device company.
In addition to the structural advantages of the new paradigm, there appears to be a fundamental cost advantage for ARM processors that Intel could never overcome, as long as the mobile device makers could pursue the new paradigm at sufficient scale.
The custom SOC paradigm offers a template for adoption of ARM servers in the data center.
Don't forget, Cavium is still kicking around out there with ThunderX. I don't know anything about their hardware schedule though. Allegedly there are test systems in the wild already, since people have been working on getting the mainline kernel to boot on ThunderX without any platform-specific kernel patching. Kernel 4.4 will supposedly allow that.
Given that the CPU has a slight say in sata speed test it s not unreasonable to assume that Kabini ,AMD s low end APU released in 2013, has a better sata controler than the best haswell dedicated chipset..