This teardown is of the EMC DD160 which is a 2U rack server, with a single processor socket. At a high level - the system is composed of a Chassis (which includes two power supplies but without motherboard), a motherboard, and other off-the-shelf components. Final assembly markings on the unit say Made in China.
Unlike the Barracuda and Zenith systems we saw that use off the shelf system level components, EMC is known for designing their own motherboards - which is a key differentiator. This gives them more control over the BOM and costs.
The EMC design leverages Intel at it's core which tends to make for a pricier solution overall. With the Zenith and Barracuda systems - because they use 'off the shelf' system level components (Mobos, etc.) it makes it easier to confirm the cost chain from discrete components like integrated circuits, to the price of finished sub-assemblies (ASUS mobo for example). With this system it's more blind - but the ratio between the sub-system cost and the final retail prices leaves a lot of mystery for us. Some of this - in fact much of this must come down to the software/IP portion of the total cost for EMC, NetApp, etc..
In fact, based on what we see in our analysis - even if some of these products are built at 'full retail' subsystem pricing (the kind of pricing that anyone can achieve buying single units) - there is still a huge delta between the total cost of material and the retail pricing which suggests there is a heavy R&D, software and 'other' non-hardware costs to be considered to make these companies profitable.
Though it's not normally part of the teardown analysis, we have provided a client-specific analysis here that is designed to answer not just the question of 'what are the component costs that going into the system', but what are the subsystem totals and markups that tend to stack up along the way that ultimately add up to a price to the end buyer.
Based on data sheet copyright date.
For this configuration - EMC sell this device (MSRP) for approxmately $15K - this is the price paid for the unit analyzed here.
5,000 Annual Production Volume
5 Total Years
For the purposes of this teardown analysis, we have assumed an Annual Production Volume of 5000 units and a Product Lifetime Volume of 5 year(s).
Teardown volume and production assumptions are primarily used for our cost analysis in terms of amortized NRE and tooling costs, especially for custom components specific to the model being analyzed (mechanical components especially). Unless assumed volumes are different by an order of magnitude, minor changes in volume (say 1 million vs. 2) rarely have a large net effect on our final analysis because of this.
Our normal methodology for compiling top cost drivers does not apply here, and we have produced roll ups that more closely emulate the modules and options that end-buyers would be able to buy or negotiate. As an example of what this means, we assume that end-buyers will be more interested in negotiating NIC cards, or RAID controller cards, for example, instead of individual ICs.
We do not normally account for logistics costs or 'other' costs beyond basic electronic systems and EMS-Level assembly, teardowns are hardware cost focused. But in order to better chart the cost links in the supply chain, and how that might affect end-buyers we have created a simplified supply chain (in the ''System Cost Analysis'' Tab) with rough mark up assumptions (which also should account for logistics cost) between the parties to better account for where our BOM costs are in the grand scheme of 'total cost', or price to the end customer.
Country of Origin
For the purposes of this analysis, we are assuming the following country(ies) of origin for each level of assembly, based on a combination of 'Made In' markings, and/or assumptions based on our knowledge of such equipment.
Backup Battery Unit - China
Box Contents - China
CPU - China
Drives - China
Memory - China
Motherboard - China
Other - Chassis - China
PCIe Network Accelerator Card - China
Country of origin assumptions relate directly to the associated cost of manufacturing, where calculated by iSuppli. In the cases of 'finished' sub-assemblies (such as hard drives or DIMM modules), we do not calculate internal manufacturing costs, but rather assess the market price of the finished product in which case country of origin assumptions may or may not have a direct effect on pricing.
Labor rates are applied directly only to hand inserted components and systems in our bill of materials, and although regional assumptions do, these new rates do not have a direct effect on our modeled calculations of placement costs for automated SMD assembly lines. "Auto inserted components (such as SMT components) placement costs are calculated by an iSuppli algorithm which allocates a cost per component based on the size and pincount of the device. This calculation is affected by country or region of origin as well.
Component counts by assembly and the number of assembly are indicators of design complexity and efficiency.
Component Qty: 12 - Drives
Component Qty: 830 - Other - Chassis
Component Qty: 2 - CPU
Component Qty: 2920 - Motherboard
Component Qty: 4 - Memory
Component Qty: 1117 - Misc Peripheral Assemblies
Component Qty: 11 - Box Contents
Component Qty: 4892 - Grand Total
The component counts of this system are inline, largely speaking, with other systems of similar functionality. Storage servers are inherently complex and we expect the kinds of elevated components we ee here. Most of these components in such a system are board level components. At a systems integration level this equates to handful of items - which are, for the most part listed above as lineitems (Motherboard, CPU, Etc.). Final integration is not terribly challenging.