In 2012, SemiAccurate reported that Intel's (INTC +13.96%) Broadwell processor family would not come in a "socketed" format for desktops and would be a "mobile only" processor. Intel acknowledged it would launch an unlocked and socketed variant of Broadwell for desktops (something confirmed in this VR-Zone leak) at some point, but only a single, premium variant.
Over the months, various leaks concerning Intel's next-generation Skylake architecture have suggested that socketed versions of that architecture will be very widespread -- in stark contrast to what Intel seems to be doing with Broadwell. While what Intel is doing with desktop PCs is interesting, what's more interesting is what the company seems to be planning with its Xeon E3 family of processors, which are derived from the company's socketed mainstream desktop offerings.
Taking a look at the public road map
Intel recently updated the public road map for its data center-focused products, reproduced below:
Source: Intel.
While this does not give us the granularity that I'd like, we know that Xeon D is scheduled to be rolled out in the second or third quarters of 2015, per CPU World. The road map above suggests, then, that the Xeon E3-1200 v4 family will likely be available in the second quarter of 2015 (which CPU World also corroborates).
While the consumer desktop PC version of Broadwell is expected to come only in a premium variant with four cores and the highest-end integrated graphics, it's not clear whether that will also be the case for the Xeon E3 family of processors.
Looking at the evidence
According to CPU World, which published details of the Xeon E3 v4 (Broadwell) and Xeon E3 v5 (Skylake) processors in October, the Xeon E3 v4 is expected to feature "up to four Hyper-Threaded CPU cores, and up to GT3 graphics with 128 MB of embedded DRAM." The operative phrase here for both CPU and graphics is up to, which implies lower CPU and graphics configurations.
Intel appears to be rolling out for its Xeon E3 line an entire lineup of socketed Broadwell chips that will be compatible with the same boards as today's Haswell processors, unlike what it is doing for consumer desktop chips.
What is the business case for doing this?
While I can think of a number of potential explanations for this launch strategy, the one that makes the most sense revolves around the different margin profiles between consumer desktop chips and Xeon E3 processors.
It is well known that Intel's yields on its 14-nanometer manufacturing process still aren't quite where the company would like. This means the average cost to manufacture each chip is higher than Intel would like and higher for Broadwell processors relative to their Haswell predecessors.
Source: Intel.
It makes sense, then, that for mobile applications that absolutely need the performance and power savings of 14-nanometer technology, Intel would take the margin hit until yields improve. Furthermore, for the entry-level workstation customers that the Xeon E3 targetings, total cost of ownership (which includes power costs) is generally more important than the initial chip acquisition costs. This means Intel might simply charge more for these products to offset the increased cost structure.
The cheapest Xeon E3 v3 per Intel's ARK database is priced at $193, with the most expensive variant listed at $774. On the consumer side of things, the cheapest Core i3 goes for $117, with the most expensive mainstream socketed Core i7 product listed at $339. While these aren't ultra-cheap by any means, I would imagine the average selling price of a Xeon E3 is much higher than the desktop-oriented Core i3/i5/i7 family of products.
What I believe Intel is doing is allocating its initial 14-nanometer wafers (which will have lower yields) for the products that will benefit most from the technology, and waiting for the yields to increase before transitioning all of its major product lines to 14 nanometers. Intel seems to be using the same strategy with its low-cost Braswell chips for low-cost desktops and notebooks.
