Italian website, Bits 'n Chips, reports that, according to its sources, Intel (INTC 3.25%) has removed the server-oriented Broadwell-EP/EX processors from its product road maps. The report claims that Intel hasn't sent out any official notice that these products have been cancelled, but the disappearance of these chips from official product road maps suggests that Intel may be skipping these parts and will instead go straight to Skylake-EP/EX.
Is this rumor likely to be true? Let's take a closer look.
Timing is everything
I would say that the likelihood of this rumor being true depends heavily on Intel's expected launch timing for its Skylake-based server chips. These chips should be quite compelling, and the substantial new platform features should make it a solid upgrade from the Grantley (Haswell/Broadwell) platform.
However, according to a fairly recent road map leak, the Skylake parts and the associated Purley platform will not launch until sometime in 2017.
If that remains the case, then it would be an incredibly bad idea for Intel to outright cancel Broadwell-EP/EX. Even though the Broadwell architecture is just a minor enhancement of the Haswell design, Broadwell-EP/EX are expected to pack in more cores than the corresponding Haswell-EP/EX parts, improving performance.
Additionally, since Broadwell-EP/EX will be built on Intel's 14-nanometer manufacturing technology, which is a solid improvement from the 22-nanometer technology used to craft the Haswell-EP/EX chips, performance-per-watt should move up substantially as well.
So, if Intel has a chance to offer higher-performing, more efficient server processors in the 2016 time frame (a fairly recent slide deck on Cisco's suggests that the Broadwell-EP parts will roll out in the first quarter of 2016), why would it instead continue to offer older, lower-performance, and less efficient parts for an entire year?
The rumor makes no sense
I'm not convinced that this report is correct. It seems that Intel has already done the hard work of building the Broadwell-EP/EX server chips, and given that the Skylake-EP/EX parts are still a ways off, the company would do well to release the Broadwell chips in order to keep its server offerings fresh.
Indeed, Intel executives generally make a big deal about how its customers continually buy products higher up the product stack, which has been a major driver of the company's data center group performance. If Intel's 2016 product line is essentially the same as its 2015 product line, then it could be hard to get customers to buy an even richer product mix next year.
Finally, also keep in mind that Intel signaled at its investor meeting last year that by the third quarter of 2015, Intel will be able to manufacture a 1.3 billion transistor 14-nanometer Broadwell processor at a lower cost than it was able to build a 0.96 billion transistor Haswell processor in the second quarter of 2014.
I suspect that because of this significant cost-per-transistor advantage that Intel should have with its 14-nanometer parts, Intel will want to transition its product line to that manufacturing technology as soon as possible.
Additionally, Intel has signaled that it has been taking 22-nanometer capacity offline and reusing that equipment and space for 14-nanometer production, potentially further incentivizing a move of its key product lines to the latest manufacturing technology.
