1 Way Intel Corporation Is Leaving Money on the Table

In the world of desktop PCs, Intel (INTC 0.14%) generally offers two major types of desktop CPUs. The first is the "mainstream" lineup, which are fundamentally built from the same designs as higher-powered notebook processors. These generally offer either two or four cores and varying levels of integrated graphics.

The other lineup, which Intel refers to as the "high end desktop" or HEDT platform, is based on Intel's one-socket Xeon processor family for servers and workstations. These platforms typically come loaded for bear, and on the current platform, six-core and eight-core CPU options are possible. These chips are typically more expensive than the top-end mainstream CPUs.

While it is nice for Intel's customers and investors that it offers this lineup, the pace at which Intel updates the chips on this platform leaves much to be desired. No only that, but in all likelihood, this policy leaves money on the table.

Intel's data-center group has the right idea ...
Diane Bryant, who heads up the data-center group within Intel, stated the company aims to deliver at a minimum a 20% increase per generation. In other words, the Xeon E5-2687 v3 is expected to be at least 20% faster than the Xeon E5-2687 v2 (and given the core counts and frequencies, I'd bet that it is).

Further, Intel is continually releasing higher processor tiers that it charges even more money for. Indeed, Intel's average selling price growth in the data center has been a substantial driver of its revenue growth in that business for quite a while. Those customers are obviously willing to pay for extra features and performance, and Intel capitalizes on that to great effect.

So it's rather puzzling that Intel doesn't do the same thing in the high-end PC space, where customers also seem to be willing to pay hefty sums for more performance.

What are you talking about?
So, in order to understand why I made the claim that I did in the prior sentence, take a look at the table below; it lists each of Intel's highest end "extreme edition" desktop processors with core count, frequency, and launch date since the "Westmere" generation:

 Source: Intel, AnandTech

Now, since the core architectures have evolved, it's not really easy to get a sense of the generational improvements here. So, using results of the popular multithreaded Cinebench performance test from AnandTech's 5960X review, let's take a look at how the performance of the "extreme edition" Intel chips have evolved over time:

Source: AnandTech, author calculations 

While the jump from the 4960X to the 5960X was solid (new architecture coupled with 33% more cores), it is pretty easy to see that performance here hasn't been advancing anywhere close to the "20% minimum" that the company targets with its data center products.

I contend that if Intel were to be more aggressive in pushing performance on a given platform in each generation -- even if it's by offering higher-core count, higher-priced products -- it might be able to drive revenue and gross profit upside from current levels. 

What can Intel do about it?
In theory, a customer who purchases a high-end desktop motherboard can plug in most any Xeon processor into said board and reap the same performance increases that the data center customers get.

Now, there's a problem here. Keep in mind that the HEDT processors come multiplier unlocked, which means that the user can adjust the clock rate of the chip up -- often pretty significantly with good enough cooling. Off-the-shelf Xeon processors come multiplier locked. So for enthusiasts who want the highest per-core performance possible, the HEDT chips are the only way to go.

This, in my view, is silly. Intel typically offers a Xeon 2687W model in each processor generation aimed at workstations. It's usually a frequency optimized part and is rated at a fairly high thermal design power. It also usually has a higher core count than the best HEDT chip.

So, why doesn't Intel just offer an unlocked variant of this family of processors for somewhere between $50 and $100 more than the locked version? There are likely enough enthusiasts out there that would be willing to buy this part, and it's hard to imagine that it would drive much in the way of incremental costs for Intel. 

Now, if Intel wanted to not brand such a product "Xeon," it would be straightforward to remove the chips ability to work in dual-processor configurations, unlock the multiplier, and sell it as a $1,500 Core i7 "extreme edition" part. I think there are enough hardcore enthusiasts out there to make such a chip worthwhile to build and sell.