While Intel's (NASDAQ:INTC) management tells a good story about their opportunities in mobile areas like tablets and phones, the reality of non-competitive parts litters their dreams with harsh reality. There can be no question Bay Trail, and its unattractive cousins, have visited Intel with disaster, regardless of how Intel wishes to frame it. On the plus side, the Silvermont architecture in Bay Trail is being superseded by Airmont in new processors like Cherry Trail, the 14nm shrink of Bay Trail.

Based on the performance in GeekBench, Cherry Trail's performance is pretty disappointing, and strongly suggests the CPU is essentially only a shrink from Bay Trail. Already non-competitive with AMD's (NASDAQ:AMD) Puma line of processors in terms of performance, this seems to be more bad news for Intel's failing MCG group. It seems that way at first glance, but it may not be.

Adding to Haswell and Broadwell, Intel has continued its streak of disappointing processors, particularly when the inferior Bay Trail is considered. However, like Haswell and Broadwell, the lack of absolute performance improvement for Cherry Trail has more to do with where Intel wants to put the processor, rather than a failure to improve on the predecessor. In reality, while the top performance did not improve significantly in any of these parts, they all show significant improvements that make them far better suited for certain markets.

Also, before we place too much importance in these benchmarks and view the chip as a regression in some ways, it's important to remember that early benchmarks do not tell the whole story. While it is clear Cherry Trail will not have strong CPU performance, there is still room for improvement as Intel tweaks the design for released products. It's also very likely Intel will hit higher clock speeds at similar power levels as the 14nm process matures. Most importantly, we also have to consider what Intel is targeting with this line of processors.

The processor line all starts with the Silvermont architecture used in Intel's Bay Trail. This design is remarkable - in the sense it's likely the worst CPU design of this century. This architecture has completely failed in its first iteration, at everything but power use, and against all competitors. Intel backers like to think it moving into 46 million tablets last year shows how successful this design can be, but this conveniently ignores they lost $4.2 billion for the year in MCG, almost entirely caused by this processor design. In fact in Q4 of 2014, Intel generated negative revenue in this division, indicating they had to pay companies to take their processor. This is fascinating in a negative sort of way, while at the same time being the purest indicator I have seen of a failed design. This exceeds the level of ineptitude even Pentium 4 and Bulldozer were able to achieve, and it's not like they didn't make an effort to top the chart. It is interesting all three of these designs took significant risks by departing from established processors, with regrettable results.

Part of Bay Trail's problem is the size, being over 100mm2, or the same size as AMD's better performing Puma parts. The power use, however, is quite good, leaving Intel with two major problems - performance and size. With this line essentially shut out of higher-end parts, and Intel having shown no evidence of an effective low-end chip design team, the company could not hope to compete on performance. Far better to try for what's attainable than fail at everything like Silvermont did. Intel instead did the best thing they could get the size down to where they can make the parts on an inexpensive basis. As mentioned, the early performance results indicate strongly the CPU has been left significantly the same, which implies this core will probably be around 50% of the size of the Silvermont core due to the fabrication shrink. Of course, while Intel can and may make some SoCs using Airmont as big Bay Trail, this in no way diminishes that they can make significantly smaller SoCs as well. For example, if we see a 100mm2 Cherry Trail for desktops, this in no way indicates Intel cannot make a tiny part for phones, since much of that size will be occupied by logic that can be easily removed or reduced in size, like the GPU, for the phone parts.

So, Intel made the right move by not attempting to compete with AMD in performance. They simply cannot do it without severe compromises for the market Intel really needs to target. Even in PCs, and tablets, both AMD's Jaguar/Puma and Intel's Bay Trail/Cherry Trail generally target the low-end of the market, making cost relatively more important than performance. More than that, although Intel did not improve the CPU much, the GPU is rumored to be three times the size of the Bay Trail GPU, giving Cherry Trail a clear advantage over AMD's current Puma line. Of course, this advantage may not hold against AMD's next releases. So, although Bay Trail will not compete well with AMD's low-end line in performance, it still may be "good-enough," and still may find favor when accompanied by an attractive price tag that does not entail Intel losing money on the part.

Intel is accepting they have no real hope for the high-end in phone and tablets, giving themselves a lot of hope in the low-end. This is significantly helped by their broad technology portfolio, particularly in connectivity. In addition, Broadwell covers the higher end tablet market (4.5 watt and up) quite well. It's also not clear if tablets will remain a growing market or will start to recede as it did in Q4 of 2014. In low power laptops and desktops, AMD will still present a better part than Cherry Trail, like it does now, but then there's Broadwell again to limit just how much AMD can hope to get. So, while AMD will continue to do well with their Jaguar/Puma line as they release new parts in 2015, it should not improve much from the current success they are having. Besides, what Intel does sell into this segment with Cherry Trail promises to be considerably more profitable than Bay Trail, by virtue of the lower cost. Naturally, this assumes they can keep the other parts of the chip within reason.

Where Airmont really does help is in the lower-end phone market and possibly low-end tablets. The low-end phone market is huge, and growing, and this has to be the primary focus for Intel. There's also no help from big brother there, so Airmont has to do this on its own. Keeping the CPU core size small is absolutely essential. The performance of this chip will be more than enough for this segment, and by eschewing performance for size, Intel also will improve power characteristics as well. Those two are critical, and Intel made the right choice by favoring them. They also made the right choice in which market they really wanted, because it's already enormous and growing very fast. It's also entirely accessible to Intel, unlike the higher end, where big players who make their own chips have a high market share.

I'm not often happy with Intel's management decisions, and perhaps bizarrely see the lack of performance of Cherry Trail as not only inevitable but also highly desirable considering all things it implies. In previous articles, I had stated there my strong belief Cherry Trail could not improve CPU performance significantly and be a good part for Intel. Being successful in low-end phones strongly implies a very small core to allow a competitive cost structure for the market. Competing against AMD's Puma line in performance would have required extensive redesign, and an increase in size and cost. Since AMD does not sell this part for phones, and will be moving to 20nm in 2015, and FinFets after that, it's clearly not a battle Intel could win while pursuing phones. It was far better to concede CPU performance, while improving the GPU, and likely cost, while optimizing for the phone market with this chip. Silvermont tried to do too much and failed at everything. Airmont appears to be more targeted for one segment.

This does not guarantee Intel success in the lucrative phone market, but given where they were, it was the right choice. In the end, although we all like to see performance parts, the only performance that really matters as an investor is the stock price. Intel made the right choices to deliver that with Airmont.

http://seekingalpha.com/article/2953786-intels-cherry-trail-disappoints-with-performance-is-this-as-bad-as-it-seems?auth_param=13qoi6:1aeujgb:a3964d013d10985d1a714c1813f999ed&uprof=82&dr=1