A report from Marketwatch yesterday has a note about JMP Securities analyst Alex Gauna downgrading Intel Corp (INTC) to market perform or neutral on the "discovery of technology roadmap issues that add to an already challenging outlook." The article at Marketwatch does not go any farther than that. But, while I was doing my daily crawl through the various technology forums - you know, so that I can have a clue as to what I'm talking about - I came across this post which further clarifies Gauna's reasoning. It also helps to confirm what has been lurking between the lines of most of the early leaks about Intel's upcoming Haswell CPU's, namely that things have not translated from design to implementation as well as we have been led to believe.
The importance of Haswell in the evolution of Intel's chip offerings is that its design imperatives centered on improving two things Intel's chips, frankly, do not do well. They are manage power and process graphics. In my previous articles on the subject I called into question the product viability of Haswell's improved graphics capabilities. While I applaud Intel for a huge improvement in theoretic - and initial testing - performance, one has to be objective and say that the performance is still not good enough overall.
I guess this is why some people are still bullish on Nvidia's (NVDA) discrete graphics business, as the need for them will exist even with Haswell's release. In the short term, they are probably right. In the end, however, when judged on a price/performance/watt basis it is looking more and more that AMD's (AMD) Richland and Kabini APUs will be more than capable of holding their own in the channel through the end of 2013.
In all of the discussion about the potential shortcomings of Haswell, however, no one questioned Intel's power management improvements. I'm not sure why since Intel's record in this regard has been so stellar in the past. I still contend that Microsoft's (MSFT) Windows 8 rollout was severely hampered by both AMD's dysfunction and Intel's unwillingness to truly embrace mobile computing.
The myriad of products that hit the market were crippled in some fundamental way - power management, performance, gaming prowess, etc. - allowing Google's (GOOG) Android and Apple's (AAPL) iPad to continue gobbling up market share.
Both Haswell and the upcoming Bay Trail replacement for the current Atom CPU have got to be good or the market will continue to move forward without them. Gauna's report, an excerpt of which I linked above and please read it carefully, is quite clear on this.
So, I keep asking this question, if Haswell's graphics are nothing to write home about, its CPU performance gains are not compelling for enthusiasts to upgrade to and now we find out that its power management scheme is less than advertised (by how much no one is saying) then what is the benefit of the chip to the market?
This is especially relevant when we note that AMD's Richland APUs made big enough strides in its power management schema such that significant performance gains - 15% to 25% for desktop parts - are being seen over last year's Trinity APUs on, essentially, the same 32nm process. While we still have yet to see Richland officially benchmarked by the technology intelligentsia the results we have seen, done by individuals, confirms AMD's released numbers, at least at the high end.
It is important to note that relative performance gains will rise as the TDP of the chip goes down because as you move farther away from the chip's operational limit, it's ability to run at the maximum allowed clock speed increases. So, a Richland A10-6700 outperforming a Trinity A10-5800K by 15% to 25% lends a lot more credence to AMD's claims of up to 40% improvement in lower wattage parts.
And this is directly the market that Haswell has to be able to compete in and compete well. At the low end there is the power-efficient Jaguar and the mid to high end Richland and by year-end Kaveri. In short, AMD has closed the overall performance gap in the areas the market deems important at price points that Intel is uncomfortable with.
What's important to take away from this is not that Intel is bad and AMD is good or anything like that. It has to do with building the right product for the market at the right moment in time. Is Haswell going to be used in a lot of devices? Yes, of that there is no question. The question is whether the OEMs will be able to build them to meet their product goals.
At this late stage of the game contracts have been signed and parts are going to ship, the question is at what price? And, if there is a performance issue, who is going to eat the margin to offset it, Intel or its OEM partners? In the past it was always the OEMs. Today? I'm not as convinced Intel still has the leverage it once did.
A blog post over at Microsoft News the other day caught my eye, it was the press release from Vizio on its spring lineup of touchscreen "thin and light" laptops - Macbook Airs for the Windows 8 set - and what jumped out at me were the prices. Vizio is one of AMD's new OEM partners - this gives us a great opportunity to compare prices. Presumably, everything in these machines is the same except for CPU/GPU, and this gives us, for the first time this year, a look at just how much that Intel branding will cost you.
Vizio Spring Pricing
Whether or not Haswell has serious power handling issues or not is still unknown but we have to consider it in light of Mr. Gauna's report and downgrade. What I can tell you is that in the real world, Intel has a serious pricing problem and if that's the case, the shareholders have a serious valuation problem.