I'll freely admit that I'm a technophile who feels a compulsive need to own a good portion of the latest technologies out there. Fortunately, my addiction to investing typically overwhelms my desire to own a whole slew of fast-depreciating pieces of technology. However, a couple of weeks ago, after all of the fanfare and hype about Intel's (INTC) Haswell had finally overtaken me, I broke down and purchased what I believed to be the most gorgeous Haswell-based Ultrabook on the market - the Sony (SNE) Vaio Pro 13. I needed a system that was light, portable, and able to offer me significantly better battery life than my gaming laptop. In this article, I'd like to detail my impressions of this Haswell-based Ultrabook and to express my concerns over the state of the PC industry today.
The Ultrabook Vision Achieved - With Caveats
I'll be upfront about it - this system is incredible. It offers fantastic battery life, great performance (although I could always use more…), and it's probably the thinnest notebook that I've ever used. Further, Microsoft's (MSFT) Windows 8 is really a whole lot better with a touch screen than without. In fact, I find myself using the touch screen about half of the time - and this is a clamshell we're talking about here. It also helps that Sony actually put a nice 1920x1080 IPS touch panel with pretty solid colors and brightness - instead of the junk 1366x768 TN non-touch panels that continue to plague the vast majority of the PC ecosystem. While the new Haswell chip (i5-4200U in this machine, to be precise) enabled Sony to do some pretty awesome things, it's clear that a close collaboration between the system designer and the chip vendor is required to really pull off a magnificent device.
Unfortunately, while I'm gushing over this Vaio Pro 13, the vast majority of the Ultrabooks selling today don't come with Haswell - they're still based on the much less efficient Ivy Bridge platform, and the systems themselves always tend to have one or more deal-breaking flaws that keep them from really achieving greatness. So, that leads me to two topics for this discussion:
- Where's Haswell?
- Can this greatness trickle down to more mainstream price points?
So, Where The Heck Is Haswell?
That's the $34B question - where's Haswell? As far as I can tell, there are very few Ultrabook designs shipping with Haswell - only the Sony Vaio Pro 13 and the Acer Aspire S7 (and the MacBook Air 11" and 13" if you count them as Ultrabooks). Sure, there are some non-Ultrabook systems using the low power "U" series of Haswell chips, but by-and-large, if you walk into your typical Best Buy (BBY), you're really not going to find much in the way of Haswell-based Ultrabooks. You'll see some Ivy Bridge based designs such as the Lenovo (OTCPK:LNVGY) Yoga 11 and 13, the Toshiba Kirabook, and a few ASUS models, but really the only devices that I'm seeing Haswell in are the big, bulky gaming laptops that utilize the very high TDP parts paired with a discrete GPU (i.e. not Ultrabooks). In short, the same parts that Intel released for the desktop (and these are certainly available in ample quantities - surprise, surprise) are the ones making it into the notebooks, while the really fancy, low power parts remain very elusive.
After a lot of digging, I doubt very much that Intel is having any problems actually building these things, since they're now being built on a very mature 22nm process. Nothing really seems to have gone wrong on the design side of things either, since Intel demonstrated working samples of Haswell back in 2011 (and, you know, Apple, Sony, and Acer seem to have plenty). It is my view that OEMs simply have too much Ivy Bridge inventory to work off, which is why the they felt it necessary to essentially try to sell the same systems that, mind you, weren't selling, during the all-critical back to school season.
This, unfortunately, likely created a rather unfortunate catch-22 situation. Intel can't sell the good stuff (i.e. Haswell for Ultrabooks) until the Ivy Bridge inventory was good and worked off at the OEMs, but these same systems - which sport largely bland industrial designs, offer poor battery life, and are generally undesirable compared to a really nice table - aren't doing too hot, and as such delay the OEMs' introduction of truly good, Haswell-based systems. It is my view that the OEMs will start to offer the new Haswell based systems in earnest during the Christmas shopping season in a last major push for the PC/Ultrabook ecosystem as a whole for the year.
Haswell's a great product for Ultrabooks, but until the great systems sporting it in its power-sipping form, the PC ecosystem will continue to suffer. However, there's still another major question on everyone's minds and it begins with a "P".
What About Price?
Haswell is a complex chip and as such - from both a COGS and from an R&D perspective - it costs more to make that a typical ARM (ARMH) or an Intel Atom based system-on-chip. This means that Haswell based systems are - rightfully, I might add - priced higher than the bevy of ARM/Atom based devices. The very high end of the market is largely saturated, while the low end of the market - that is, where Intel sells die-harvested (or artificially castrated - which is more likely given Intel's typically exceptional yields) parts with features disabled - is still growing like a wildfire. Unfortunately, while these parts are offered fairly cheaply, they are still too expensive to power a really inexpensive device without some serious sacrifices being made. Do you want a $330 Celeron based thin-and-light notebooks? Sure thing! Are you willing to live with sub-par battery life and an even worse screen? The CPU/SoC still takes up too much of the bill-of-materials to really enable a good system at these price points, particularly when Intel is selling both a 100+mm^2 CPU/SoC die + 65mm^2 PCH - the latter of which cost a fortune to build as Intel uses in-house tools, lays out the majority of the core by hand, and so on.
In an ARM-less world, this would probably fly just fine - nobody really had a choice. However, the ARM vendors (think Qualcomm (QCOM) and Nvidia (NVDA)) were used to designing for cheap. I don't just mean from a COGS perspective, but from an R&D perspective. Intel cannot win a battle against Cortex A15's and Qualcomm Krait's with an expensive, high performance core that is being shoehorned into an expensive package. But it can compete - and, in my view, even win - with a core and accompanying system-on-chip built to go head-on with these designs.
That's where Atom comes in. Not only will Intel's upcoming Bay Trail system-on-chip designs based on the upcoming Silvermont processor core likely be much smaller than its Haswell-based brethren, but it will be much more cost effective from an R&D standpoint. This is part of why I find the argument that Bay Trail will necessarily "cannibalize" traditional Core chip sales a little bit fishy. Sure, if somebody is buying, say, a $300 Bay Trail tablet over a $999 Ultrabook, then that's absolutely a problem. However, the trend seems to be much more that the entry level/mainstream PCs (i.e. the ones used for basic tasks and not for business/performance users) that use these artificially disabled versions of these very expensive processors. That is, from a raw gross margin dollars perspective, a shift from a low end PC into which Intel is selling two pretty large dice for a combined total of ~$50 to a tablet using a single-chip, cost/size optimized die for $30 isn't necessarily bad. Although, of course, I will have to get a measurement of Bay Trail's die size before I can really make any definitive conclusions here.
More importantly, though, the new Atom core is built for a low power design point. While the Core processors have a surprisingly large dynamic range, the best chips for this space will be those designed for the right thermal envelopes. Leaks suggest that the highest end tablet SKU (Atom Z3770) will sport an SDP of 2W - less than half of the lowest SDP Haswell which comes in at 4.5W (and of course the Atom part can be marketed as "quad core" since it is a quad core device against the Haswell which is a "dual core" - albeit a much faster pair of cores).
How does this all tie into my Haswell rant? Simple. Intel plans to release notebook oriented version of the Bay Trail product (dubbed Bay Trail-M) that should go into fanless, super-thin notebooks (not Ultrabooks though). These chips will be much cheaper to make than the low end "big core" parts, and they'll be much lower power. This means that, if I'm right, Intel can still make the same raw gross margin dollars per chip, but sell the chip for a lower price. This frees up a good part of the bill of materials for things such as flash storage, a nice high resolution screen, and a smaller chassis (not needing a fan saves on both the fan as well as the rest of the system). In short, low end PCs and convertibles become an order of magnitude more desirable than they are today - which could help to win back wallet share in the 10"+ category of devices.
Conclusion - Haswell Needs To Get Here Pronto With Bay Trail In Tow
I believe that once Haswell starts to more broadly permeate the notebook ecosystem that it could help to drive customers who have been holding off on upgrading their current systems to finally make that new higher end purchase. The current crop of Ivy Bridge systems just isn't desirable, and I'm really surprised that the OEMs decided to actually go through the entire back-to-school season with stale designs. However, while Haswell and its successors will be important for the future of higher end client devices (and I'm shocked that most designs still use Ivy Bridge), the real show starts when Intel finally releases its Bay Trail chip. The bottom line is that the netbook idea (i.e. cheap enough laptops to drive secondary and tertiary device purchases) was actually a pretty good one - it drove volume. Unfortunately, the original Atom was pretty lousy from a performance perspective, and the systems built around it looked archaic once Apple (AAPL) dropped its iPad bomb. With Bay Trail, Intel gets into the tablet game proper, but probably equally importantly, low end PCs can finally become interesting again, particularly in "convertible" form. At that point, there will be very little distinction between tablet and "PC" (since many people buy separate keyboards for their iPads), and it'll be a question of whether the OEMs and Intel can do a good job marketing these things.
Time will tell, but - if it weren't obvious already - my money's on Intel.