Xbox One graphics capabilities, odd SoC architecture, and bus bandwidth confirmed by Microsoft

Microsoft has finally lifted the curtain on the Xbox One, with a great deal of technical detail on display at the Hot Chips conference. For the first time, we’ve got a view into how the architecture is laid out and what its capabilities are. The chip is built on a 28nm process by TSMC and measures a sizeable (though not enormous) 363mm sq. It’s capable of running at as little as 2.5% of active power thanks to aggressive power gating — leaving the system running won’t destroy your power bill. The chip is built on TSMC’s HPM process, which is designed to offer simultaneous benefits of high performance and low leakage power.

What follows is a quick dive into what we now know.

Inside the SoC

The Xbox One SoC appears to be implemented like an enormous variant of the Llano/Piledriver architecture we described for the PS4. One of our theories was that the chip would use the same “Onion” and “Garlic” buses. That appears to be exactly what Microsoft did.

That slide is from Llano/Piledriver. Microsoft’s slide from the Hot Chips presentation is here:

This image courtesy of SemiAccurate. Better images hopefully coming soon.

Here’s the important points, for comparison’s sake. The CPU cache block attaches to the GPU MMU, which drives the entire graphics core and video engine. Of particular interest for our purposes is this bit: “CPU, GPU, special processors, and I/O share memory via host-guest MMUs and synchronized page tables.” If Microsoft is using synchronized page tables, this strongly suggests that the Xbox One supports HSA/hUMA and that we were mistaken in our assertion to the contrary. Mea culpa.

You can see the Onion and Garlic buses represented in both AMD’s diagram and the Microsoft image above. The GPU has a non-cache-coherent bus connection to the DDR3 memory pool and a cache-coherent bus attached to the CPU. Bandwidth to main memory is 68GB/s using 4×64 DDR3 links or 36GB/s if passed through the cache coherent interface. Cache coherency is always slower than non-coherent access, so the discrepancy makes sense.

Let’s talk about the cache blocks at the bottom of the image, as there’s some very curious data here.

The interesting ESRAM cache

First, there’s the fact that while we’ve been calling this a 32MB ESRAM cache, Microsoft is representing it as a series of four 8MB caches. Bandwidth to this cache is apparently 109GB/s “minimum” but up to 204GB/s. The math on this is… odd. It’s not clear if the ESRAM cache is actually a group of 4x8MB caches that can be split into chunks for different purposes or how its purposed. The implication is that the cache is a total of 1024 bits wide, running at the GPU’s clock speed of ~850MHz for 109GB/s in uni-directional mode — which would give us the “minimum” talked about. But that has implications for data storage — filling four blocks of 8MB each isn’t the same as addressing a contiguous block of 32MB. This is still unclear.

The other major mystery of the ESRAM cache is the single arrow running from the CPU cache linkage down to the GPU-ESRAM bus. It’s the only skinny black arrow in the entire presentation and its use is still unclear. It implies that there’s a way for the CPU to snoop the contents of ESRAM, but there’s no mention of why that capability isn’t already provided for on the Onion/Garlic buses and it’s not clear why they’d represent this option with a tiny black arrow rather than a fat bandwidth pipe.

Even with this new information, the use and capabilities of the ESRAM remain mysterious. It’s not clear what Microsoft expects it to be used for — if it’s for caching GPU data, why break it into 8MB chunks, and why does the CPU have a connection to it?

CPU, GPU, what we now know

The GPU and CPU blocks look like we expected and detailed in our Xbox One graphics preview. DirectX 11.1+ is listed as supported and the GPU stats are matched well to the prediction that this was a Bonaire-derived core. Microsoft claims a total of 15 non-CPU processing blocks, which works out to 12 for the GPU, two for audio, and an “other” possibly related to I/O or Kinect. The CPU, as previously reported, is an eight-core Jaguar variant. Microsoft claims the audio engine contains multiple discrete function blocks with their own impressive hardware stats, but a description of audio capability is beyond the scope of this discussion.

Save for the clockspeed bump on the GPU, the data from earlier leaked slides was accurate.

The big picture takeaway from this is that the Xbox One probably is HSA capable, and the underlying architecture is very similar to a super-charged APU with much higher internal bandwidth than a normal AMD chip. That’s a non-trivial difference — the 68GB/s of bandwidth devoted to Jaguar in the Xbox One dwarfs the quad-channel DDR3-1600 bandwidth that ships in an Intel X79 motherboard. For all the debates over the Xbox One’s competitive positioning against the PS4, this should be an interesting micro-architecture in its own right. There are still questions regarding the ESRAM cache — breaking it into four 8MB chunks is interesting, but doesn’t tell us much about how those pieces will be used. If the cache really is 1024 bits wide, and the developers can make suitable use of it, then the Xbox One’s performance might surprise us.

This technical unveil confirms much of what we suspected about the chip, but throws us for some curves in other areas. All in all, a great session.

Storage Pricewatch: Get out your wallet, hard drive and SSD prices have dropped

The last time we dug into storage prices was back in April, when the trends for both hard drives and SSDs showed a modest decline for the former as prices returned to pre-flood levels and a relatively stable track for the latter. Since then, we’ve seen reports of volatility in the SSD market.

We’ve teamed up with Dynamite Dataonce again to examine long-term price trends and what they show us about the storage market.

The SSD split

SSD prices are simultaneously increasing and decreasing depending on which capacity you focus on. This is a fairly new trend. Previously, high capacity drives were typically the most expensive as these products were aimed at enterprise/business customers or offered the highest performance money could buy. Low capacity drives were typically cheaper, since they offered fewer features or lower performance. Now, that’s changing.

When we break out the price per GB at the 128, 256, and 480-512GB capacities, we see three different cost trends. The 128GB SSDs have been the most volatile, with average prices ranging between $1.05-1.13/GB over the past six months. Costs had fallen to around $1.07 from a high of $1.12 in late July. 128GB drives tend to be a touch slower than 256-512GB products since they have fewer memory channels, and they are often based on older NAND and slower controllers.

We’ve dropped 60GB drives entirely from this comparison due to age and price; 60GB drives are often $1.50 per GB or higher now as manufacturers shift to higher capacities. The 256GB drives have also edged upwards compared to where they were at the beginning of the year. Prices have peaked and fallen again several times in the past few months, with total pricing currently sitting just under $1 per GB. At the 512GB level we see the slowest, most consistent trend — a general reduction in price with current prices hovering around 80 cents per GB. There are several factors driving this. Scan NewEgg and you’ll find a number of cheap, high-capacity SSDs based on old technology as well. This has apparently become something of a popular move — build out slower versions of a drive with huge storage space and offer it at a bargain price.

Where prices go from here

There are two big reasons to be optimistic about the future of SSD pricing. First, there’s TLC NAND. While I was initially dubious about the long-term efficacy of TLC (triple level cell) memory, Samsung’s 840 and 840 Evo have convinced me that this NAND has a strong place in consumer products. The Evo, in particular, is a great example of how pairing a TLC drive with a small SLC (single level cell) NAND cache can improve performance dramatically and guard against problematic wearing.

Second, there’s vertical NAND, also called V-NAND. As the name suggests, this is NANDthat tilts the cell structure up on edge. V-NAND has only just entered production, so it’s not going to hit consumer markets yet, but the long-term feedback is promising. It’s not clear if TLC and V-NAND can currently be deployed in the same hardware, but the ITRS has made it clear that it expects V-NAND to drive SSD prices downwards over the next few years. This kind of improvement is critically important given that NAND flash faces major scaling problems below the 19nm node.

One thing to note is that while pricing in these segments as a whole has only fluctuated by ~8%, fluctuation on some drives is considerably more volatile.

The below chart shows the biggest movers and shakers for the entire SSD market since February. Keep in mind that pricing can vary dramatically depending on drive model but, again, the largest capacities tend to show less shifting than the higher ones.

I suspect the 100GB and 110GBs listed here might be OCZ’s RevoDrives, which have been volatile of late. Still, it can pay to keep an eye on favorite drives when looking for a deal.

Prices should continue to edge downwards through 2013 and into 2014. I don’t expect any enormous drops but if Samsung’s 840 Evo continues to win favor, it could push the average selling price lower.

Hard drives: Desktop in great shape, but laptop drives show troubling trends

Desktop hard drives have returned to pre-flood levels and continued downwards from that point. I don’t have an updated version of our hard drive graph from April, but a quick check of NewEgg shows 2TB drives starting at $79.99, or roughly 3 cents per GB. 4TB drives, even WD’s high performance Black family, are available for 6-7 cents per GB. That gap means hard drives remain more than an order of magnitude cheaper than SSDs. SSD performance has increased much more rapidly than spinning disc performance, but the rate of change may slow dramatically over the next few years. TLC and V-NAND are designed to improve NAND density and cost, but not its speed. This is particularly true with TLC, which is markedly slower than MLC.

There is, however, one negative trend I want to call out here, as it impacts laptop hard drives in particular. Back in 2009-2010, companies like Seagate confidently predicted that nearly 50% of laptop HDD volumes would be 7200RPM drives by now. Instead, the manufacturer has since announced that it intends to phase out 7200RPM drives altogether by the end of the year. Already, we see the trend kicking in — of the 212 laptop HDDs in stock at NewEgg, only a handful of them are 7200 RPM models. Seagate’s Solid State Hybrid Drives (SSHDs) are available, but they’ve staked out a price point well above $100.

Fortunately, there are still 7200RPM drives from Western Digital, but Seagate’s decision to exit that market could be a sign of things to come. While SSDs remain a far better performance choice in mobile, 7200RPM drives are still noticeably quicker than 5400RPM counterparts, and it’s somewhat frustrating to be stuck with just two choices — bottom end, or SSD.

Low temperature combustion could drive efficient, clean diesel engines back into the limelight

Cold fusion, despite probably being impossible, is a good idea. Take a process known to create a meaningful amount of power and remove the harmful byproducts (like explosions), and you’ve got a winner in production. The same logic holds true throughout the energy world: we have solutions, but they are often saddled with too many downsides to be workable in the real world. The combustion engine is a great example of this, making use of a readily available and energy-dense fuel while creating increasingly dizzying amounts of pollution. If only we could change the process of extracting that energy, we could make fossil fuels at least moderately more sustainable.

Sandia National Laboratories has released a review article detailing the feasibility of Low Temperature Combustion (LTC) diesel engines. It should really be called lower temperature combustion though, but as you’ll see ,even a moderate decrease in heat can have a powerful effect on pollutants. Sandia concludes that LTC diesel could be a major step forward for diesel-powered industry. The study suggests that low-temperature combustion could drive diesel engines forward in a way desperately needed for both the environment and for competition with gas and electric competitors. Since so much of the industrial sector runs on diesel fuel, particularly in transportation of goods, that’s an important step forward for a technology that is often thought of as on its way out.

Two of the main pollutants in diesel exhaust are both the results of heat. One, nitrogen oxides (NOx molecules) are formed primarily at the highest-temperature point in the engine, the flame itself. When released into the air, these compounds act as inveterate pollutants but also react with sunlight and other particles in the air to form smog. Particulate matter (PM), on the other hand, results from the roiling cauldron of the combustion chamber, and the small pockets of too-high fuel concentration that arise during the combustion process.

According to the Sandia researchers, both of these problems can be mitigated by meddling with the engine’s fuel-air mixture. The engine’s own exhaust can be fed back into the fuel-air mixture to absorb some of the heat of the combustion process and ferry it away. This keeps many of the nitrogen oxide products from forming in the first place. Mixing the fuel with additional air just prior to combustion ensures there will be no fuel-rich regions to give rise to airborne particulate pollutants. It’s an elegant refinement of the basic process, and one of that would not make the engines much more expensive to produce — though it might very well make them more difficult to maintain. The air-injection mixing apparatus is not unlike the fuel injection system in current engines, which is already one of the most finnicky parts of the whole arrangement.

A conventional diesel engine, with a fuel injection system that results in uneven fuel density during combustion.

LTC has traditionally increased the amount of carbon monoxide released, the main greenhouse gas in car exhaust, along with unburned hydrocarbons. It’s frustrating because both of these problems reduce efficiency, which is one of the main advantages of diesel in the first place. This review looks at research showing an improvement in this area after they figured out that it was due to poor mixing of the fuel-air mixture. By splitting the whole the injection process into a main spurt followed by several smaller ones, they were able to make a more even distribution of the fuel-air mixture that burned to completion all throughout the cylinder.

This research concerns itself with diesel engines, which can be more efficient but also more polluting than gasoline engines that incorporate high-end catalytic converters to filter their exhaust. If diesel can be made less polluting in turn, the biggest objection to its use on the large scale will disappear. A largely diesel-driven nation would consume less fuel per kilometer, which translates to a big chunk knocked off total national emissions. More immediately, it offers a possible solution to high emissions of vehicles that already use diesel, particularly long-haul cargo trucks.

To build their detailed maps of the combustion process, needed to figure out things like the under-combustion of fuel nearest to the injection site, researchers have had to pioneer some truly impressive modes of analysis. The use of two-photon laser induced fluorescence was used for the first time to track the movements of combustion byproducts like formaldehyde, which in turn showed the researchers exactly how the fuel was burning, where, and when.

Tesla readies mass-market Model E while other EV makers tread water

Tesla Motors has a new model pending, the Tesla Model E, according to a trademark application filed earlier this month. Does E stand for Economy? Efficient? Easy? Environmental? It’s in the trademark office category of “Vehicles and Products for locomotion by land, air or water,” but we’ll assume this not a competitor to Terrafugia and the Model E will remain earthbound. Most likely it will be a more affordable Tesla electric vehicle. You can’t grow market share with cars that cost more than $65,000 entry point of the highly rated Tesla Model S.

Tesla currently has a model lineup one car deep, the Model S, a five-door hatchback with a starting price twice that of the average car sold in the US. With more powerful batteries and home chargers, you can spend $100,000. A second model is announced: The Tesla Model X crossover with three rows of seats and room for up to seven passengers. Price isn’t announced but expectations are that it will fall into the realm of the Model S. First production was pushed back a year and Tesla now says while we’ll see some Model X’s in 2014, volume production won’t happen until 2015. Tesla said the push-back was to insure Tesla will be able to produce 20,000 Model Ses in 2013 and to underscore Tesla’s commitment to profitability.

Model E due around 2015

When the Department of Energy set up loans to Tesla in 2009 – loans, not Detroit-style bailouts – Tesla talked about producing a range of electric vehicles: urban delivery vans, convertible or roadster (similar to the first Tesla, the Roadster of 2008-2012), minivans, and a smaller EV-sized like a BMW 3 Series, Acura TL, or Chevrolet Cruz. Most believe this will be the downsized, economy Tesla with a price in the neighborhood of $30,000. It would seat four or five with the same comfort you’d expect in an economy car — pretty good if the economy car is a Honda Civic, snug in back if it’s a Ford Focus. Right now, $30,000 gets you a comparatively spartan EV from other automakers and the assumption is Tesla wouldn’t hurt its brand with a cheap interior.

The Model E would most likely be an electric vehicle only. Tesla’s plans in the past included small combustion engine range extenders as on the Chevrolet Volt but no such cars were built and Tesla’s current plans are for EVs only. Instead, Tesla is working on better batteries, quicker charging, replaceable batteries, and more charging stations.

Tesla has the look of a winner (no missteps so far)

Some automakers’ EV sales have stalled, others have done well, and Tesla has done well in virtually every aspect. It is likely to meet its goal of selling 20,000 Model S sedans this year. In California, the Tesla Model S outsells every model from Porsche and Jaguar (also Lincoln, Land Rover and Volvo), with 4,714 Teslas registered there in the first half of 2013. The Model S got the highest possible ranking on the US crash tests and Tesla, not content with that, issued a press release that its interpretation makes Tesla safer than the other top scorers. Tesla also got the highest rating in a Consumer Reports road test and won a host of car of the year awards, including from Extreme Tech.

Tesla sees itself as much a tech company as it is a car company with production facilities in the old GM-Toyota NUMMI plant in Fremont, California. USA Today notes Wall Street values Tesla more than Google based on revenue: $20 billion in valuation (market cap), or 10X expected annual sales. Google has a market cap of $300 billion, more like five times expected revenue. In other words, when you say Tesla, investors see sizzle as well as steak.

Tech has been good to Tesla’s bottom line in ways beyond creating desirable cars. Like Porsche has done for decades, Tesla has a healthy consulting business. It has worked with Mercedes-Benz (A-Class EV), Smart Fortwo (EV), Toyota (RAV4 EV), Freightliner (Custom Chassis Electric Van). Because it sells only zero-emission cars, Tesla banks environmental credits it can sell to other automakers. Jerry Hirsch of the Los Angeles Times says Tesla profits by as much as $35,000 for every Model S sold.

Tesla has suffered just one recall of about 1,000 cars for a weak rear seat latch. It also got into a public feud early this year with The New York Times over whether reviewer John Broder followed Tesla’s review procedures and kept proper notes. (Tesla kept copious notes by means of an on-board tracker that would do the NSA proud.) The upshot was that The Times came off looking less than perfect and Tesla was overly thin-skinned. Times public editor Margaret Sullivan concluded there were “problems with precision and judgment but not integrity” on the writer’s part.

Tesla needs a mass market car to grow

In a market rebounding to 16 million sales this year, there are still only a couple hundred thousand high-end premium cars sold with list prices at or above what Tesla charges. (An Audi A4 is expensive, but it’s not that high end to marketers.) The best premium luxury cars (Mercedes-Benz S-Class, Cadillac Escalade) top out at about 25,000 units a year, and Tesla is virtually there already with the Model S.

That’s why the betting is on the Tesla Model E being an affordable ($30K), compact or midsize car with exclusively electric power. It makes sense. Multi-car families might find it could be the second or third car, so long as there’s one big SUV or sedan for long trips. The more cars Tesla produces and sells, the more it can amortize R&D costs.

iPhone 5S

It’s been nearly a year since the introduction of Apple’s A6 SoC, so the recent rumors surrounding the launch of the iPhone 5S and its A7 chip next month is making the whole internet begin to froth at the mouth. Not only are we supposedly going to see a substantial CPU speed boost over the A6 processor, but sources indicate that Apple is finally making the switch to a 64-bit CPU in its mobile devices. While it likely won’t make much of a difference right off the bat, this does set the stage for faster apps and more memory down the road.

Fox’s Clayton Morris recently claimed that we’ll see a 31% increase in speed over the iPhone 5′s A6, but that alone is nothing to write home about. Apple’s next chip will obviously be faster, but what else can it do? 9to5Mac is reporting that while the A7 won’t make the jump to a quad-core CPU, it will feature a 64-bit processor. That, in and of itself, is something to get excited about.

64-bit processors, now the standard on desktops and laptops, do have speed advantages. Applications specifically written to take advantage of 64-bit CPUs can see a nice increase in performance. For example, when Handbrake released a 64-bit version, users saw a speed boost of 10-15%. More importantly, 64-bit CPUs are capable of addressing more than 4GB of RAM. In the coming years, smartphones and tablets will undoubtedly need more than 4GB of RAM, so it’s smart for Apple to start making the transition now when most devices still only ship with 1 or 2GB of RAM. By the time iPhones start shipping with 6 or 8GB of RAM, at least some apps will be optimized to take advantage of the additional memory.

While the iPhone 5S will likely stick with a dual-core CPU, there’s no telling what Apple might do with the A7X for the iPad line. When the first Retina iPad was announced, Apple flaunted its quad-core GPU built into the A5X SoC — designed to help drive the huge 2048×1536 display. If Apple wants the 9.7-inchiPads to remain at the top of the tablet market, this year might be the perfect time to include a quad-core CPU in the A7X as well. Since the full-sized iPad has such a gargantuan space for its battery, the increased power draw of quad-core chips could certainly be easier to compensate for. With any luck, Apple’s top-notch engineers will have something truly compelling to show us next month.

Apple has been relatively low-key in 2013, so expectations are running very high for the upcoming annual iPhone announcement. We’re hearing new information about the iPhone 5S and 5C every day, and speculation is running rampant. After a year of lackluster stock performance, September’s event is clearly Apple’s best opportunity to turn the tides in its favor.

Amazon testing licensed WiFi, could lead to cheap unified global WiFi for Kindles

Amazon is privately testing WiFi technologies on licensed frequencies that belong to satellite service provider Globalstar. Called TLPS, it has the potential to become a uniform globally available terrestrial wireless network.

TLPS, or Terrestrial Low-Power Service, is a new category for defining wireless operations on an expanded version of Globalstar’s licensed 2.4GHz frequencies. Globalstar currently owns a license for 2483.5-2495MHz nationally, denoted as AWS-5. However, Globalstar has not been operating any wireless service on the band, because it was previously allocated as MSS (mobile satellite service) spectrum. Globalstar is proposing to not only convert MSS 2.4GHz to terrestrial spectrum (as Dish has done with the 2GHz S-band), it wishes to expand the range and absorb 10.5MHz of spectrum adjacent to it. That would expand the channel to 2473-2495MHz, which is a 22MHz channel.

Normally, this does not mean that much. However, there is one key fact about this band: it is globally licensed to Globalstar. That means that Globalstar can potentially turn this into a harmonized range of frequencies to operate a common terrestrial network all over the world. Of course, there are challenges to this. Globalstar is currently licensing this frequency range globally for satellite services only. If Amazon wants to launch terrestrial wireless networks all over the world on these frequencies, Globalstar will need to get approval from every regulatory authority to convert its satellite license into a terrestrial one.

The advantage of Globalstar owning the license for the channel all over the world in a frequency range that it supported by WiFi equipment, is that the cost of building the network would be very low. WiFi gear made today support the frequencies, but the firmware on the radio block usage of Globalstar’s channel in order to prevent interference with any satellite services operating on the channel. It is merely a matter of providing hardware that does permit access to this channel. No real hardware changes are required to make equipment work on the frequency range. It is even possible to use something like WiMAX as the basis of a truly global cellular system for the frequencies, since WiMAX can already operate on the band as well.

Should Globalstar convert it for use in WiFi-based TLPS by Amazon, it would become possible to cheaply cover the world in WiFi so that Amazon could use it for Kindle e-readers and tablets. It could completely bypass the mobile network operators in providing universal coverage.

RIP Ubuntu Edge, but Ubuntu is still coming to smartphones in 2014

The Ubuntu Edge, a smartphone that doubles up as as a desktop PC when docked, has unsurprisingly failed to meet its funding goal of $32 million. As the Indiegogo campaign drew to a close at midnight last night, the Edge had raised just $12.8 million — a massive funding deficit of just over $19 million. Canonical is fairly stoic about its failure, issuing a statement that $12.8 million is still the “biggest ever fixed crowdfunding campaign.” Canonical founder Mark Shuttleworth, speaking to the Guardian newspaper, says that carriers and OEMs are still “definitely interested” in making mid-range Ubuntu-powered smartphones. For now, though, the dream of an ultra-high-spec superphone like the Edge, which doubles up as a desktop PC, is dead.

There’s no denying the fact that, despite not reaching its target, the Edge accrued a ton of press coverage, introducing Ubuntu to a lot of people who had never heard of it before. In this regard, it’s hard to call the Edge campaign a failure. According to Canonical, the massive interest in the Edge has allowed the company to negotiate better prices with component suppliers for future Ubuntu-powered devices. The Edge, though, is dead: In all likelihood, these devices will be low- and mid-range devices that run Ubuntu Touch (the version of Ubuntu tailored for smartphones and tablets), and won’t have the ability to switch into desktop mode when docked.

The problem with Indiegogo, Kickstarter, and similar crowdfunding sites is that they tend to bring devices to market before they’re ready. Even Shuttleworth himself implies as much: Speaking to the BBC, he said that if the funding had succeeded, “we would have been bringing the future forward a year or two at least.” While the concept of a desktop-replacement smartphone is exciting enough to drum up almost $13 million in funding, the simple matter of the fact is that it isn’t ready for prime time. In a couple of years, $13 million will probably be more than enough to set up a production line in China; but today, it just isn’t feasible. With enough interest, it’s possible for some crowdfunded projects to come to fruition ahead of their time — such as the Pebble smartwatch and Ouya game console — but their paths to market are still usually fraught with strife.

There are some who would argue that Canonical always knew that the Edge was vaporware, and thus why it opted to use Indiegogo instead of Kickstarter. There is almost no reason to use Indiegogo, except in cases where Kickstarter refuses to host your project — and if rumor is to be believed, Kickstarter refused to run the Edge project because it is not a store. If you think about it, the Edge was never really a crowdfunded project; there was never any risks or challenges, nor an open development process. Canonical just wanted you to buy a phone. That’s how Canonical ended up hosting a ludicrous $32 million project on Indiegogo — whose previous funding record was just $1.37 million for the Scanadu Scout.

Moving forward, the first Ubuntu phones are expected to arrive in the first quarter of 2014. These, like the recently released Firefox OS phones, are likely to be mid-range phones that carriers primarily carry because they’re not beholden to Google or Apple. Any phones that have 1GB of RAM, 32GB of storage, and a quad-core Cortex-A9 or Intel Atom processor will be capable of switching into desktop mode — but as far as we know, none of these superphones are in the pipeline. In my opinion, we are probably still a few years away fromsmartphones that are capable of replacing our PCs.

For Canonical, though, which has been desperately trying to find a space in the consumer electronics market since it realized that desktop Ubuntu would never take off, the emergence of some Ubuntu-powered phones, irrespective of how lackluster the hardware turns out to be, will be a very exciting prospect indeed. Whether Ubuntu Touch really stands a chance against iOS, Android, Windows Phone, and dozens of also-ran mobile OSes, is another question entirely.

2014 Ford Fusion hybrid review: Best midsize hybrid in a crowded field

The Ford Fusion Hybrid nicely balances fuel economy, passenger comfort, and big car technology in a midsize sedan with few compromises. The lithium-ion batteries don’t steal much trunk space and fuel economy is 40 mpg or better. Few midsize cars — hybrid or not — offer lane departure warning, blind spot detection, adaptive cruise control, and automated parallel parking. Ford Sync will be a challenge initially but can be mastered if you do one important thing: read the manual.

The Ford Fusion Hybrid’s overall excellence earns our ExtremeTech Editors’ Choice award, for the car category of midsize hybrids (one size bigger than the Toyota Prius). Feel free to disagree in the comments section. In response to comments not yet posted: Yes, we know you won’t get the 47-47-47 mpg EPA claimed by Ford but 41 mpg isn’t bad … yes the list price is higher than the competition but dealer prices are competitive … and you really can make Sync work.

Democratizing technology with affordable options

Ford has the industry’s widest array of affordable driver assist technology. Cool tech doesn’t do most people much good when it’s on a $50,000 car. The most impressive Ford Fusion tech is adaptive cruise control (ACC) that has been $2000-$2500 on high-end cars. On the Fusion and other Fords, ACC runs $1000. Set the maximum speed as with traditional cruise control, and then adaptive cruise control paces the car in front, slowing and accelerating as needed. To make it affordable, Ford’s ACC only works above 20 mph where stop-and-go adaptive cruise control (also: full-range ACC) goes down to a full stop and then back up to speed.

Do you need ACC? No. Does it help on long trips? Absolutely, especially in heavy traffic and late in the day when your attention isn’t 100%. If you can’t stand the car driving for you, ACC doesn’t cost you anything if you take a pass, and still it’s available for other buyers. Live and let live. ACC is rare on compact and midsize cars.

The optical ACC system on the Subaru Forester, called EyeSight, has the potential to drive the cost down to $500 and the same device can do pedestrian detection and lane departure warning. (The best-car-you’ll ever-drive, the Mercedes-Benz S-Class, uses both optical and radar sensors for adaptive cruise, pedestrian detection, city safety braking, and knocking down low-flying drones.) The smaller Toyota Prius offers ACC in part because some buyers had the ability and desire to spend well into the thirties on a small car that suited their lifestyles.

The Fusion also offers lane departure warning (LDW) and blind spot detection (BSD). Ford calls it Blind Spot Information System, as in ignorance is BLIS for drivers who don’t turn their heads to check traffic behind and to the sides. The same sensors used for BLIS also provide rear cross-traffic alerts (CTA) that warn of traffic coming from the side as you back out of a parking spot.

Ford’s last cool tech offering is active parking assist. Drive slowly along a street and the car finds open spaces. You do the simple stuff — put the car in reverse, apply throttle and brakes — while the car does the tough part of steering into the space.

Hybrid engine with few compromises for a sedan

“Where’d my trunk space go?” That’s the question a new hybrid sedan owner asks when seeing the big battery pack up against the rear seatback, stealing as much half the available space. By switching to lithium-ion batteries, there’s less trunk space lost (four cubic feet out of 16) and the rear seats fold flat, allowing a pass-through so you can take the 55-inch flat panel TV home with you instead of paying for delivery. What’s left is still awkwardly shaped, which is typical in all hybrids except SUVs that raise the entire load floor an inch or two.

Most hybrid buyers are looking for high mpg first, performance second. No surprise. For its second-generation Fusion Hybrid, Ford tweaked the drivetrain with a new Atkinson-cycle engine that extracts more power from gasoline and used a new constantly variable transmission (CVT). Engine and electric motor generate 188 hp combined, powering the front wheels.

Will you hit the EPA rating of 47 mpg? Probably not

The EPA rates the 2013 Fusion Hybrid at an unbelievable 47 mpg city, highway, and combined, making for a best-in-class (midsize hybrid sedans) rating. The outgoing 2012 Fusion Hybrid was at 41/36/39. “Unbelievable” turned out to be the case for the 192-inch, 3,615-pound four-door sedan when testers found they got economy around 40 mpg.

In a week’s driving I got 41 mpg. The 13% (6 mpg) shortfall isn’t uncommon for a hybrid engine, which doesn’t appreciate being pushed. I was able to reach or exceed 47 mpg for short periods, but in urban driving that meant a long line of unhappy motorists behind me as I hypermiled away from traffic lights.

The smaller Ford C-Max crossover hybrid with a similar hybrid drivetrain will be restating its EPA rating downward from 47-47-47 to 45 mpg city, 40 mpg highway, 43 mpg combined and Ford will rebate $550 as a make-good, $325 if you leased. How did this happen? Automakers run tests and certify mileage figures using EPA procedures. They can certify similar cars with a single test. But the EPA can retest if, say, owners complain. They did, the EPA retested the C-Max, and Ford lowered the ratings for the C-Max.

For now, at least, the original 47-47-47 rating stands for the Fusion Hybrid, even if the restated C-Max numbers more in line with what I and other testers got for the Fusion Hybrid as well.Consumer Reports said it got 39 mpg with the Fusion Hybrid. Before the C-Max make-good, Ford ran a customer satisfaction campaign — no, that’s not a recall — with a Fusion Hybrid software update initiated in July. It reprograms the EV mode to allow electric-only drive up to 82 mph, adjusts cooling fan and AC compressor timing, and tweaks active grille shutter closing times. The grille shutters divert air away from the engine at speed.

If you drive 12,000 miles a year, you’d burn 255 gallons of gasoline if you got Ford’s claimed 47 mpg, 300 gallons a year if you get a more realistic 40 mpg. If the car you leave for the 40 mpg Fusion Hybrid got 30 mpg, you’d have burned 400 gallons, or an extra $360 a year at current prices. The gas engine Ford Fusions get 31-37 mpg combined mpg, depending on engine, transmission, and front- or all-wheel-drive.

On the road: just like a real car

Driving the current Ford Fusion Hybrid is like the gas-engine Ford Fusion. Other than a small badge, it doesn’t look different from other Fusions. Acceleration is decent, at about nine seconds to 60 mph. You can reach 62 mph (100 kph) under electric power alone, (soon to be 82 mph), but you won’t do it in nine seconds. The car has all the power you need for city and highway driving, just not enough for comfortable two-lane passing. The brakes were a bit sensitive; some hybrids have touchy brakes and the moment you touch them, they aggressively invoke the motor-generator resistance to refill the battery. You’ll get used to it, as well as angling across parking lot and driveway entrances if they’re sloped; otherwise you risk scraping the air dam under the bumper.

The cabin is roomy in both front and back; the current Fusion gained a lot in looks while the previous Fusion was a bit bigger inside. The sloping roofline and rear pillars reduce vision, all the more reason you’ll want the blind spot detection option.

All the engine room accessories are electrical rather than belt-driven — accessories like the water pump in the engine cooling system that you don’t care about, and the air conditioning compressor for passenger cooling that you do care about. When the car stops at a light or in traffic, the gasoline engine can suspend and the AC keeps running, all to make the car seem as normal as possible.

Ford Sync and other dashboard quirks

Ford was early to integrate USB and support for digital media, Bluetooth, and voice control in 2008 with Ford Sync and then with MyFord Touch for center stack touchscreen control of phone, infotainment, HVAC, and navigation. The Ford Sync interface is accessible to developers and Sync supports center stack control of streaming media such as Pandora. If you connect a cellphone via Bluetooth and you’re in an accident, the car calls for help without the need for on-board telematics at $10-$30 a month.

In its current form, Sync and MyFord Touch remain challenging to use without a couple hours of study or instruction at the dealership. But you can master Sync. And Ford can and probably will continue to upgrade the touchscreen interface based on user feedback and JD Power satisfaction studies.

Other Ford interface design choices are more challenging than Sync. Many of the center stack controls are capacitive touch plus one big round dial in the middle. Capacitive touch looks cool in the showroom but out on bumpy pavement, it’s hard to hit the control you want dead center, the same problem as with the touchscreen. My technique was to brace my wrist on the gearshift and a thumb on frame of the center stack, and then rotate the index finger into place over the appropriate switch. Sometimes it worked, sometimes not, and sometimes I felt hot flashes about two minutes later. Since that’s unlikely in males of any age, I had realized that I once again had brushed the seat heater button in the lower corner of the center stack touch panel.

Ford’s current instrument panel design has an analog speedometer in the middle flanked by two 4-inch LCD displays that can show phone, audio, navigation, trip computer, efficiency, or driver assistance information. Even if the passenger takes over the 8-inch center stack display for audio, the driver still has a small navigation screen in his or her line of sight.

The driver assist indicators are on the left side of the left LCD screen, but unlike the Ford-supplied photo (above) that shows a pair of fully visible LCDs, from the driver’s line of sight the steering wheel obscures information about lane departure warning and adaptive cruise control, including whether ACC is on (don’t worry, be happy that the microprocessors and radars are pacing the car ahead) or off (full attention required).

Should you buy a 2014 Fusion Hybrid?

There are plenty of capable midsize hybrids, especially the Hyundai Sonata and Toyota Camry, and other good gas-engine sedans including the Nissan Altima, Honda Accord (a current ExtremeTech Top Ten car despite quirky tech), Kia Optima, and Chevrolet Malibu. Ford does the best job of making a hybrid seem mainstream and offering driver-assistance features. Fuel economy is excellent even if it’s not what Ford’s EPA numbers say. (The Sonata Hybrid was about 5 mpg less for me than the Fusion Hybrid.) Unlike the Prius with a distinctive shape, this is more of a stealth hybrid.

Ford also has gas-engine versions of the Ford Fusion for about $5000 less (with less standard equipment such as Ford Sync and LCD displays) and now a plug-in hybrid, the Ford Fusion Energi, which sells for $39,000. Ford has gotten mixed feedback on the turbocharged (EcoBoost) four-cylinder engines in the Fusion relating to fuel economy and acceleration, especially the entry 1.6-liter engine. The entry Ford Fusion Hybrid SE starts at $28,000; the upscale Fusion Hybrid Titanium is about $33,000.

Options you might consider are adaptive cruise control ($1,000), active parking assist ($900), and lane departure warning and blind spot detection. The Fusion Hybrid with help tech will be the mid-to-high $30,000 range (list price) but Ford seems to discount the hybrid enough to make it competitive.

If you’re set on a hybrid, the Fusion is our Editors’ Choice among midsize sedans. But should you buy one? If you’re not a high mileage driver, look at gas engine cars, and be sure to calculate whether you’ll earn back the hybrid price premium.

As for any model year questions: Keep in mind, because the Fusion was newly designed for 2013, changes for 2014 are modest, though the gasoline Fusion will get an even smaller, 1.5 liter, EcoBoost engine. The pre-2013 Ford Fusion Hybrid is a different car with less in the way of fuel economy and looks. The 2013 and 2014 Ford Fusion Hybrids are the best on the market. Nothing else this size gets 40 mpg city and highway.

Tesla Model S gets best-ever crash score, but that might mean tests are outdated

Tesla claims its Tesla Model S is the industry’s safest car by virtue of Tesla’s deconstruction of government crash results that give Tesla a 5.4 rating on a scale of 1 to 5. Is Tesla blowing its own horn loudly in a public place or is the electric vehicle pioneer suggesting we need better crash tests? Perhaps both. Three years ago, the government revamped test procedures because the results looked like report cards at Harvard where mostly everyone got an A. It may be time for another rework.

Tesla this week crowed about getting 5-star results across the board on all crash tests run by NHTSA, the Department of Transportation’s National Highway Traffic Safety Administration. In a release, Tesla took information provided to manufacturers on the overall vehicle safety score (VSS), reworked the data and presto, “the Model S achieved a new combined record of 5.4 stars.” Once a car meets the 5-star standard for the highest rating in a particular crash test, NHTSA gives no extra credit for how far an overachiever goes beyond the 5-star threshold.

Here’s an example. The Tesla S and Volvo S60 both got 5-star ratings in all categories. In the side pole intrusion test, Tesla says, “the Model S preserved 63.5 percent of driver residual space vs. 7.8 percent for the Volvo.” Both merit 5 stars, but which car would you rather ride in? Watching the frontal impact test video vs. other crash tests I’ve seen on video and in person, the Model S seems (from the video, which is not scientific proof) to do an exceptional job limiting most of the impact crumpling to the wheel and forward, providing a solid margin of safety for occupants.

A rollover test measures the likelihood of a vehicle rolling over (the bane of tall SUVs) and the roof crushing its occupants. Tesla aced that one, too. But wait, there’s more. During validation testing that Tesla ordered up at an independent lab, the Model S wouldn’t roll using traditional test methods, and in a test of roof strength, “the testing machine failed at just above 4 g’s [meaning] that at least four additional fully loaded Model S vehicles could be placed on top of an owner’s car without the roof caving in.”

However, Tesla notes that it’s possible to game the test by strengthening a car in the locations the machines test. A more neutral explanation would be that automakers make sure the car is engineered to pass the tests; over the years the fear of not doing well on test results has led to safer cars. NHTSA evolved from a head-on front crash test to an offset crash test to a marginal offset (very little overlap) crash test, automakers re-engineer crumple zones and even wheel-and-tire placement to do well.

Tesla determined it would meet the NHTSA front crash test with a 5-star rating and “Tesla then analyzed the Model S to determine the weakest points in the car and retested at those locations until the car achieved 5 stars no matter how the test equipment was configured.” In other words, Tesla expects to perform at a 5-star level no matter how much or little offset in the actual crash. Or better.

Clarence Ditlow, director of the safety advocacy Center for Auto Safety, analyzed NHTSA’s unreleased data for the Los Angeles Times and found the Tesla Model S had the best safety rating of any car, 0.42 (lower is better), with the Chevrolet Camaro next at 0.47. Ditlow told the LA Times that other cars have performed better on individual tests, but the Tesla has the best combined score.

Are EVs inherently safer because of their design?

Tesla suggests electric vehicles have a design advantage that can make them safer. A Tesla-generated chart (above) shows how the Model S performs on NCAP, the New Car Assessment (crash test) Program: at the top of the charts. The red line is the relative risk score where lower is better; the boxes 3-4-5 are star ratings and higher is better. With an EV, the battery is typically mounted low in the car (sometimes it forms the floor), yielding a low center of gravity and excellent resistance to tipping. With the electric motor in back (the front is trunk space), that gives Tesla even more freedom to create crash protection crumple zones. The mass of a front-mounted engine doesn’t protect in a crash; it’s more of a hazard for being driven into the passenger compartment.

Time for new safety tests, again?

Current NHTSA tests are the moderate overlap frontal test, side impact, rollover, and rear tests, plus the new small overlap frontal test. NHTSA already beefed up its standards starting with the 2011 models. Why? Pre-2011, as automakers improved, many cars got 4 or 5 stars. The 2011 tests provided some separation between good and very good. Tesla, since it scores so well, sees enlightened self-interest in additional reporting, perhaps with decimal results rather than 1-2-3-4-5 results, or possibly even stiffer tests.

How to read and understand the test results

US crash results dating to 1990 are on the NHTSA Safer Car site. Others interpret the results and attempt to roll the ratings into a one-phrase summary. The Insurance Institute for Highway Safety rates the better cars as an IIHS Top Safety Pick if crash scores are the best currently possible on the four established tests: moderate overlap frontal test, side impact, rollover, and rear tests. The TSP+ rating adds in the new small overlap front test, the one a lot of automakers are struggling to meet, and requires four or five of the highest ratings; one score can be lower. Most likely the lower score will be on small offset front crash.

If you’re shopping for new or used cars, the NHTSA tests are helpful. NHTSA says you can compare across car categories in the breakout tests except in the frontal impact test, because it measures the effect of running into another car of the same size, both going 35 mph. IIHS says not to compare its TSP and TSP+ scores between categories (don’t compare a TSP in full-size cars with a TSP in compact cars) because the IIHS rollup includes the front impact result. Ironically, one car not listed on the IIHS Top Safety Pick page this week is the Tesla Model S.

When shopping for a good used car, look at the NHTSA tests or something like the IIHS TSP rollup scores. Consumer Reports has good reliability ratings, although the difference between best and worst has compressed. Csaba Csere, erstwhile editor of Car and Driver and an MIT engineer, says the most important feature to look for on a used car is electronic stability control. Seat belts and airbags protect you once you’re in an accident; stability control helps prevent loss of control that leads to accidents. Anti-lock brakes are important, too, but they’re almost universal on recent model cars; not yet stability control.