I’ve had my iPhone 4S for a few weeks now, and have gotten to know it pretty well. Mine is the 64GB model, mostly because I want to play around with the new 1080p video recording capabilities, and don’t want to worry about running out of storage space all the time.

I’ve read many reports online about runaway battery consumption with the new 4S, but I haven’t noticed any huge change vs. my iPhone 4 in battery behavior. I still charge it every night and most of the time when I’m at my desk in my office.

From a hardware perspective, the iPhone 4S pretty much delivers as expected. The dual-core A5 processor zips along with plenty of headroom, smoothing out many of the rough spots in everyday usage that have started to creep in since the iOS 4.3 update. The camera upgrade boosts both still image and video (1080p) performance into the realm of most modern point-and-shoot cameras, although the lack of any optical zoom is still a big limitation. On the other hand, the wide availability of camera and photo enhancement apps pretty much make up for the lack of optical zoom. Battery capacity has increased minimally, but not enough to make any practical difference.

From the power efficiency standpoint, by far the most interesting new hardware feature is the new dual-antenna design, which eliminates the infamous “death grip” effect of the iPhone 4 and improves cellular reception in general. Combined with the new communications chip that boosts data rates (for GSM networks), the new radio setup is worth looking at—especially since we’ve seen that the radio can contribute even more to the energy equation than the display.

The runaway star on the software side with the iPhone 4S and iOS 5.0 is Siri. Reminiscent of a cross between HaL and the famous Star Trek “Computer,” Siri listens, seems to think, and generally does a better-than-expected job of carrying out your wishes. While still clearly early in the development cycle, Siri feels to me like the beginning of a paradigm shift where we may actually become just a productive without a keyboard as with one. I’ve dictated quite a few e-mail and text messages with Siri (sometimes while driving!), and accuracy is very good—or very bad. That to me is an indication of evolving and improving AI on the recognition side. And as is usual for Apple, the real genius of the Siri interface is the simplest part of it: You simply hold your phone up to your head to start talking to Siri. To everyone else, it just looks like you’re answering a phone call! I wish I thought of that…

Running the 4S though my usual battery (pun intended) of power tests running the Star Trek movie resulted in the following:

The results were surprising in several respects. First, the 4S seemed to be extremely efficient in the “Max Battery” mode. It played through the entire movie consuming just 0.6 Wh of energy. That would be more than 18 hours of continuous movie playing, although you can’t see much at the lowest brightness setting. This is almost a 40% improvement in energy efficiency vs. the iPhone 4! The new lower power A5 chip is likely at the heart of this result. These days video decoding is an almost-routine task, and can probably easily be handled on one of its cores.

Turning the display to full brightness (the “Max Brightness” test) shows the expected result. The energy cost of running the display at full brightness is about 0.6 Wh for the two-hour movie. The display on the 4S isn’t notably different than on the iPhone 4, so this is expected.

Turning up the sound to maximum (“Max Movie” mode) shows one other interesting change. On the iPhone 4, there was virtually no change in energy consumption between the runs with the sound muted or with the sound at maximum. On the iPhone 4S there is definitely a measurable difference, both in energy consumption as well as in the perceived loudness of the sound. In fact, with the sound at maximum, the tiny speakers in the 4S produced enough sound to make it annoyingly loud as I was trying to do some other work. I had to resort to my “manual” muting method (putting a piece of tape over the speaker) to conduct my tests.

Well, I’m about out of space and time for this post. Next time I’ll describe the results of testing the new iPhone 4S dual-antenna and modem chip setup. These results are very interesting.

In late breaking news, Apple just announced a software update, iOS 5.0.1, which among other things “fixes bugs affecting battery life.” No big surprise, as the complexity of today’s smartphones rivals any other computing platform—and dwarfs the others when it comes to power management. In particular, the interaction between hardware and software to minimize energy consumption is very difficult to model and predict, but the ramifications on battery life are immediate and sometimes ugly.

And by the way, if you are interested in learning more about low power hardware design, my colleague Josefina Hobbs is hosting a new series of short videos covering everything from introductory concepts to selected advanced low power design topics. Check them out here.

The iPhone 4S was announced today and frankly I was a little disappointed. No new iPhone 5, no 4G support, no bigger display or smaller form factor. But now that the initial disappointment has worn off, let’s take a look at what we DID get—pretty much the rumored iPhone 5 in an iPhone 4 package. There’s a faster processor with the A5, higher-resolution eight-megapixel camera for reasonable stills and 1080p video, a “fat” 64 GB version, wireless mirroring to HDTV displays via Apple TV, a world phone (GSM and CDMA support), and a whole host of new software features including expanded voice control, many new iOS 5 features, and iCloud to support wireless synchronization and cloud storage.

While I don’t have one to play around with yet, my first interest from a power standpoint will be to evaluate the impact of the A5 processor. We saw a measurable improvement in energy efficiency in the iPad 2 vs. the original iPad, partly attributable to the A5. Roughly doubling overall performance compared to the A4, the A5 is a dual-core processor and adds significant additional power-saving features to do more with less.

Even more interesting, though, will be an evaluation of a less-talked-about new hardware feature: A revamped antenna setup for the iPhone 5 is said to significantly improve reception and data speeds. That should mean fewer dropped calls. We’ll see about that because it won’t be hard to verify this claim. And if data reception is improved by any noticeable amount, I’m certain we’ll be able to see it in our standard “Star Trek streaming test.” With the Retina display remaining constant, we should be able to get some good comparison points on the new hardware.

And here’s the big secret in Apple’s strategy with the iPhone 4S—I expect that Apple TV unit sales will go through the roof. Wireless streaming to today’s huge HDTV displays for $99 is only a little more than double the cost of Apple’s $39 HDMI adapter (Apple Digital AV Adapter), which doesn’t even include the cable! Sign me up.

So no new iPhone 5, but plenty of new features to fiddle around with until the “5” arrives. See you in line!

We’ve been examining power efficiency of iOS devices for a while now, and it’s hard not to notice the relative trajectories of mobile operating systems and more traditional PC operating systems. With the recent release of OS X Lion, Apple is moving in the direction of converging capabilities of these platforms, with the clear goal of a more unified environment coming down the line.

When I first played with the iPad and iPad2, I thought I had purchased my last laptop computer. The portability and battery life of these tablets were so compelling that surely the days of the laptop were over, and it was only a matter of time before tablets ruled the portable computing category. And sure enough, in the last couple of years, tablets have multiplied faster than rabbits. But interestingly, none have taken significant share away from the runaway success of the iPad.

Then in late 2010 came the refresh of the MacBook Air, transforming an over-priced, under-powered specialty gadget into a mainstream computing device that has breathed new life into the entire laptop category. Sure, it was still at least one generation behind in raw compute power, but as we all know by now, it’s the combination of compute power with all of the other system parameters that determines utility today, and the 2010 MBA hit the center of the target. And to top it all off, less than one year later, the July 2011 MacBook Air refresh brings latest technology to this form factor, completing the repositioning of the MBA from a “snob’s machine” to one that can satisfy 80% of the market. I got my 13” MBA a few weeks ago, and have been impressed not only with its speed (1.7GHz Core i7 with 3MB cache, 4GB memory, and 128GB flash disk), but also its power efficiency (about five to six hours of typical use.) On the “Star Trek” power efficiency test, the 13” MBA fared very well—It made it through the entire 2:06 movie at maximum brightness and consumed about 22.5 Wh of energy, nearly the same as my 2010 11” MBA at 21.35 Wh.

My biggest dilemma now is which device(s) to bring with me? My arsenal now includes the iPhone 4, iPad 2, 11” 2010 MacBook Air, and 13” 2011 MacBook Air. All are very compelling, but a few factors make the determination easy, at least for now. First, the iPhone 4 is in. It’s the one device that I ALWAYS carry with me. Small, light, and utilitarian, it’s the 21st century Swiss army knife. The dilemma is, which additional device makes the cut? I’ve already decided that Internet access is best achieved through my phone via the personal hotspot feature, so that’s a wash. The iPad 2 is a wonderful machine. It still tops the list for the most compelling portable movie-watching device. Compared with all of the other devices, the display is big, crisp and clear, with deep rich colors, and exhibits the fewest artifacts. It’s a winner, but unfortunately iOS apps still restrict serious usage for entering or editing the standard documents that we all need to access, namely those created in Microsoft Office. Sure, it’s possible to upload/convert to Google Docs and edit online, but access isn’t 100%, and the interface is still a kludge in iOS, at best. I’m convinced that storage and editing in the cloud is the way of the future, but unfortunately, we’re stuck here in the present for now.

Which leaves me with two great choices for larger-format portable computing, the 2010 11” Air, or the new 2011 13” Air. The 2011 13” Air is a fantastic machine—blazingly fast and extremely power efficient. The display is significantly bigger than the 11” version, but so is its form factor. I’ve decided that my ultimate road warrior combo is my iPhone 4 coupled with the 11” MacBook Air. Everything I need, and super portable!

From the software standpoint, being able to run standard apps is certainly a compelling feature. Note to Microsoft: Where is that iOS Office app? Or maybe the operating systems are converging even faster. I’ve tapped the screen of my Air many times in the last few months, expecting a document to open, or a Web link to be followed. Certainly there are prototypes of touchscreen laptops deep in the research facilities at Apple. Plus, the large track pad and additional interface features in Lion (MacOS 10.7) are starting to make this laptop feel a lot like a tablet!

With this many great choices out there today, I can tell that my ultimate combo probably won’t last long. A converged iMacOS or big jump in performance and apps might well nudge the iPad (3?) back into the lead. But one thing’s for sure: I haven’t purchased my last laptop!

With the current popularity of all things extreme, from extreme dieting, extreme couponing and extreme hoarding, all the way to extreme sports and even extreme programming, I thought, “Why not Extreme Power Efficiency?” After all, power efficiency has been improving at a blistering pace for the last few years. Where will the hotspots and power bottlenecks be looking into the future?

Well, let’s start by rounding up the usual suspects. Dynamic and static power are the buckets into which we partition the energy that is used for computing (flipping bits), vs. the energy used to maintain power to the circuitry (sometimes also called standby power). For dynamic power, much of the focus today is on the back-end of the implementation flow—making sure that capacitances are minimized, dealing with many voltage areas or “islands,” and allowing dynamic variation of voltages and clock frequencies to conserve power. These problems aren’t completely solved today, and continue to expand as power architecture complexity increases, but they are reasonably well understood, with lots of people working on improvements in tools and methodologies. “Extreme” dynamic power efficiency might instead be measured in units of “transitions per function” to gauge the transition-efficiency of any implementation, combined with “joules per transition” for the physical layout and technology efficiency, to arrive at energy consumption estimates. As with any process, you can’t improve what you can’t measure, so thinking about measurements and metrics isn’t a bad place to start.

For static power, we are now pretty good at power gating or “shutdown” to minimize leakage power in unused blocks, and new technology improvements have at least postponed the dreaded explosion in leakage at smaller geometries. However, these problems won’t go away, so as we move forward, “extreme” thinking dictates that power gating will continue to become finer-grained—and to a certain extent the current move toward “3D” transistors is a move in this direction—with much better on-off characteristics such as faster performance and lower leakage. So as the technology enables new transistor designs that approach the “perfect switch,” the tradeoff between finer-grained power-gating vs. more efficient technologies continues to shift.

Finally, while it seems there’s an endless list of things that we need to (and can) worry about, remember that part of what we do everyday is to make practical decisions about priorities. Power efficiency is no different. Worrying about power consumption for one transistor may not seem like much, but multiply it by 3 billion transistors on a chip and suddenly you’re talking real power. At a macro level, an average no-load power (that means no phone on the other end) of 0.1W (0.5W just 3 years ago!) for a cell phone charger isn’t much, but multiplied by the 5 billion mobile phones in the world and 24/7, and you can see we’ve got a big problem. Extreme thinking doesn’t always point us to practical problems that need addressing immediately, but it does allow us to step “outside the box” for a bit just to see what might be out there.

I have an opportunity this month to do some tests using my iPhone (AT&T iPhone 4, 32G, iOS 4.3.3, Personal Hotspot) as a true business productivity tool. I’m travelling to Austin, and am hosting a two-hour WebEx video conference call through our WebEx/MeetingPlace infrastructure. I’ve decided to use my iPhone 4 as the wifi connection from my laptop for the WebEx session, as well as for the audio portion of the conference call in MeetingPlace. Supporting hardware is my HP EliteBook 2540p laptop, which is connected via VPN to our company infrastructure through my iPhone 4 personal hotspot.

The first thing to note (if you haven’t seen the TV commercials) is that this setup isn’t even possible on the Verizon iPhone 4 because it requires voice and data transmission simultaneously. So if it works, chalk one up for AT&T.

Setting up the hotspot is very easy. This is a regular tool in my arsenal now, usually used for connecting iPads and iPod Touches to the Internet. Interestingly, I’ve also found around the house that my other family members, who are on more data-limited plans, also like to use my personal hotspot to save their data allocation. The personal hotspot feature from AT&T doubles your data allocation to 4GB per month, so I’ve never had a problem with the data limit on this plan.

With my laptop connected to the Internet via WiFi, the VPN connection into the company is a no problem, and I’m ready to try things out. Starting up the WebEx session goes surprisingly smoothly; I’m worried that latency might be an issue, but no problems. Now the next trick: start up the conference call, as well. Our latest software update in WebEx allows the system to call each participant, including the host, so I enter my phone number, and voila! My phone rings, and I start the conference call. During the next 15 minutes, about 15 to 20 participants join the call and the WebEx, and we’re off.

In order to get as much data as possible during the call, I am recording the time, battery percentage, and signal strength, and have decided to run the first half of the meeting with my iPhone display ON, and the second half with the display OFF (except for occasional interrupts for me to record the data).

Running a full WebEx videoconference with a voice conference in parallel resulted in the following:

– Power consumption during first hour with display ON: 1.87W

– Power consumption during second hour with display OFF: 1.38W

Our first obvious conclusion is that the display seems to draw about 490mW of power at full brightness. This is interesting because in our previous tests (watching Star Trek), the difference in power between the display on full brightness vs. minimum brightness was about 300mW. So now we can approximate the missing piece of information about the iPhone 4 display. It seems to draw around 190mW from powered down to minimum brightness, and an additional 300mW for full brightness.

In addition, the data also continues to fill in our chart of power consumption as a function of signal strength. Previous tests showed the hotspot feature drawing about 920mW when used to stream Star Trek with two to three bars of signal. In this case we can approximate the power for streaming by taking the total power required with display off (1.38W), and subtracting the estimated power for the voice call (750mW, given Apple’s seven-hour talk time estimate), leaving about 630 mW for the personal hotspot with 4 bars of signal strength.

OK, I know this calculation is a stretch. I really have only one data point. The call power is purely a guess from the datasheet, and the volume of data for the WebEx conference vs. the Star Trek movie isn’t controlled (I forgot to reset the data monitor on my phone!). But it’s fun to see that 1) completely mobile video/audio conferencing is a reality, 2) it can all be done on a single smart phone, and 3) the order of magnitude of energy usage is very similar to streaming a movie. Point No. 3 reinforces our previous notion that power dissipation on today’s mobile devices is dominated by communications (radios and antennas), as opposed to computations (gigahertz or gigabytes). And that’s very interesting.

Last week was the annual pilgrimage of hardware designers to the Design Automation Conference, where the latest in tools and technologies are displayed for use in upcoming generations of computer and electronic devices. And it was no surprise that low-power design continues to expand across all facets of the design space, from LED lighting to smart phones and cloud computing.

For me, one of the memorable moments was the mention of a new acronym in Alan Gibbons’ tutorial “System-Level Design and Software Development for Energy Efficient Platforms: Challenges from Models to Methods” – RFTS stands for “Run Fast, Then Stop.” Could this be a viable new strategy in our unending quest for energy-efficient hardware design? Haven’t we shown that voltage reduction is the real winner, resulting in a squared reduction in power, and even more as frequency is reduced? Won’t the inevitable leakage current in idle mode swamp any potential gains, anyway?

The answers, as expected for any complex question, are “maybe,” “kinda,” and “sorta.” While by now we’re all used to clock gating to reduce dynamic power and multi-threshold transistors to control leakage, and even voltage reduction, power shutdown, and dynamic voltage/frequency scaling, the demand for more techniques to minimize energy consumption continues to grow. Once we’ve lowered voltages to the point where noise (and timing) margins are a concern, and we’re shutting down everything possible, what’s next? RFTS is one answer to the question of whether it’s more efficient to work slowly and constantly, or finish the job as fast as possible and take a break. Clearly the general answer depends on the effectiveness of any shutdown strategy, the level of speed control, and some knowledge of the input stream of work to be done.

I was struck by the similarity of these questions to the questions we’ve raised regarding iPhone power efficiency. When we tested the AT&T iPhone vs. the Verizon iPhone for video streaming a few months ago, the results seemed to come down in favor of coverage (and a stronger signal), but in areas of similarly weak and variable signal, having a faster network was a clear advantage. Receive fast, then stop – interesting!

Optimizing for energy efficiency continues to get more complicated, requiring better and higher level analysis to make the right decisions, and extending the challenges to the design community. That’s a good reason to return to DAC next year.

It’s been an exciting week for designers of low-power electronics, as well as consumers of mobile devices. Intel announced on May 4 that its Tri-Gate 3D transistor technology is ready for production on its 22nm process. The “Ivy Bridge” processors are being demo’ed now and scheduled for full production this year. Products containing Ivy Bridge processors will ship in 2012.

Analysts have been salivating at this new technology announcement, with much written about how this technology unlocks Moore’s Law for the future and bolsters Intel’s competitive position vs. ARM for low-power processors. But there’s been some inconsistency and confusion in the press about the technical claims for this technology, so I thought it would be fun to take a closer look.

Intel’s press release states: “The 22nm 3-D Tri-Gate transistors provide up to 37% performance increase at low voltage versus Intel’s 32nm planar transistors. This incredible gain means that they are ideal for use in small handheld devices, which operate using less energy to “switch” back and forth. Alternatively, the new transistors consume less than half the power when at the same performance as 2-D planar transistors on 32nm chips.”

Sentence No. 1: A 37% performance increase for low voltage applications (mobile devices) sounds pretty impressive, but that’s pretty much in line with what I’d expect in moving the technology node from 32nm to 22nm.

Sentence No. 2: This one’s just in there for marketing purposes, although the grammar implies that the “handheld devices” are switching “back and forth.” That’s more than a little confusing, even for an engineer—or maybe especially for an engineer.

Sentence No. 3: The key word here is “Alternatively”. This implies that you get 37% better performance in low-voltage (low-power) mode, OR 50% less power for the same performance in high-performance (high-power) mode. “Same performance“ here isn’t well-defined, but I’m assuming that means equivalent transistor performance (delay), not operating voltage.

Several press accounts have reported 37% better performance AND 50% less power, which isn’t consistent with the press release.

So what’s the bottom line, and how big of a deal is this? First of all, we’re going into production at the 22nm node. Second, the so-called 3D technology is indeed a breakthrough because it enables volume production at 22nm without the anticipated and feared exponential rise in leakage current. Intel uses the analogy of a skyscraper in its press release. It looks to me to be more of a rooftop deck–but decks are still a big improvement. And yes, I am excited!

By now, you’ve heard about AT&T’s recently introduced feature on the iPhone 4 – along with iOS 4.3, the iPhone 4 tethering feature now allows you to create a “personal hotspot” that can be shared with up to 3 other devices. One interesting twist on the plan from AT&T: the additional $20 per month cost of the plan comes with an additional 2 GB of data, for a total of 4 GB per month.  It seems as though all of the major carriers are pretty much settling on around $10 per GB at this point, and that seems like a fair price for now. Don’t look for new unlimited data plans anytime soon – the move to 4G (or at least 3½G) has already started, and carriers are looking for clean ways to get rid of all unlimited data plans – that makes sense, since the transmission speeds will soon allow you to use MUCH more than your “fair share” of wireless bandwidth on an unlimited plan.

I signed up for the personal hotspot option when it was introduced, finally replacing a jailbroken iPhone 3GS running MiWi as my portable hotspot solution. I still haven’t convinced my nephew that this was a smart or even reasonable move. In fact, I was very happy with the performance of MiWi, although I was never totally comfortable with the idea of jailbreaking my phone, both from the standpoint of getting new software updates and keeping the software stable, as well as violating the terms of agreement with AT&T. Now I’m happy to pay my $20 per month to have it all, and sleep soundly at night (actually that’s never been a problem).

But I digress…the really fun part about having the personal hotspot on my phone is that I can now run some additional power efficiency tests, and watch Star Trek a few more times!  Recall that our previous testing running the 2 hour, 6 minute movie used about 1.1 Wh of energy on the iPhone 4, with the screen dimmed completely. Adding the screen and sound (which was negligible) pushed the total energy consumption to 1.7 Wh.  And streaming the movie via WiFi bumped the total energy to 2.3 Wh. Our 3G streaming tests were even more interesting, resulting in energy consumption of 2.8 Wh to 5.3 Wh, depending on the strength of the 3G signal as measured by the number of bars displayed on the phone.  I‘ve gotten a few comments regarding inaccuracy and instability of the numbers of bars method of measuring signal strength, but I’ve actually purposefully used these instead of the dBm numbers available in field test mode because I don’t want to imply any more accuracy than there is in the measurement.  I’ve actually found that the number of bars is reasonably consistent for a general swag, and if anything, the resolution should only be viewed in three “quanta” rather than five. So I’m using 1-bar, 3-bars-ish, and 5-bars as my signal strength measurement.  Anything finer implies more accuracy than I can measure consistently.

For my personal hotspot test, I’m streaming Star Trek through my iPhone 4 using the personal hotspot feature to connect with my iPad 2 via WiFi.  This test was done in a different location, so I first ran a test on the iPhone 4 to calibrate everything. Streaming Star Trek via 3G in my room in Orlando (on vacation) with about 3-bars-ish of coverage with full brightness and sound took about 3.15 Wh of energy, the same as when I did the same test in my house in Palo Alto! The picture started out somewhat pixilated, but the 3G connection buffer caught up after a few minutes, and the movie generally looked great (note that my previous testing on the Verizon iPhone showed that under the same conditions the pixilation continued throughout the movie, suggesting that the connection bandwidth wasn’t high enough, and it was having trouble keeping up).

Next, streaming to the iPad 2 using the personal hotspot on the iPhone 4 – and it worked as advertized!  I was able to watch the whole movie (2:06) with no problems.  Again, the picture was a little pixilated in the beginning, and degraded at one point during the test, but in general the quality on the iPad 2 was very impressive. Final energy usage: iPad 2 used 6.5 Wh of energy (brightness and sound at maximum), and the iPhone 4 consumed 1.94 Wh to act as the bridge between the 3G signal and WiFi.  Previous tests showed the iPad 2 using 6.25 Wh to stream the movie using Wifi so this was a little higher, but within a reasonable margin of error. Also, from previous tests, the iPhone 4 used 3.15 Wh to stream the movie via 3G at 3-bars-ish, and 1.7 Wh to play the movie locally – the difference of 1.45 Wh would be a good initial guess at energy to receive the data via 3G.  Throw in the added cost of broadcasting the WiFi signal, and the 1.94 Wh seems to be just about right.

Not a bad vacation, and “The Wizarding World of Harry Potter” was fun, too!

Our stable of test devices continues to expand.  After only a couple of hours in line at the Stanford Shopping Center last Friday, I have my iPad 2, complete with a funky lime-green cover! It’s the low-end 16 GB model, with wifi connectivity only. After our recent experiments on power efficiency versus communications, I figure I’d rather isolate the biggest energy leak in the system by using my iPhone 4 as the communications hub, and go for the new Personal Hotspot feature (replacing MyWi) to get mobile data to other devices. Draining batteries on multiple devices in parallel just doesn’t make any sense to me.

So our test device is a white 16GB iPad 2, complete with two cameras, gyroscope, and a new processor, and is thinner and lighter than the original.  Not bad right out of the gate.  I also got the two add-ons: a magical magnetic cover, and the new HDMI cable.

First, the cover – I thought when I bought it that $39 was a ridiculous price to pay for a one-side cover, but after trying it out, I love it!  It has a whole slew of magnets – 21 in the cover, and 10 in the iPad 2 itself (check out the “iPad 2 cover teardown” on iFixit) that all work together to allow the cover to work as a typing stand, a display stand, a screen cleaner, and (of all things) a cover.  Totally gadgety – right up my alley.  The other add-on is the HDMI cable – this is a HUGE improvement over all of the original cables and interfaces (VGA, component, HDMI, DVI, composite) that Apple and third parties offered.  Not because the cable itself is better or worse, but because the iPad 2 now supports full-time mirroring through HDMI, straight onto your big screen TV! Netflix looks great, presentations are useful, surfing the web is fun, and a little (err, BIG) “Pig Rush” certainly passes the time nicely. Thumbs up on the add-ons!

I haven’t looked much at power efficiency of the cameras or gyroscope subsystems, so let’s jump straight to the processor. The original iPad used a 1 GHz A4 processor, based on the ARM Coretex A8.  The iPad 2 utilizes the new A5 processor, built upon the ARM Coretex A9, with dual processor cores at 1 GHz, and clock speed variability to save power. There are many teardowns and reviews of the A5 processor for more gory details – one of my favorites is at AnandTech (a great site if you’re interested in hardware design and analysis by the way). Bottom line from all of the benchmarks that have been run so far is that the A5 in the iPad 2 can nearly double the performance in real apps, and possibly up to 6-8x performance for floating point intensive apps compared to the A4.  That’s quite a boost in horsepower, and the most interesting question to me is, “how much does it cost me in power efficiency?”

There is clearly plenty of performance headroom in the iPad 2.  But what does it cost in terms of energy?  A 400 hp car might accelerate twice as quickly as a 200 hp car, but in many cases it also gets about half the mpg – carrying around the extra weight of a more powerful engine is typically a drain on efficiency.  In our case, my original iPad power efficiency tests showed about 3.2 Wh of energy required to view the Star Trek movie (2:06:46) in the “maximum battery life” configuration (airplane mode, brightness & sound minimum).  Since updating the software to iOS 4.2.1, that number has dropped to 2.73 Wh, a 15% improvement! There were some rumored improvements in power performance in iOS 4.2 – I’m still checking into the details on this.  For the iPad 2 in the same configuration running iOS 4.3, the energy required is just 2.25 Wh.  Power efficiency on the iPad 2 has increased enough to more than make up for the “bigger engine”!  That’s pretty incredible.  One clue is that the A5 processor is manufactured using a 40 nm process, compared to the 45 nm process of the A4.

Similarly, in maximum movie watching mode (display and sound at maximum), the iOS 4.1 iPad required 6.2 Wh, updated to iOS 4.2 it requires 5.95 Wh, and the new iPad 2 requires just 5.75 Wh for the Star Trek movie test.  Nice improvements in power efficiency all around.

Many people have asked me if the iPad 2 outperforms the original iPad.  The answer, of course, is “not yet”.  That’s because virtually every iPad app out there is designed to run on the original iPad, and in general, software developers (the good ones anyway), have tuned performance, and even limited functionality, to suit their target platform.  Six months from now, when the latest apps require the additional performance of the iPad 2, the original will undoubtedly feel a little sluggish.

In fact, I have no complaints about performance on my iPad, so I wouldn’t expect new hardware to suddenly seem much faster. This behavior is a testament that device performance today isn’t the bottleneck – there’s plenty of horsepower to accomplish amazing things in today’s apps. What will change down the road is that apps will get even more amazing.  I’m still waiting for the “headless man” Halloween costume app, which would utilize two iPads, one in front and one behind your head, connected via FaceTime, effectively making your head “invisible”! You have to think about it a bit, but this might be the first cut at the famed Romulan Cloaking Device! 

First test – I’m going directly for the jugular.  Stream Star Trek to the Verizon iPhone 4 in my office, where I can just barely get a signal on my AT&T iPhone 4.  This is the case where I wasn’t even able to get through the movie on a full battery charge.  The AT&T iPhone 4 extrapolated to 5.3 Wh of energy to play the roughly 2 hours and 6 minutes of Star Trek, with just about 1 bar of signal strength.  On Verizon, I got an average 3+ bars of signal strength on average, and streamed the movie with no problem using 3.15 Wh of energy.  It’s looking like a landslide victory!

Driving home, I monitored the signal strength on the Verizon iPhone – it varied between 2 and 5 bars, even in my well-travelled and well-known “dead spots” where I’ve dropped many calls and data connections.  Another moral victory for Verizon.  At this point, I’m about ready to chuck my iPhone back at the AT&T rep, and jump ship.  But a little more testing is in order…

Next test is simply to rerun the Star Trek streaming experiment at my house – I chose to do it downstairs, where coverage is typically a little worse.  And here’s where it gets strange.  Both phones showed about 2 bars of coverage – the Verizon iPhone varied between 2 and 3 bars, and the AT&T iPhone fluctuated more, between 1 and 3 bars.  No big surprise yet, but the bottom was about to drop out. 

Streaming Star Trek under these conditions, the first big observation was that the picture on the Verizon iPhone was clearly pixilated – it was having trouble getting the streamed data fast enough, particularly in the fast-moving action sequences.  The AT&T iPhone showed some similar artifacts, but only occasionally.  This continued, and got somewhat worse throughout the movie – the Verizon iPhone stopped playing a dozen or so times during the movie, waiting for the data to catch up.  The final results: the Verizon iPhone consumed 4.3 Wh of energy, and the AT&T iPhone consumed 3.15 Wh!  No that’s not a typo – the winning number 3.15 Wh, or 60% of the 5.25 Wh battery capacity, was exactly the same as where we started this round of the experiments, except on the other phone!  That’s some poetic justice.

It’s now past midnight, and I still have a few more tests to run – it seems our showdown wasn’t nearly as conclusive as I expected.  One possible explanation relates back to network coverage vs. network speed.  Better coverage is a good thing, but there’s an interesting inversion in equally low coverage areas – network speed might help make up for a poor signal in data intensive applications by allowing the application to buffer more data until the signal quality improves.  This, of course, is only a late-night random theory, but an interesting one to pursue.  Things just get curiouser and curiouser.  If you have any interesting ideas, send them along, and we’ll continue the experiments together.