The Apple iPad 2 Review
by Brian Klug, Anand Lal Shimpi & Vivek Gowri on March 19, 2011 8:01 PM ESTWiFi and baseband
In our iPad 2 preview I speculated that Apple might have moved entirely to Qualcomm for the iPad 2 baseband, and the most likely guess was to the same Qualcomm MDM6x00 used in the Verizon iPhone. Turns out that I was only partially right. Looking at the difference in weight between the Verizon and AT&T models on the Apple website should've been the dead giveaway, but I initially suspected the reason was more area dedicated to power amps for the GSM/UMTS version.
Black RF window at the top, even on a white iPad.
The Verizon iPad 2 shares the same Qualcomm MDM6600 baseband as the Verizon iPhone 4. The AT&T (GSM/UMTS) iPad 2 on the other hand uses the same Infineon, ...errr Intel X-Gold 618 baseband as the GSM/UMTS iPhone 4. It's interesting to see Apple using two different basebands when one could have covered both CDMA2000/EVDO and GSM/UMTS, and faster HSPA at that rate. The X-Gold 618 just isn't super impressive anymore in a world dominated by HSDPA 14.4 and 21.1 compatible basebands. That said, it's a welcome improvement from the X-Gold 608 shared between the iPhone 3G, 3GS and GSM/UMTS iPad, which lacked HSUPA entirely, and thus had a maximum upstream rate of 384 kbps (UMTS speeds).
| GSM/UMTS (AT&T) iPad 2 - Network Support | |||||
| GSM/EDGE Support | 850 / 900 / 1800 / 1900 MHz | ||||
| UMTS/HSDPA/HSUPA Support | 850 / 900 / 1900 / 2100 MHz | ||||
| HSDPA/HSUPA Speeds | 7.2 Mbps / 5.76 Mbps | ||||
| Baseband Hardware | Infineon/Intel X-GOLD 618 | ||||
| CDMA (Verizon) iPad 2 - Network Support | |||||
| CDMA2000-1xEV-DO | 800 / 1900 MHz | ||||
| Theoretical Downstream/Upstream | 3.1 Mbps / 1.8 Mbps | ||||
| Baseband Hardware | Qualcomm MDM6600 | ||||
The cellular enabled versions of the iPad 2 still are easily picked out by the black plastic strip at the top which is an RF window. Curiously, the color of the RF window is black even on the white iPad models, which seems a bit un-Apple. Maybe black made more sense since white plastic might eventually show some discoloration. Through testing the window seems to also be used for WiFi in addition to cellular. The GSM/UMTS version has a microSIM slot and comes with a SIM ejector tool in the package.
I found the microSIM incredibly difficult to eject on my AT&T iPad 2 compared to every other iDevice. You have to insert the ejector tool normal to the curved surface, which ends up being at an angle of nearly 45 degrees with respect to the back surface. It isn't how I expected things to work, and the tray itself is a bit awkward.
I carried around the AT&T version of the iPad 2 and ran over 150 speedtests using Ookla's speedtest.net application. I averaged throughput of 2.76 Mbps down and 802 kbps up, with latency of 390 ms. Anand carried around the Verizon version and saw downstream speeds averaging 872 kbps down, 434 kbps up, and latency of 382 ms.
| 3G Speed Comparison - Speedtest.net to best host | ||||||
| Provider / Data | Verizon - (EVDO) | AT&T - (UMTS/HSPA) | ||||
| Max Downstream | 1.679 Mbps | 6.135 Mbps | ||||
| Average Downstream | 0.872 Mbps | 2.759 Mbps | ||||
| Max Upstream | 0.837 Mbps | 1.635 Mbps | ||||
| Average Upstream | 0.434 Mbps | 0.802 Mbps | ||||
| Average Latency | 382 ms | 390 ms | ||||
| Minimum Latency | 298 ms | 239 ms | ||||
Pricing between the two providers is interesting. It appears that Verizon was careful to not compete with AT&T at the same data package points, instead targeting data packages other than the two 256 MB and 2 GB bundles AT&T already offers. At the same time, pricing at 3 GB on Verizon is exactly where you'd wind up should you run over 2 GB on AT&T and incur the automatic $10.00 overage charge. Interestingly enough, whereas AT&T bills you automatically when you run over your data caps, Verizon simply stops and lets you add more data before the billing cycle ends. It's an interesting distinction, certainly competition is good. In addition, there doesn't appear to be any activation charge for either AT&T or Verizon.
| 3G Enabled USAiPad 2 Data Plan Pricing | ||||||
| Provider | Verizon (CDMA2000/EVDO) | AT&T (GSM/UMTS) | ||||
| 250 MB | NA | $14.99 / mo - $14.99 for each additional 250 MB | ||||
| 1 GB | $20.00 / mo | NA | ||||
| 2 GB | NA | $25.00 / mo -$10 for each additional GB | ||||
| 3 GB | $35.00 / mo | NA | ||||
| 5 GB | $50.00 / mo | NA | ||||
| 10 GB | $80.00 / mo | NA | ||||
So which of the two is the one to get in the US? Ultimately the decision about which is "better" is really a function of local coverage profile, speeds, and how much data you use (which will determine your pricing). If you're an iPad 1 AT&T user and used unlimited, you also have the option of grandfathering it in for some time as well. Unless you’re in an overloaded metropolitan area, AT&T 3G data speeds are undeniably faster. At the same time, Verizon has more EVDO coverage that, while slower, is generally more consistent. Verizon’s pricing also includes heavier data numbers than AT&T does, so if you’re looking for something to completely replace a notebook with and intend to not use WiFi, that’s a fairly big draw.
Speaking of data use, I find the fact that the “Post PC” era includes implicit data caps (neither Verizon nor AT&T offer any unlimited data iPad packages to new customers) somewhat disturbing. I burned through just shy of 1 GB of my 2 GB AT&T plan in 3 days of use. By the time I got finished writing everything on my review sections, I had already used more than half.
Words cannot express my frustration - nay, rage - at the telecom industry's calculated assault on unlimited data plans, first in the mobile space, and now alarmingly with terrestrial DSL. Glass caps on both have always existed (200 or 250 GB for cable, 5 GB for essentially all mobile), but unless you clearly were abusing it, nothing happened. This move to tiered billing is a much more alarming trend. The above screenshot coupled with my use patterns (I have under 37 MB per day for the rest of the month? HA!) is precisely why I can't possibly see myself using the iPad 2 on a day to day basis. I hate watching a meter, but at least I do it. The problem isn't so much the number so much as the psychological effect of knowing you're eventually going to run into the celing.
WiFi
The iPad 2 uses the exact same 802.11a/b/g/n WiFi + Bluetooth 2.1 EDR stack as its predecessor - the BCM4329. Like the iPad 1, the iPad 2 has 65 Mbps (long guard interval, 20 MHz channel, 1 spatial stream) 802.11n support on both 2.4 GHz and 5 GHz. Essentially WiFi speeds and behavior close to the AP is exactly the same as it was before. I’ve seen other WiFi stacks on mobile devices with “802.11n” support provide short guard interval 20 MHz rates of 72 Mbps, but nothing above, yet, so it’s fair to say that the iPad 2 WiFi authentication rate is about par with the competition. That said, Bluetooth 3.0 is already here on a bunch of other devices.
Unfortunately, WiFi range on the iPad 2 is noticeably less than its predecessor, and throughput slows down much earlier on the iPad 2 than it does on the iPad 1. I tested with a WiFi - 3G (AT&T) model and found that I’m able to make it further away from the AP by a good 10 meters or so, but throughput even after you drop one WiFi bar on the iPad 1 is continually better than on the iPad 2.
WiFi throughput also tells a similar story, it just isn’t quite as good as its predecessor. It’s a bit unfortunate since Apple certainly could’ve used this refresh opportunity to be the first to introduce a two spatial stream 802.11n device.
For now, the wrap up with regards to the iPad 2’s WiFi performance is that close to the AP, things are just like they used to be. At the edge of WiFi coverage, the iPad 1 edges out the iPad 2 at maintaining higher negotiated throughput and actual speeds.
Facebook
Twitter
RSS
189 Comments
View All Comments
JarredWalton - Sunday, March 20, 2011 - link
Considering the source (ARMflix), you need to take that video with a huge grain of salt. It looks like they're running some Linux variant on the two systems (maybe Chromium?), and while the build may be the same, that doesn't mean it's optimized equally well for Atom vs. A9.Single-core Atom at 1.6GHz vs. dual-core A9 at 500MHz surfing the web is fine and all, but when we discuss Atom being faster than A9 we're talking about raw performance potential. A properly optimized web browser and OS experience with high-speed Internet should be good on just about any modern platform. Throw in some video playback as well, give us something more than a script of web pages in a browser, etc.
Now, none of this means ARM's A9 is bad, but to show that it's as fast as Atom when browsing some web pages is potentially meaningless. What we really need to know is what one platform can do well that the other can't handle properly. Where does A9 fall flat? Where does Atom stumble?
For me, right now, Atom sucks at anything video related. Sorry, but YouTube and Hulu are pretty important tools for me. That also means iOS has some concerns, as it doesn't support Flash at all, and there are enough places where Flash is still used that it creates issues. Luckily, I have plenty of other devices for accessing the web. In the end, I mostly play Angry Birds on my iPod Touch while I'm waiting for someone. :-)
Wilco1 - Sunday, March 20, 2011 - link
The article is indeed wrong to suggest that the A9 has only half the performance of an Atom. There are cases where a netbook with a single core Atom might be faster, for example if it runs at a much higher frequency, uses hyperthreading, and has a fast DDR3 memory system. However in terms of raw CPU performance the out-of-order A9 is significantly faster than the in-order Atom. Benchmark results such as CoreMark confirm this, a single core Atom cannot beat an A9 at the same frequency - even with hyperthreading. So it would be good to clarify that netbooks are faster because they use higher frequency CPUs and a faster memory system - as well as a larger battery...somata - Sunday, March 27, 2011 - link
CoreMark is nearly as meaningless as MIPS. Right now the best cross-platform benchmark we have is Geekbench. It uses portable, multi-threaded, native code to perform real tasks. My experience with Geekbench on the Mac/PC over the years indicates that Geekbench scores correlate pretty well to average application performance (determined by my personal suite of app benchmarks). Of course there will be outliers, but Geekbench does a pretty good job at representing typical code.Given that, the fact that a single-core 1.6GHz Atom (with HT) scores about 28% higher than the IPad's dual-core 1GHz A9s in the integer suite leaves me little doubt that the Atom, despite being in-order, has as good or better per-clock performance than the A9s.
Even the oft-maligned PowerPC G4 totally outclasses the dual A9s, with 43% better integer performance at 1.42GHz... and that's just with a single core competing against two!
tcool93 - Sunday, March 20, 2011 - link
Tablets do have their advantages despite what the article claims. For one thing, their battery life far out lives any Netbook or Notebook. They also run a lot cooler, unlike Notebooks and Netbooks, which you can fry an egg on. Maybe they aren't as portable as a phone, but who wants to look at the super tiny print on a phone.Tablets don't replace computers, and never will. There are nice to sit in bed with at night and browse the web or read books on, or play a simple game on. Anything that doesn't require a lot of typing.
Even a 10" tablet screen isn't real big to read text, but its MUCH easier to zoom in on text to read it with tablets. Unlike any Notebook/'Netbook, which its a huge pain to get to zoom in.
tcool93 - Sunday, March 20, 2011 - link
I do think the benchmarks shown here do show that there is quite an improvement over the Ipad 1, despite what many seem to claim that there isn't much of an upgrade.secretmanofagent - Sunday, March 20, 2011 - link
Anand,Appreciate the article, and appreciating that you're responding to the readers as well. All three of you said that it didn't integrate into your workflow, and I have a similar problem (which has prevented me from purchasing one). One thing I'm very curious about: What is your opinion on what would have been the Courier concept? Do you feel that is the direction that tablets should have taken, or do you think that Apple's refining as opposed to paradigming is the way to go?
VivekGowri - Sunday, March 20, 2011 - link
I still despise Microsoft for killing the Courier project. Honestly, I'd have loved to see the tablet market go that direction - a lot more focused on content creation instead of a very consumption-centric device like the iPad. A $4-500 device running that UI, an ARM processor, and OneNote syncing ability would have sold like hotcakes to students. If only...tipoo - Sunday, March 20, 2011 - link
Me too, the Courier looked amazing. They cancel that, yet go ahead with something like the Kin? Hard to imagine where their heads are at.Anand Lal Shimpi - Monday, March 21, 2011 - link
While I've seen the Courier video, and it definitely looked impressive, it's tough to say how that would've worked in practice.I feel like there are performance limitations that are at work here. Even though a pair of A9s are quick, they are by no means fast enough. I feel like as a result, evolutionary refinement is the only way to go about getting to where we need to be. Along the way Apple (and its competitors) can pick up early adopters to help fund the progress.
I'm really curious to see which company gets the gaming side of it down. Clearly that's a huge market.
Take care,
Anand
Azethoth - Monday, March 21, 2011 - link
Gaming side is a good question. Apple will have an advantage there due to limited hardware specs to code to. They are a lot more like a traditional console that way vs Android which will be anything but.Are actual game controls like in the psp phone necessary?
I am also curious what additional UI tech will eventually make it to the pad space:
* Speech, although it is forever not there yet.
* 3D maybe if its not a fad (glasses free)
* Some form of the Kinect maybe to manipulate the 3d stuff and do magical kinect gestures and incantations we haven't dreamed up yet.
* Haptic as mentioned earlier in the thread.
Speech could make a pad suitable for hip bloggers like the AnandTech posse.