NAND Performance

The $199 Nexus 7 ships with 8GB of NAND on-board in the form of a standalone eMMC 4.41 package from Kingston. The single package storage solution features NAND and controller, although as you can guess there's no room for DRAM. The NAND in use is MLC although it's not clear whether we're looking at 2 or 3-bits per cell. This basic architecture shouldn't be all that unfamiliar to long time AnandTech readers - the cacheless design is exactly what plagued some of the early MLC SSDs for PCs a few years ago.

More recently users have been noticing very poor performance with ASUS Android tablets whenever there's substantial background IO going on, particularly multitasking while writes are taking place (e.g. downloading a torrent and browsing the web). The slowdowns range from multi-second long pauses to sluggish response time. The video below shows you what can happen on a Transformer Pad Infinity while downloading a bunch of large files in the background:

Even though downloading a large file is a largely sequential write operation, any other action performed while that download is in progress will effectively turn the IO stream from purely sequential to pseudo-random. Most of these eMMC and other cheap MLC NAND flash controllers are really glorified SD cards, primarily optimized for reading and writing large images as if they were used in a camera. What they're not designed for is to run and be used in a full blown multitasking OS. Some manufacturers seem to do a better job of picking their storage solution, and the Kingston eMMC in the Nexus 7 is faster than the Hynix e-NAND ASUS has used in previous tablets. Neither is perfect however. There's a tangible impact on simple multitasking if you're downloading a lot of files or installing apps in the background. Even operations in memory are quite negatively affected by background IO. Take a look at how SunSpider performance is hurt by a background file download:

SunSpider 9.1 Performance
  Default With 2MB/s Background Download
Google Nexus 7 (8GB) 1665.9 ms 1984.9 ms

To put all of this in perspective, I turned to Androbench, a storage benchmark that allows an ok amount of flexibility in configuring the test environment. At its default settings the benchmark varies a bit too much for my liking, but if you up the buffer sizes to 100MB it helps smooth out some of the variance.

Here we're looking at both sequential and random IO, the two sides we pay attention to in our SSD reviews as well. It's not clear to me whether or not Androbench is clamping the random IO to a particular range of addresses or not, but the results are telling nonetheless:

Androbench (100MB read/write)
  256KB Sequential Read 256KB Sequential Write 4KB Random Read 4KB Random Write
ASUS Transformer Pad Infinity (64GB) 15.9 MB/s 7.13 MB/s 4.90 MB/s 0.22 MB/s
Google Nexus 7 (8GB) 23.1 MB/s 8.43 MB/s 4.77 MB/s 0.32 MB/s
Samsung Galaxy Nexus (16GB) 26.8 MB/s 7.29 MB/s 7.46 MB/s 0.23 MB/s

Sequential read/write speed isn't bad, but it's the random write speed that's really a problem. We're talking about write speeds of a couple hundred KB/s. Remember what I said earlier about how multitasking can take an otherwise sequential IO stream and make it look fairly random? I suspect the low random write performance is one reason we're seeing significant slowdowns with background IO. Not all NAND controllers do well with concurrent reads and writes, which could be another contributing factor to poor performance.

Note that for light usage this isn't a problem. Similar to the first generation of affordable MLC SSDs for PCs, as long as you're doing a lot of reading you'll be ok. It's really for the heavier usage models that this is a problem. On a tablet however, simple background installation or downloading of files counts as heavy these days.

The real solution to this problem is to integrate better NAND flash controllers on-board, or even onto the SoC itself. Tablet makers are still mostly focused on cost cutting, but eventually we'll see real SSDs with good controllers in these things. Windows 8 tablets, at least the high end ones due out later this year will be among the first to incorporate real SSDs.

WiFi Performance, NFC

The Nexus 7 includes support for single spatial stream 802.11b/g/n on 2.4 GHz only. That ends up being a PHY rate of 65 Mbps (single spatial stream with no SGI). ASUS has been using AzureWave modules for basically every single one of its tablets, and the Nexus 7 is no exception. Thanks to the FCC test reports, you can see an AW-NH665 module alongside the new NXP PN65 NFC controller. The AW-NH665 includes a BCM4330 inside - remember there are different vendors who will sell you a completed module.


NXP PN65 NFC, AzureWave AW-NH665 (Source: FCC Internal Photos)

If you're looking for a cost cutting measure, the lack of 5 GHz WLAN support is most definitely one. In fact, the lack of 5 GHz WLAN made for probably one of my most memorably hilarious Google I/O experiences, since the event had multiple signs noting that only devices with 5 GHz WLAN would work well (or at all - 2.4 GHz is unuseable at any major event) on the conference WiFi. To mitigate this, Google hooked up USB OTG to MacBook Air ethernet adapters for its Nexus 7 demo stands, which I of course hijacked for my Galaxy Nexus at one point. More and more, having 5 GHz WLAN in a tablet is an expectation, and soon we'll even start seeing 2x2:2 antenna configurations.

The resulting performance is pretty par for the course when it comes to WiFi speeds among single spatial stream 2.4 GHz only devices. I tested using iperf the same way I do smartphones. 

WiFi Performance - iPerf

Unlike some of the other ASUS tablets of note, there are no reception issues with the Nexus 7 either. I have no problem getting good reception and see the Nexus 7 fall off the network where I expect it to in my area. I think it's worth being explicit about where the antennas are located since wireless connectivity issues have been a recurring issue for just about every product. 


BT/WLAN (red), NFC (blue), GPS (green)

I've gone ahead and marked where the antennas are for BT/WLAN, NFC, and GPS on the case from the FCC teardown photos. Knowing exactly where the NFC coil antenna is located is actually hugely important for correctly positioning the device for beaming or reading tags. It ends up being right behind the "nexus" recessed lettering on the back, but until I saw the antenna in the FCC teardowns I wasn't completely sure. With a smartphone the size makes it relatively forgiving, but tablets are less forgiving about positioning not being exact thanks to the larger size. 

GPS

The Nexus 7 uses a monolithic Broadcom BCM4751 GPS receiver, which we've seen in other devices like the iPad 2. This isn't the absolute newest broadcom GPS/GNSS, and again I'm sure ASUS was able to secure great pricing for this since the newer GNSS (GPS+GLONASS) modules are making their way into devices right now. 

That said, I had no problems with GPS locking quickly from cold and warm starts. I walked around San Francisco with the Nexus 7 using Google Maps (tethered to an SGS3) and saw pretty good performance in the urban canyon scenario. With no occlusions to the sky, performance is excellent, and locks indoors are possible too. 

 

GPU Performance Battery Life
Comments Locked

118 Comments

View All Comments

  • chrnochime - Thursday, July 26, 2012 - link

    Great product for those who don't mind reading on a 7" screen, but for me even the 10" gets tiring after a while, though considerably less so than a 7".

    I am amazed at those who are able to read the tiny texts on a 7" for a prolonged period of time. Tried that with a 10" and couldn't last more than 1hr. Uh.
  • amdwilliam1985 - Thursday, July 26, 2012 - link

    Tell that to the girls who are reading on iPhone with 3.5", I can barely see the tiny texts on their screen.
    Going from 3.5" to 7" will 4x their reading area, will make a heavenly difference for them.
  • jamyryals - Thursday, July 26, 2012 - link

    What does being a girl have to do with it? I read books on my 4s quite often when traveling. Last time I checked I was a man... Yup, still am.
  • amdwilliam1985 - Friday, July 27, 2012 - link

    Sorry to bring girls into the topic.
    I'm only saying that because I see a lot of girls in NYC subways with iPhone 4/4S. Most guys I saw in NYC subways are with bigger phones. Just reporting my observation.

    And the point of the topic is that if people can read on 3.5", then reading on 7" will be a dramatic(4x) improvement.
  • Akilaehunter - Thursday, July 26, 2012 - link

    Every buy a paperback with uncomfortably small lettering? Or even a hardcover that tries to look artsy by using an inconvenient font?

    All (worthwhile) e-readers and e-reader apps allow for margin size, text size, and font style changes.

    Argument is invalid; You can read a book on a 4" screen as optically comfortably as on a 10" screen, you'll just wear out your fingers flipping resized pages.
  • chrnochime - Thursday, July 26, 2012 - link

    It's not invalid; I'd rather not have to enlarge everything and scroll a whole bunch just to read one 8" x 11" webpage.

    Yes I can actually see the texts all shrunken on a 7", but I have no intention to torture myself in doing so. Rather spend that time on something more productive.

    Besides I *did* say it's a great device didn't I?

    And not all e-readers are stuck in 7". The Amazon DX is 10" which I also have, so.
  • Akilaehunter - Friday, July 27, 2012 - link

    Webpages, you're 100% right. I don't own a 10" tablet and even in landscape mode on a 7" many webpages can get uncomfortably small. Plus the scrolling.

    Still, if the nook/kindle app is uncomfortable to read books with it's one's own fault for not making it comfortable with the many combinations of size options.
  • CeriseCogburn - Sunday, July 29, 2012 - link

    It's never comfortable as the screen needs to be closer so the eye fry syndrome overtakes any blurry eyed goon who can't read small text.
    I suspect a far sighted elderly person with bifocals can hold it at arms length and "telescope" in to the text, but then, have never seen that.

    Next we'll have a flurry of articles analyzing peeps with facial skin cancer then we'll have the scare articles that quote the hidden study.

    In the mean time the tiny teens and teeny texts can increase corporate visine profit margins.

    They should call these devices "fryballs!"
  • Super56K - Monday, July 30, 2012 - link

    Yes, because before e-readers/tablets all the books I purchased were 10" + in physical size, and I had to hold mass market paperbacks 5" away from my face to read the tiny words. You're a feisty one Cogburn.
  • Akilaehunter - Thursday, July 26, 2012 - link

    First off, this is my first article response, though I've been lurking since 2004.

    A nitpick to start things off: I know this site tests a ton of mobile hardware so why are some models shown for some tests in an article and not others -in the same article- as comparisons? For instance, my Droid Razr Maxx is in a few comparisons (not this article) but not all. Should maybe be a database cataloguing them all, ala the gfx card benches section? (Or if there is I'm too dense to see it...)

    Otherwise, excellent article as always. Especially interested in how the flash storage is the main bottleneck now. Tegra 3 could be a beast at multitasking if all its cores are accounted for, but would then get strangled by IO as soon as it is told to multitask- something a multi-core cpu should be amazing at.

    Lot to think about. :D

Log in

Don't have an account? Sign up now