Smartphone download and upload speeds are becoming more important as Internet gaming, graphic rich apps, and multitasking are further enabled by improvements in Smartphone technology. The latest Smartphones have both 3G (and/or 4G) cellular and Wi-Fi technologies, but the rated speed of the cellular (or even Wi-Fi) technology doesn’t necessarily indicate that your handset will perform as expected – download and render a website to its screen similar to a laptop – which is the ultimate handset user experience. This performance requirement goes beyond just rendering a website, it bleeds over into time critical applications such as online gaming.
Several factors influence the Smartphone’s download (DL) and upload (UL) speeds as well as its rendering to the screen:
1) cellular technology (3G or 4G)
2) available bandwidth
3) website complexity and size
4) traffic to that website and its server’s ability to deliver content
5) the internal hardware capability
The entire link between Cloud-based content and a Smartphone’s display is primarily influenced by the above five (5) factors and affects Web-surfing as well as online gaming. Performance expectations along with novel uses of the Smartphone are continuing to grow as new phones are released. In addition, Smartphone capabilities have improved at an increased rate this past year (2010), as seen by releases and announcements of 4G (WiMAX and LTE) technologies, A9 dual-core processors, video encoder/decoder that supports 30fps video playback, etc. What other applications can a Smartphone be used for other than to communicate directly with the Cloud for Web-surfing and online gaming?
Although not a new idea, the Smartphone can be used as a Mobile Hot Spot, thereby allowing a person with a laptop to Web-surf remotely – an indirect use for Web-surfing. Yet to make this a desirable user experience the Mobile Hot Spot should perform similar to a Wi-Fi connection at home or the local café; a performance short-fall, until recent Smartphone entrants. Whether a website is rendered to the Smartphone’s display, or to a Laptop’s display when used as a Mobile Hot Spot, the entire system from website to the device’s screen makes an impact on the performance and the user experience.
“Mobile Hot Spot” Performance Testing
Most Smartphone’s offering the Mobile Hot Spot feature can connect up to five (5) laptops or any portable device that has Wi-Fi technology. Laptop Mag (laptopmag.com) conducted a series of tests using some of the latest Smartphones that are Mobile Hot Spot enabled. The Mobile Hot Spot feature uses the phone’s cellular (data) network to connect to the Cloud, while using its Wi-Fi to connect to a laptop’s Wi-Fi, acting as a bridge or “pipe” for the laptop, see Figure 1.
Source: Imagines taken from Internet
The test by Laptop Mag to determine the Best Mobile Hotspot Smart Phone took place this past September (i.e. Sept 2010). The results are summarized in a matrix in Figure 2 of download (DL) and upload (UL) speeds and Web-surfing times, in seconds. The testing used a Toshiba Satellite A665 notebook as the Wi-Fi connected test device, several Smartphones acting as the Mobile Hot Spot, and SpeedTest app on the notebook to test the DL/UL connection speeds through each Smarpthone. For the 3G/4G Smartphones data matrix in Figure 2, the number on the left side in the cell is the 3G time (in seconds), and the right side is the 4G time (in seconds). The DL/UL speeds for the HTC Droid, HTC Evo 4G and Samsung Epic 4G are comparable to speeds achieved with “average” home connected cable technology available today.
Figure 2 – Results of Laptop Mag’s Download and Upload Testing, Mobile Hot Spot
The testing was conducted in New York City but used only Verizon and Sprint networks because these carriers only had phones with Mobile Hot Spot functionality. AT&T and T-Mobile will soon offer phones with Mobile Hot Spot functionality, however.
An overview of the testing is available on the Laptop Mag’s blog. The data in Figure 2 shows that the DL/UL speeds and rendering times to the Laptop’s screen produce a desirable Web-surfing user experience, but the blog write-up doesn’t describe the details that can affect the DL/UL speeds; for example, the amount of activity on the website during the time of testing, which can slow a server’s response; or multiple Laptops using a single Smartphone that divide up the available bandwidth, reducing the DL speed and increasing the time to download. Also, the size (in kilo-bytes) and complexity (e.g. flash) of the website can affect the download response.
All these websites have different sizes and complexities that change throughout the day, week, etc. For instance, the download time of the Laptop Mag’s website versus NY Times’ website (see Figure 2) suggests that Laptop Mag’s website has twice the amount of content to download, which may not be the case. However, Laptop Mag’s article does describe and examine the innovative use of the Smartphone as a “pipe” to link a Laptop to the Cloud using cellular + Wi-Fi technologies, i.e. turning the Smartphone into a Mobile Hot Spot. I recently took Laptop Mag’s testing to the next level to better understand and pinpoint the potential bottlenecks of the Smartphone and Cloud link that slow the rendering of a website on the handset’s screen – is it the cellular (data) network or the handset’s internal hardware?
Additional Data Gathered
Since Laptop Mag already provided data using the Smartphone as the connectivity pipe, I wanted to first test the Web-surfing experience without the Smartphone – using a (Ethernet) connected Desktop/PC and a Wi-Fi connected Laptop, to a modem connected via cable to the Internet. Second, I tested a (standalone) Smartphone’s cellular (data) connection and then its Wi-Fi connection (in separate tests). This additional data helped to understand and differentiate the user experience between downloading and Web-surfing through the Smartphone (using it as a Mobile Hot Spot) and downloading and Web-surfing to the Smartphone (using Wi-Fi, 3G and 4G technologies).
Figure 3 – DL/UL Test Results for Desktop/PC and Laptop, non-Mobile Hot Spot
Note:
1 - Used tool from Pingdom.com. Sends out ping packets and provides network response times
a. Times of less than 50 milli-secs are considered “fast” response times
2 - Rendered full website on screen
3 - Connected to server in Palo Alto , CA
4 - Connected to server in Wichita , KS
Several tests were taken for each case and the average provided, with the Samsung Intrepid in 3G (CDMA) mode having a much greater variation in DL/UL speeds, see Figure 3. Also, the number reported within the matrix cell on the left side is the Pingdom.com test time, and the number on the right side in the same cell is the download and rendering test time, both in seconds. For example, for the Desktop/PC test case and for the espn.go.com website, which was 1,025k bytes and 31 objects, it downloaded in 2.9 seconds using the Pingdom.com tool, and downloaded and rendered (in real-time) on the Desktop/PC screen in 4.2 seconds. In comparison, the Samsung Intrepid/Wi-Fi test case downloaded espn.go.com in 4.3. seconds using the Pingdom.com tool, and downloaded and rendered (in real-time) on the handset’s screen in 45 seconds – a bottleneck?
A noteworthy outcome of the testing was the rendering of the website to the Samsung Intrepid’s screen in both 3G (CDMA) and Wi-Fi modes, which took a long time (in comparison) and translated to a poor user experience. Although the varying bandwidth in 3G (CDMA) mode could be the part of the cause for the slower download speed, rendering to the screen is significantly affected by the handset’s internal hardware design and this affects performance as well. The download and rendering times for the Samsung Intrepid using either the 3G (CDMA) or Wi-Fi connection were almost the same (~45 seconds), which points to the internal hardware design as the bottleneck. This is most apparent when compared to other (newer) Smartphone’s in Figure 2, which rendered the website to the Laptop’s screen quickly (in less than ~10 seconds) using either 3G (CDMA) or 4G (WiMAX) technologies.
To further understand and compare the performance of the new 3G and 4G phones with the Samsung Intrepid data, I then tested the HTC Evo 4G phone in both 4G (WiMAX) and Wi-Fi modes (same phone tested in Laptop Mag’s testing), and tested the HTC Hero phone in both 3G (CDMA) and Wi-Fi modes. The results of the data gathered are shown in Figure 4.
Figure 4 – HTC Evo 4G in 4G/Wi-Fi modes and HTC Hero in 3G/Wi-Fi modes, to handset
Note:
1 - Used tool from Pingdom.com. Sends out ping packets and provides network response times
a. Times of less than 50 milli-secs are “fast” response times
2 - Connected to server in Palo Alto , CA
3 - Connected to server in Los Angeles , CA
4 - Mobile website version was downloaded
The numbers in Figure 4 included downloading each website to the Smartphone and rendering it to its screen. The Pingdom.com tool was not used for supplementary testing since it became apparent in the previous testing results (in Figure 3) that this tool provides information about the website’s objects and download time only to the Smartphone, and was consistently faster than downloading and rendering to the Smartphone’s screen. The testing results in Figures 2 thru 4 provided additional insight into various connectivity uses of the Smartphone and its performance regarding Web-surfing. The connectivity uses can be summarized into three basic categories:
A. Smartphone using its “Mobile Hot Spot” functionality – Wi-Fi + 3G/4G
B. Smartphone using its Wi-Fi technology
C. Smartphone using its cellular (3G/4G) technologies
The data in Figure 4 clearly shows the 4G (WiMAX) network downlink performance (at 9.5 Mbits/secs) surpasses the 3G (CDMA) network (at ~0.4 to ~0.9 Mbits/secs) and Wi-Fi connection (at ~3.8 Mbits/secs). The ping test varied between technologies and the ping performance provided better results when the Smartphones were connected via Wi-Fi to the network, at ~117 milli-seconds. The faster the ping response, the better the network response is for time critical applications. The ping tests for Smartphone-Wi-Fi and cellular cases, however, were not as fast as the Desktop/PC-Ethernet and Laptop-Wi-Fi tests, which were on the order of ~30 milli-seconds (see Figure 3).
The Web browsers for the newer Smartphones detected espn.go.com and nytimes.com websites’ mobile versions, and these mobile versions reduced the download times significantly, as seen in Figure 4. In fact, from the testing I was able to conclude that the 3G (CDMA) network with a sustained minimum downlink of ~0.7 Mbits/sec (and an uplink of 0.15 Mbits/sec), would have just enough bandwidth to download the larger and more complex websites, while providing a good user experience. The only anomaly I came across was the downloading and rendering of the Laptop Mag website (which didn’t have a mobile version), and that took ~37 seconds using the 3G (CDMA) network This longer time (i.e. slower download speed) may have been caused by a slow 3G downlink speed at the time of testing, which I found varied often during testing.
With data in hand and looking back at the five (5) criteria that can influence Web-surfing using a Smartphone, the DL/UL speeds and rendering of the website to a Smartphone’s screen is well understood. The difference in cellular (data) technology, 3G vs. 4G, can make a performance impact, especially since the 3G technology’s bandwidth varied considerably throughout the testing, bandwidth limiting the connection and affecting the download times; however, the 3G download speeds in Laptop Mag’s testing are several hundred to a mega-byte higher than my testing, which makes a difference regarding the time to download and render content to the Laptop’s screen. Yet it was determined that a certain sustained minimum bandwidth on a 3G network could still produce a good user experience for Web-surfing on the Smartphone. In addition, mobile versions of the websites enabled faster download times for the newer Smartphones, thus providing a very good 3G (CDMA) experience.
Traffic to the website and the website’s server ability to deliver content wasn’t directly tested but the websites were tested during the day, evening and weekend, and found little difference in performance (when 3G bandwidth limiting wasn’t an issue!). The performance of the newer Smartphone’s compared more closely to a wired Desktop or Wi-Fi connected Laptop, thus “a sign” of closing the performance gap between portable and mobile devices.
The quick display (i.e. render) of content on the Smartphone’s screen is a salient feature that users are becoming accustomed to with improvements in technology. There are fundamental (wired vs. wireless) network limitations as noted in the testing, but the quick rendering of text and media (for Web-surfing and online gaming) is necessary for all new Smartphones released in the future, thus continuing to “raise the bar”. Besides the cellular (data) network limitation, it was determined that the bottleneck which limits the rate of rendering is affected by the Smartphone’s internal hardware design.
A Look At The Internal Hardware Design
The movement of data (in bits or bytes) from its wireless connection through the phone’s connected hardware and to the display depends on the hardware, inter-connectivity, and related software/firmware. See Figure 5 for a general diagram of a Smartphone’s internal hardware connectivity.
Source: Imagines taken from Internet
Figure 5 – Exploded View: General Internal Hardware Connectivity of a Smartphone
There are a number of factors as mentioned within the Smartphone’s design that affects the final display (or rendering) as the bits move from the cellular or Wi-Fi hardware to the screen. The following matrix, see Figure 6, shows the application processors and GPUs for each of the Smartphones used in the Laptop Mag’s and my testings.
Figure 6 – Matrix of Smartphones, the OS, and Apps Processors and GPUs
The high-end Smartphones that contain the latest application processors and GPUs will provide a better Web-surfing experience, as noted in the testing results. Qualcomm’s Scorpion and Snapdragon platforms, as well as Samsung’s Hummingbird and TI’s OMAP platforms, are in the Smartphones tested. These platforms not only improve the rendering time of a webpage, but enable better gaming experience and high-definition (HD) video decode and playback. A comparison of these three processors is available at PhoneArena.com. Other competitors offering high-performance application processor and/or GPU products include: Nvidia’s Tegra/2 platforms, Freescale’s i.MX multimedia applications processors and ST-Ericsson’s NovaThor platform.
Qualcomm’s GPU – Adreno platform – can process 32M triangles/sec (Adreno205) and 22M triangles/secs (Adreno200), enabling quick processing of rich graphic content. “Triangles per second” is one measure of a GPUs graphic computing power. The PowerVR used in the other Smarphones tested doesn’t have the same “triangle” computation spec/capabilities as the Adreno, but visually (i.e. empirically) produced similar rendering results. From my testing, the HTC Evo 4G had superior Web-surfing performance versus the competition. However, an article by Digital Trends mentions that the Samsung Epic 4G produced better graphics performance for gaming tests (Epic 4G: 55.1 fps, frames per second vs. Evo 4G: 42.9 fps) and Neocore benchmarking than the HTC Evo 4G, yet the sum total of all areas tested gave the HTC Evo 4G the “win”.
As determined through testing, both the cellular (data) connection speed and application processor/GPU performance contributed to bottlenecks that limited the content rendering to the handset’s screen. Gray areas exist with even the newer Smartphones that prevent a consistent high-level user experience for Web-surfing. These gray areas include: varying 3G cellular connectivity speed and the application processor design (which includes the GPU). Further improvements in the application processor and GPU performance will bring near-Laptop performance levels to newer Smartphones in the future. Newer 3G technologies, such as HSPA or HSPA+ will reduce the probability that your cellular (data) connection will throttle-back to an undesirable level, as can happen in densely populated or high-use areas.
Minimum sustained speeds could be imposed by the carriers but that may also block other users from data network access, creating an even worse experience – no network access – or “minutes” to download a complex website. Higher data rate cellular technologies, i.e. 4G – WiMAX and LTE, that are becoming available in the newer Smartphones will “open the pipe” to further accommodate data network traffic. Motorola’s announcement at CES 2011 of their Atrix 4G Smartphone is a performance monster. Due to be released in March 2011, it uses an Nvidia Tegra 2 dual core processor (Dual-Core 1GHz Cortex A9 + GeForce ULP), cellular technology – HSDPA (category 10 @ 14.4 Mbps), and has Wi-Fi (802.11 a/b/g/n – 2.4 and 5 GHz band technologies), amongst other technology upgrades relative to the devices tested in this article. HSDPA category 10 isn’t considered 4G technology but offers significantly more bandwidth (and nearly 4G speeds). The additional cellular data speed and the Nvidia Tegra 2 processor will make for a superior Web-surfing experience as well as online gaming experience using the Atrix 4G.
Perspective
As mobile users continue to utilize their Smartphones for Cloud connectivity, the demand for improved Web-surfing will grow. Most websites in the Cloud have sizes that range from 40 kbytes to 1,500 kbytes in size and contain 6 to 140 objects. Larger websites can be demanding for Smartphones, thus providing an average user experience, where as a wired or wireless Laptop performance sets the benchmark for Smartphones to achieve.
Mobile website versions do make the Web-surfing experience “better” for a Smarpthone, especially when data network speed limitations come into play. Once the data network limitations are minimized, then the Smartphones’ internal hardware design becomes the potential bottleneck to render the full website on the display in a time that meets the user’s expectation.
From the testing conducted, Desktop/PC-wired and Laptop-wireless downloaded and rendered websites in less than 7 seconds, thus establishing a target for Smartphones to achieve. Most websites were rendered within 10 seconds on the newer Smartphones. My testing of the Samsung Evo 4G in WiMAX (4G mode) took about 15 seconds to render with no network limitations (where as Laptop Mag showed less than 5 seconds to render on the Laptop’s screen).
My user experience was still very good since I could scroll around on the screen with my finger to view content as it downloaded. My experience with the HTC Hero was good, but Laptop Mag’s website took about 37 seconds to fully download; however, I could still scroll around the screen with my finger to view the content as it downloaded. The Samsung Intrepid was an agonizing experience and lacked flexibility to maneuver on the screen to view content.
The up-and-coming Smartphones – such as the Motorola Atrix 4G and others to follow in 2011 – will have the higher performance, and there is much anticipation to see if they provide a wired-like experience that accommodates the demands by the gaming enthusiast and multi-taskers, prosumers, and Gen-X and Gen-Yers.
Smartphone or Tablet, together with compelling features (e.g. Mobile Hot Spot functionality) and offered at an array of prices, does make a consumer wonder what Cloud-accessible devices will the industry start to consolidate towards in the near future. And does the consumer want to always wait two (2) years before they buy a new Smartphone or will the carriers consider new contracts to allow yearly upgrades by the consumer so they can take advantage of the newer technology, sooner?
No comments:
New comments are not allowed.