A Look at Real World Performance of an Overclocked Phone
In order to test just how practical overclocking an Android device really is, we set up a test device, ran it through a bunch of benchmarks and tests, overclocked it, and ran it through those benchmarks and tests again. Our test device was the Samsung Nexus S. At fifteen months old and starting to show a bit of stutter with a bunch of apps installed, it made an excellent candidate.
We were easily able to bring the 1GHz processor up to 1.4GHz, which you might think would suggest a 40% increase in performance, but in most benchmarks, we saw nowhere near that kind of a gain. The CPU is just one part of the entire system. Overclocking your Android typically won’t affect the GPU or memory and storage busses. Overall, that 40% clock boost was more like a 10-20% system-wide boost on average.
|Smartbench 2012 Productivity Index
|Smartbench 2012 Game Index
Looking at the benchmarks that test overall system performance, it’s clear that some do give preference to CPU speed, particularly Smartbench 2012’s Productivity Index, but the average increase across all of the benchmarks is just under 20%.
As we can see with the gaming benchmarks, GPU dependent games see very little gain from overclocking, though we think it’s important to note that the results from Nenamark1 and Neocore appear to be limited by the refresh rate of the screen. Nenamark2 and Electopia aren’t affected by this issue, however, and neither show much improvement in 3D gaming. It’s a pretty safe bet that overclocking your Android phone won’t make a choppy game more playable.
|Linpack Score (lower is better)
|Score in MFLOPS (higher is better)
Linpack, which is heavily CPU dependent, showed the most gain after overclocking. Actions that are primarily CPU intensive, such as system navigation, web browsing, and general productivity apps, appeared noticeably more responsive; however, this was not entirely due to overclocking. We started noticing these improvements after flashing the custom kernel, but before overclocking. Overclocking did improve overall smoothness further, but tweaks from the custom kernel also played a big part.
While on paper, overclocking while undervolting results in increased battery life, actually testing and measuring this increase was a bit more difficult. Most smartphone battery testing metrics run the system at 100% load with the screen on full brightness in order to get a consistent test. During this test, the overclocked settings got roughly 15% worse battery life than the stock speeds due to the 1.4GHz clock rate requiring a higher voltage to sustain.
In a real-world scenario, your phone would spend a large portion of its time in a low power state which would be largely benefited by the custom kernel’s undervolted settings. Additionally, the overclocked high power state would allow it to finish its workload faster and return to a low power state sooner. Theoretically, this would result in less power draw overall. Unfortunately, it’s very difficult to test in a short amount of time; however, while trying to keep daily use as equal as possible, we did see a small increase in battery life of about 12% with the overclocked and undervolted settings. Due to time constraints we weren’t able to test how much of this was strictly thanks to special kernel features versus the overclocking itself.
Ultimately, we discovered that overclocking your Android really isn’t about performance gains. It’s enough to take a borderline laggy device and make it run smoothly again, but it likely won’t make your games run much faster. The byproducts of overclocking, such as the improved kernel, added features, and extra battery life, are really what make it worthwhile.
Current page: A Look at Real World Performance of an Overclocked PhonePrev Page Step Three: Using SetCPU
Get the BEST of Tom’s Guide daily right in your inbox: Sign up now!
Upgrade your life with the Tom’s Guide newsletter. Subscribe now for a daily dose of the biggest tech news, lifestyle hacks and hottest deals. Elevate your everyday with our curated analysis and be the first to know about cutting-edge gadgets.
I have my HP touchpad overclocked to 1.89GHz which is somewhat decent from the stock 1.2GHz. Really breathes new life into a dual core ARM CPU.Reply
PS wanted to add that there is a way to save battery life without losing performance.
Justy like how every CPU has it's own max overclock, every CPU also has it's own minimum voltage per given clock speed
You can extend your battery life by adjusting the voltage for each clock speed step to a lower level (to test for stability, use your overclocking app to cap the clock speed to the desired level, then run a few benchmarks)
you can generally lower almost all of the voltages and thus give your self a decent battery life boost without losing any performance, and best of all, you can still keep your overclock
simply set the lowest stable voltage for each clock speed in the list.
During what most would consider normal use, you rarely ever hit the CPU's max speed, and if you overclock, you are even less likely to hit the overclock speed (many overclocking apps will record the CPU clock speed history and you will see how little those higher clock speeds are used (but when they are used, they really show their benefit, (especially when loading web pages and launching applications)
All in all, find your max clock speed and fine your lowest voltage for each clock speed in the table, and you will have a device that bas better battery life and much faster performance
(the lower voltage tweak takes longer to do than the max overclock because you will be testing every single clock frequency in the list for stability each time you lower it's voltage
wanted to also addReply
(sry for the typos in the previous message sees they still did not add a edit feature)
anyway, most of those benchmarks base their scores on more than just the CPU but the overclocks only really effect the CPU, (depending on the device, you can overclock the L2 cache and in some cases, even the GPU and RAM )
I think your Linpack results are a bit off. A higher score in Linpack is better, and the Nexus S can get 60+ MFLOPSReply
HurrpancakesI think your Linpack results are a bit off. A higher score in Linpack is better, and the Nexus S can get 60+ MFLOPSReply
Very sorry, you're right! The Linpack table is incomplete. The numbers listed are the times in seconds it took to complete the test.
The stock Nexus S speeds scored an average of 16.8 MFLOPS and the 1.4GHz speeds scored 23.6 MFLOPS. While the Linpack "top scores" page lists the Nexus S getting 60+ MFLOPS, it's not actually true. Those scores are from other devices running custom ROMs ported from the Nexus S.
Ugh, I've tried rooting my HTC Aria but it never worked. DARN YOU AT&T!!!Reply
It is no surprise that overclocking the CPU did not do much for game benchmarks, this is no different to PC gaming unless it is a heavy CPU intensive game. You should of overclock the GPU as well. Tegrak Overclock Ultimate is what you want to use to overclock both CPU and GPU.Reply
I think people would be interested in seeing real world performance difference between overclock GPU and CPU
After getting my Epic 4G (Galaxy S) when it had Eclair on it, I've found the performance increase from overclocking pales in comparison to optimization brought on by Google and the excellent ROM community. Froyo brought a JIT compiler, Gingerbread EXT4 partitions and ICS finally has a hardware accelerated GUI, making it iOS smooth. The ROM community has really cleared out the crap, making my phone lean. Compared to my Eclair Epic, my ICS Epic feels like it's a full generation ahead. Overclocking... yeah it helps a bit, but software optimization does so much more in the Android world.Reply
if you look at the benchmark details you will see that the 40% overclock actually netted a 40% boost in the CPU benchmark scoreReply
PS while overclocking helps, a device with more RAM helps even more especially when paired with a custom ROM. if your device has 1GB of RAM, then you can have custom roms where more elements of the OS are buffered into RAM, this allows for a more iphone like experience when it comes to transitions in the UI. By default, android unloads UI elements and other graphical elements from memory when they are not visible, but it can be modded to keep it so you get a smoother UI experience, eg widgets not having to redraw when you rotate the screen, and all native UI's within the OS load instantly. (takes more RAM but not an issue if you have 1GB
PS for really good demo of what overclocking can do, try playing HD video that is not GPU accelerated, on my HP touchpad, I can have 3 720p flash videos playing at the same time with no lag when at 1.89GHz
I can also play a single, large 1080p 10 bit movie (10+GB in size) smoothly and that MKV file is not supported by the GPU so that is all CPU
If you have an android or any other device that can be overclocked, then you owe it to yourself to overclock it, the benefits are not noticeable at all times but during the times when they are noticeable, they are really really noticeable.
HI FRIENDS! The Linpack chart has been updated.Reply
-Devin Connors, Tom's Guide
It's been a good nearly two years since this article got put out. And it's been a good nearly four years since my LG Optimus S released to the public. From my experience with that old phone, adding a custom ROM helps to speed it up quite a bit and overclocking really makes it fly in comparison to its stock performance.Reply
On newer hardware the benefits of overclocking may be less noticeable, but it is very noticeable on old devices.
Right now my LG Optimus S runs Android 4.1.2 and is overclocked to 768MHz, using SetCPU, when the battery isn't being charged and the screen is turned on and in use. Not only am I overclocking it, but I'm also attempting to balance out performance and battery life by using profiles in SetCPU to underclock it when not in use. This is compared to when it ran stock Android 2.3.3 at 600MHz. Yes, the performance increase is very noticeable and makes installing a custom JellyBean ROM and overclocking very worthwhile.