POST A COMMENT

56 Comments

Back to Article

  • g00ey - Tuesday, April 02, 2013 - link

    The Google Nexus 10 has a considerably higher screen resolution (2560x1600) than the Nexus 4 (768x1280) or the HTC One (1080p). This might explain why the Mali T604 doesn't look so good on these benchmarks. Reply
  • MonkeyPaw - Tuesday, April 02, 2013 - link

    If I understand the article right' resolution doesn't matter. Scenes are drawn at a set res and then scaled to native. The only overhead will be from the scaling. Reply
  • bearxor - Tuesday, April 02, 2013 - link

    Uhhh... Yeah.

    Someone didn't read the article and instead decided to skim the charts.
    Reply
  • theduckofdeath - Tuesday, April 02, 2013 - link

    The GPU and processor still has to work harder to actually output the extra pixels, even though the rendering is done at a specific resolution. With a bit of reverse calculations, someone with a bit of extra time should be able to calculate how much of an impact actual output resolution has on a benchmark like this. though, maybe it's not that easy, as different devices with "the same internals" but different resolution displays might have memory with different performance, for instance. Reply
  • Exophase - Tuesday, April 02, 2013 - link

    Mobile SoCs often have support hardware for scaling a screen overlay to native resolution. This is often built into the display controller so there's no performance impact even to available memory bandwidth.

    No idea if such a thing is actually used here though.
    Reply
  • ellis2323 - Friday, May 03, 2013 - link

    I'm a gamedev (DX, GL, OGLES...).This a OpenGL ES 2.x scene. Rendering is done on a FBO and your draw it on screen with 2 triangles. So the resolution impacts the memory bandwidth. The hardware scaling (if available) is available for compositing but for Android UI, not applications. Reply
  • blackmagnum - Tuesday, April 02, 2013 - link

    Very interesting. Will people pick their mobile devices only based on performance, or they are more concerned about looks. What is the optimum balance? Reply
  • xTRICKYxx - Tuesday, April 02, 2013 - link

    I think everyone wants a fast device; but as we all know, benchmarks don't mean everything. I would prefer the Nexus 10's dual A15 over the quad krait cores. Reply
  • alexvoda - Wednesday, April 03, 2013 - link

    Performance is not the only important factor.
    In addition to performance (CPU & GPU) features are very important (screen resolution, microSD, battery size&removability, radios).
    Also basic characteristics (device size) are probably the most important.
    And almost forgot, the software.
    Add to this the price and performance isn't such a major criteria any more.

    For example: I want a device that I can use single handedly so all 1080p 5" devices are way too big for me. As such I prefer a 4.2"-4.3" screen size. I don't really care about 1080p but I really want no less than 720p. I find MicroSD to be a must. I accept built-in batteries but I prefer removable ones.
    For my use case no phone available (at the time of purchase - December) is good enough so top performance is not as important since it isn't available. (I watch anime encoded in h264 Hi10P which can not be hardware decoded and must be software decoded. Not even the Exynos A15 Dual and the Snapdragon S4 Pro can handle it. Maybe just the new maybe just the Exynos Octa and Snapdragon 600).
    I currently have a Sony Xperia T which was very cheap when I bought it. It kinda satisfies those criteria but not perfectly. At 4.6" it is still too big, and the battery drains very fast.
    Right now the ideal device for me would probably be the Xiaomi M2S (http://www.engadget.com/2013/04/03/xiaomi-m2s-with... but I'm not so sure about the software and I doubt I could easily get service in case it breaks. I will probably wait for this years nexus and decide if it is worth sacrificing the microSD slot.
    Reply
  • alexvoda - Wednesday, April 03, 2013 - link

    link didn't work because of extra ")".
    Xiaomi M2S http://www.engadget.com/2013/04/03/xiaomi-m2s-with...
    Reply
  • UpSpin - Tuesday, April 02, 2013 - link

    Well written article. You tried to explain each anomalitiy in those charts this time, and give those charts much more descriptive text, much better than the T-Rex article.
    And as you said, it's really odd that the same hardware (Nexus 4 vs. Optimus G) on different Android versions scores so differently on this benchmark, so let's hope that all get an update to 4.2 soon or Futuremark fixes this glitch.
    Reply
  • ChronoReverse - Tuesday, April 02, 2013 - link

    It could be something to do with the graphics drivers as well as the Android version. Reply
  • jonup - Tuesday, April 02, 2013 - link

    Well there has to be something on the software level. N4 and OG have virtually the same hardware and 6 months down the road Quadrant still scores completely different on the two devices. We kept hearing that Quadrant is not optimized for 4.2. how long does it take them to fix the app. I am starting to doubt that hypothesis.
    p.s. I wonder how high my N4 would score in 3DMark. It is capable of high 23K in Antatu with some extra juice. :banghead I left the phone home. Any speculations?
    Reply
  • psychobriggsy - Tuesday, April 02, 2013 - link

    I guess here we can see a progression of the capability of ARM SoC GPUs over time too. The Mali-400 MP4 was available over two years ago in the Samsung Galaxy S2 for example. Might be interesting to see a graph of this progression over time to get an idea of how fast mobile graphics are improving. Reply
  • StormyParis - Tuesday, April 02, 2013 - link

    I just hope that more gaming performance tests will not distract the reviewers from the fact that games, thus graphics performance, are a very minor concern, as opposed to ... everything else.

    I'd hate finding myself in the same situation as for PCs, with realms and realms of graphics performance data, at the cost of more relevant stuff (reliability, sound quality, ...)
    Reply
  • LetsGo - Tuesday, April 02, 2013 - link

    Its going to be very interesting to see how well the IOS machines do in the multithreaded physic tests.

    Should shut up many a Apple Fanbois.
    Reply
  • Diagrafeas - Tuesday, April 02, 2013 - link

    As far as i can calculate Krait 300 @ 1.7GHz has the same performance in physics as my 8 year old Athlon 64 X2 4400+ @2.2GHz.
    It would be interesting to see how it does against NVidia 7800GTX of that era...
    I have one sitting around and if i find some time i will try to run a test...
    Reply
  • JPForums - Tuesday, April 02, 2013 - link

    Do you have some scores you can post?
    Given the multithreaded nature of this benchmark, they may be able to pull off an upset due to higher core count. I would love to get confirmation on it. The highend 7800/X1900 series graphics cards would be an excellent point of comparison as well given their console brethren. Perhaps the Anandtech crew would be so kind as to run 3DMark on a 2003, 2006, and 2009 era gaming rig as points of comparison to current SoC technology. The venerable Radeon 9700 Pro is just asking for one more round.

    I'm admittedly skeptical, though, given the massive difference between my system's score in ice storm vs the SoCs in the article (we're talking an order of magnitude or more). I'm not entirely convinced ARM SoCs have progressed quite as far as people tend to say. I see benchmarks here and there that are supposed to demonstrate the ARM SoCs prowess relative to the x86 PC realm. Then I see benchmarks that leave the two in different ballparks. I tend to fall back on the fact that the 5 year old Atom architecture is competitive with most SoCs out there (CPU performance), yet was relative trash compared to mainstream x86 processors even when it was released. I can see the progress ARM has made and would love to see them put more pressure on Intel outside of the mobile industry, but I think their single threaded performance holds them back at the moment.
    Reply
  • theduckofdeath - Tuesday, April 02, 2013 - link

    Well, the score comparison between the Razer Edge and the ARM's shows that it is a bit pointless to compare x86 with a GPU with ARM. The margin of error on the Razer Edge is probably somewhere around the total score on an ARM device. :) Reply
  • Diagrafeas - Tuesday, April 02, 2013 - link

    I can tell you the physics score of my Athlon 64 X2 4400+ overclocked @2.5GHz.
    It's 12600-13000 and 40-42 fps.
    Currently i have a ATI Radeon 4890. By old card 7800GTX is in a box at the basement...
    I was thinking about consoles too.
    Reply
  • jenesuispasbavard - Tuesday, April 02, 2013 - link

    Please do! I'm curious how today's phones hold up to yesterday's PCs. Especially something about on par with the Playstation 3 GPU. Reply
  • Diagrafeas - Wednesday, April 03, 2013 - link

    OK let me first say that there is something wrong with the drivers because of the GT1 low score.
    I think that the GPU is not used. Here it goes.
    3DMark Score:6104
    Graphics Score:
    Reply
  • Diagrafeas - Wednesday, April 03, 2013 - link

    Ooops accident press...
    Graphics Score: 5313
    Physics Score: 12763
    GT1 Score: 12.7 fps
    GT2 Score: 124.5 fps
    PT Score: 40.5 fps

    The card was overclocked to 486MHz from 432MHz...256MB GDDR3...
    So today smartphones have half the power of an 8 year old PC.
    4 cores help equal physics score against 2 cores but in general day to day use HALF the power.
    Reply
  • arthur449 - Tuesday, April 02, 2013 - link

    The LG Optimus G you have on hand is either still running ICS or has Eco Mode enabled. The one I have here with Android 4.1.2 pulls the following scores for Ice Storm (Normal) benchies: Graphics Test 1: 52.6 FPS. Graphics Test 2: 49.1 FPS. Physics: 8397 / 26.7 FPS. Ice Storm Score: 10746. Demo: 54.0 FPS.

    This lines up much closer to its sister the Nexus 4, which shares the same CPU/GPU.
    Reply
  • arthur449 - Tuesday, April 02, 2013 - link

    I suppose I should admit that this this Optimus G (LS-970) is running a trimmed down custom ROM (Imperium Initiative 2.0.1), and the ambient room temperature was 72ºF at the time of testing. I kept the phone in its rather insular Otterbox Commuter for the duration of the test, but it didn't get particularly warm, so I don't suspect any significant throttling occurred.

    Extreme Test results (in the same order as above): 30.3 / 23.0 / 8368 / 26.6 / 6413 / 25.8
    Reply
  • piroroadkill - Tuesday, April 02, 2013 - link

    Thermal throttling? Keep a can of "air" ready, invert that can and spray on the back. Mhmm.. chilled phone. Reply
  • xTRICKYxx - Tuesday, April 02, 2013 - link

    I'm patenting this idea first! Reply
  • RandomUsername3245 - Tuesday, April 02, 2013 - link

    All these new graphics benchmarks are very pretty and provide interesting statistics, but what's the point? Unless you carry around a bluetooth controller with your mobile device, is it possible to play a real 3d game? Most mobile games I see people play use very limited, single finger user interaction which would never allow you to play a serious FPS or driving / flying game. Maybe I'm finally getting too old and mobile gaming has passed me by... Reply
  • UpSpin - Tuesday, April 02, 2013 - link

    Car racing games like Need for speed, Asphalt ..., Real Racing, ... have high quality 3D graphics. You can control the cars with the accelerometer or touchscreen, works fine on a smartphone and makes fun.
    There are also many 3D jump and run games which are a joy on a smartphone.
    And even simpler 2D looking games can have stunning graphics with lots of lightning and particle effects, more realistic physics and 3D objects.

    With Miracast etc. you connect your smartphone to your HD TV and play on it, wireless. No need for a console any longer.

    Finally, it shows us that soon your smartphone will be as powerful as your desktop. So you might not need a desktop any longer, just a smartphone which docks, wirelessly, to keyboard, mouse, display, hard drive, ... The margin between smartphone/tablet and laptop/desktop gets smaller and smaller.
    Reply
  • pakotlar - Tuesday, April 02, 2013 - link

    As a datapoint:
    Note 2 @ 1800mhz CPU / 800mhz GPU, running Clean Rom ACE 4.5 (4.1.2), "AOSP" launcher
    Overall: 4391
    Physics: 11505
    Graphics: ~3770
    Graphics Test 1: 12fps
    Graphics Test 2: 25fps
    Graphics Demo: 25.2fps

    So this puts it ~ One X+ overall, and above Adreno 225/S3 AT&T on pixel shader heavy loads
    Reply
  • Mugur - Tuesday, April 02, 2013 - link

    Well... all I want to say is that I regret that OUYA don't have an Adreno 320 GPU instead of a Tegra 3... :-) Reply
  • ChronoReverse - Tuesday, April 02, 2013 - link

    What's interesting about this benchmark is that it agrees with benchmarks like Basemark's Taiji test or Epic's Unreal Citadel test in that Adreno does a lot better than traditionally given credit for in GLbenchmark.

    I'm glad there'll be another tool in the stable to provide a more balanced view of GPUs.
    Reply
  • Infy2 - Tuesday, April 02, 2013 - link

    The fastest physics score is around 11,000 while my Core7-3770K at 3.9GHz (4 cores+HT) scored only 52,000. That can't be right can it? If it is it would mean a mobile CPU at TDP of few watts scores 22% performance of a 77W TDP desktop CPU!

    The mobile GPU graphics score seems more plausible at 11,400 to my Radeon 7970 GE's score of 320,000, which makes it some 28 times faster.
    Reply
  • thecoldanddarkone - Tuesday, April 02, 2013 - link

    On my surface pro it got 24.5k and used 100 percent of the cpu. On my 3820@4.3 it never reached above 50% thread load. Reply
  • thecoldanddarkone - Tuesday, April 02, 2013 - link

    Look at that I had 55k with hyperthreading and 65k without. Yea... (all 4 cores were pegged) Reply
  • mwildtech - Tuesday, April 02, 2013 - link

    They must have a crappy Droid DNA? Mine scored much higher @ Total score of 8800 in standard Reply
  • Pentius_5123 - Tuesday, April 02, 2013 - link

    Pretty much the same here, I posted my stats and they seem to be 10 FPS under on everything... Reply
  • KompuKare - Tuesday, April 02, 2013 - link

    Can we have the SOC or better the vendor, corec count, GPU and speeds listed in the graphs please?

    Not everyone follows phone / tablet SOCs in so much detail that they can remember which is which; I know it would make the graphs a bit wide - or maybe have the device take up two lines?
    Reply
  • Gambit2K - Tuesday, April 02, 2013 - link

    I got 11502 on the Ice Storm with my Lg Optimus G E975 @ stock settings. Nothing tweaked in anyway oter than it's rooted and the bootloader is unlocked. Android 4.1.2. The scores on the the for the OG seems a bit on the low side. Reply
  • Pentius_5123 - Tuesday, April 02, 2013 - link

    Going to Cross-Post my stats from a similar Cnet article to provide a bit of insight to the HTC DNA (Internationally, Butterfly 'American Version')

    Ice Storm Score: 8768

    Graphics Score: 10262
    Physics Score: 5809
    Graphics Test1: 46.1 FPS
    Graphics Test 2: 43.2 FPS
    Physics Test: 18.4 FPS
    Demo: 49.8 FPS

    OS Version: 4.1.1
    1.5GHz Quad Core Krait
    Adreno 320 GPU
    2GB Ram
    Reply
  • Pentius_5123 - Tuesday, April 02, 2013 - link

    My phone is stock, I got it Saturday, so no junk running in the background, I'll have to get some averaged scores. Reply
  • XenIneX - Tuesday, April 02, 2013 - link

    HTC Droid DNA
    Purchased 3/26/2013
    Stock firmware (ver. 2.04.605.2 710RD); unrooted
    Ambient temp.: 68F

    ::Default::
    Ice Storm score: 10341
    Graphics score: 10558
    Physics score: 9646
    Graphics test 1: 46.5 FPS
    Graphics test 2: 45.4 FPS
    Physics test: 30.6 FPS
    Demo: 49.7 FPS

    ::Extreme::
    Ice Storm score: 5609
    Graphics score: 5490
    Physics score: 6069
    Graphics test 1: 28.1 FPS
    Graphics test 2: 20.8 FPS
    Physics test: 19.3 FPS
    Demo: 23.4 FPS
    Reply
  • Pentius_5123 - Tuesday, April 02, 2013 - link

    Jeeze, 2k points, I must be doing something wrong, lol. Reply
  • XenIneX - Tuesday, April 02, 2013 - link

    I think the physics test is prone to significant thermal throttling. If I really heat it up, I can drive that score down by about a third. Also, there may be something odd occurring with device orientation; it seems to post lower physics scores when it's in landscape than in portrait. Reply
  • Pentius_5123 - Tuesday, April 02, 2013 - link

    Thermal throttling, but it isn't too toasty here in the office, and the portrait Vs Landscape makes sense. Reply
  • ChronoReverse - Tuesday, April 02, 2013 - link

    I got it to run on my HTC One X (ATT version) on stock clocks. But the score results seem really different from the GS3 (ATT) results:

    Ice Storm: 5874
    Graphics: 5911
    Physics: 5750
    GT1: 27.7
    GT2: 23.9
    Physics: 18.3
    Demo: 28.5

    Ice Storm Extreme: 3358
    Graphics: 2999
    Physics: 5779
    GT1: 17
    GT2: 10.6
    Physics: 18.3
    Demo: 13
    Reply
  • JMC2000 - Tuesday, April 02, 2013 - link

    I wanted to run this in my 7" Galaxy Tab 2, but the Play Store says it's incompatible, though, oddly enough, I can install it on my T-Mobile Galaxy Exhibit, which I think has a MSM8255/Adreno 205... Reply
  • SetiroN - Tuesday, April 02, 2013 - link

    It's pretty obvious that Android 4.2.2 includes the latest blobs from Qualcomm, that HTC built into their version of 4.1.2. It's not really an OS version related difference, although 4.2 for the DNA/Optimus G will incidentally include the new blobs as well. Reply
  • Rontalk - Wednesday, April 03, 2013 - link

    How is Tegra 4 vs. Galaxy S4 vs. Iphone 5? Reply
  • kascollet - Wednesday, April 03, 2013 - link

    It's not available yet ! Reply
  • mitcoes - Wednesday, April 03, 2013 - link

    I miss a ·3dmark test for Ubuntu phone and even Ubuntu desktop. With Steam for GNU/Linux it would be great to have benchmarks even at the same machines with Ubuntu vs MS WOS 32 and 64 bits and Android vs Ubuntu phone. Benchmarking not only devices but OSs and drivers. Reply
  • Arbie - Wednesday, April 03, 2013 - link

    OK OK... but will they run Crysis? Reply
  • slatanek - Friday, April 05, 2013 - link

    Happy to see my HTC One SV going head-to-head with the One X!

    I wonder whats the reason for that? Is Tegra 3 such a weak performer overall or is there anything specific about these benchmarks which makes it underperform? Or is Qualcomms 8960 such an efficient chip?
    Reply
  • yvanize - Monday, April 08, 2013 - link

    I scored 11646 with my optimus g E973! Reply
  • DrChemist - Wednesday, April 17, 2013 - link

    Most remarkable thing is still the fact that the integrated intel HD4000 beats everyone of these hands down by huge margins. I await when they at least place the Gen7s to the intel atom mobile devices and decimate the competition. They just need to speed up the timeline for getting it out. Reply
  • kennet2000 - Sunday, October 27, 2013 - link

    Great article.

    Of course the scale up will matter, the devices with higher will have a disadvantage.

    The upscaling program probably requires 1-2 cycles for each pixel on the screen. So your GPU will spend time doing the upscale as well running the benchmark.
    The devices with higher resolutions has more pixels that needs to be upscaled, they will therefor spend more cycles on upscaling than the devices with lower resolutions. These cycles could be used for the benchmark but instead they are used for upscaling.

    On a big PC GPU this is probably neglectable since they have so much processing power.
    But remember that a mobile GPU often only have a few cores with a few execution units in them.
    So maybe this could make a difference.

    However it would probably not change the results that much but maybe a few percents.
    Reply

Log in

Don't have an account? Sign up now