Jan 05 AT 7:36 PM Taylor Wimberly 51 Comments

Intel Medfield vs NVIDIA Tegra 3 performance preview


Next week at CES we expect a few surprises, but the biggest story in Android land is likely to be the introduction of Intel’s mobile processors for smartphones and tablets. We’ve heard this same line from Intel the last couple of years, but it looks like 2012 will finally be the year that we see some Android devices with Intel inside go on sale in stores.

In the last couple of months we have seen Intel boasting that they beat the competition hands down, proclaim they are years ahead of the competition with their process technology, and show off a prototype phone. Last week someone from Intel leaked some early performance numbers from the Java benchmark CaffeineMark to VR-Zone.

*Medfield numbers come from leaked report

According to sources inside Intel, their Medfield single-core processor running at 1.6 GHz was able to score 10500 in the CaffeineMark benchmark. We tried the same benchmark on a couple other devices and found NVIDIA’s Tegra 3 was able to produce similar results with a score of 10399. This benchmark appears to be single-threaded, so Tegra 3 should have the advantage in applications that take advantage of multiple cores.

Keep in mind this is just a single benchmark and we are comparing rumored numbers, so I wouldn’t draw too many conclusions from this. Having said that, it looks like Intel’s Medfield should be competitive in browser performance with the latest ARM-based processors from NVIDIA, Samsung, Texas Instruments, and Qualcomm.

The main concern with Intel’s mobile processors has been their power consumption. We have heard reports all over the place from saying it came in 2nd or 3rd compared with the current generation of mobile devices, to saying it was the best.

With CES just days away, we won’t have to wait much longer to see how Intel really performs. Rumors suggest we could see devices from both LG and Samsung, but nothing concrete has leaked out just yet.

Hopefully Intel fares better with Android phones and tablets than they did with Google TV.

Via: VR-Zone

Taylor is the founder of Android and Me. He resides in Dallas and carries the Samsung Galaxy S 4 and HTC One as his daily devices. Ask him a question on Twitter or Google+ and he is likely to respond. | Ethics statement

    Most Tweeted This Week

  • aykutb

    i admit, i love competition between these guys, they take technology one step further in the name of profit :)

    • damambt

      the best part is, we all win

      • mac08wrx

        Yeahhh this is great for Android

    • http://mihai.discuta-liber.com/ tmihai20

      This looks promising. If Intel is able to deliver the Medfield SoC to compete against Qualcomm and Nvidia, we couldn’t be any happier. A real bomb would be if Google would choose Intel Medfield as the base for their Google tablet.

      • honourbound68

        hmm does anyone think intel will one day be subjected to anti-monopoly laws if they start dominating the mobile chip arena like they do desktop/server chip arenas and start driving companies out of business?

  • Oskar Wismierski

    A lot of terms like dual core, quad core and others are mainly used to market phones and such and sell them to people who do not know much about those things. They hear DUAL CORE and go all crazy wanting the thing where as they would be perfectly fine with a simple single core device for half the price..

    Its like when new iPhones come out.. I mean come on.. what difference did the 4s make? It’s just same thing with siri on it but idiots who don’t know anything about technology want to pay shit loads of money to “upgrade” from iphone 4..

    It’s sad how big companies brainwash people :S

    • nick

      Well thats not entirely ture, the A5 CPU in the 4S is considerably more powerful that the A4 in the 4. The GPU on the A5 is pretty much unrivaled, its an absolute beast and smashes the performance of the 4 out the water…


      • Oskar Wismierski

        Yes it’s true, but what Im trying to say is that most people don’t really need all that speed and they are just paying extra for the “S” after 4..

        • kazahani

          In the iOS world, you are correct. You don’t need a dual core A5 to run iOS smoothly.

          However, an Android phone is pretty much a Linux computer in your pocket. It’s much better able to take advantage of high-end tech.

          They will keep making better and better tech to power Android until they have pushed it to the upper limit of performance and capability. I think that’s still a year or two away.

          • mustybooks

            Hopefully android will always be pushed onwards as well. So either hardware or software are always playing catch up, driving innovation. :)

    • aranea

      As far as I know dual-core actually makes a difference in battery consumption as under the same load CPUs can run at a lower clock speed and use less power. That said if medfield is not better at battery life than tegra 3 I would go for tegra 3 as the numbers are very similar and as metafor points out this test maybe for single thread any way. That means on multi threaded situations tegra 3 (and maybe even tegra 2) will beat medfield.

      Yet competition is always better for us, the consumers. :)

  • Nate B.

    Even if its true or not the difference in numbers arent huge. Very close

    • kazahani

      You do realize that this is a single core chip that is able to keep up with the 4+1 core Tegra3? That’s impressive.

      • wuzelwazel

        I can find no information anywhere on whether this benchmark supports multiple threads. I expect, based on the other numbers on the list, that it is single threaded.

        In that case you’re looking at the Java performance of a SINGLE Tegra 3 core.

        • Josh

          I think you’re right. Look at the comparisons with Tegra-2. It must be single threaded. Medfield supports hyperthreading also. Based on this single benchmark, Medfield seems underwhelming–it’s slower per clock than the ARM A-9′s and by the time handsets launch, it will be competing with Krait and OMAP5.

  • scores87

    But how is battery life? and If I overclock tegra 3 to 1.6ghz I think tegra would win, but I like intel so GO Intel GO

    • scores87

      Holy crap I didn’t notice it was single core, just imagine the performance in dual or quad core

      Too bad this is x86 and current apps aren’t compatible

      • metafor

        If you’ll notice, there isn’t a significant jump going from dual to quad (Tegra 2 to Tegra 3). T3 is in fact, performing roughly 30% better than T2. Which is expected since it is running at a 30% faster clockspeed. That would suggest this is a single-threaded benchmark.

        • sunrise

          It is possible the benchmarking software used doesn’t take full advantage of 4 cores, thus the results could be not truly representative of the power behind tegra-3.

          I don’t know if this is true, but just a consideration to ask yourself.

      • bacalou

        These Intel CPUs are not x86.

        • Tarwinia

          Actually they are. Intel doesn’t have an ARM license of any type (it used to when it made scale processors but it sold that division to Marvell). It is x86 qndo if I remember correctly an atom variant/derivative. So yes, current apps are incompatible. That may change with time but seeing that the only x86 compatible android apps would be Google TV apps (it’s the only commercially available x86 variant of of Android), any device with medfield had better come loaded with a good set of apps for it to be truly worthwhile in the beginning.

          • Adnoxaei

            Since the apps are written in Java shouldn’t they run just fine? The incompatibilty would come in if Android and its compiler wasn’t modified to work on x86, but they did modify the OS. Making displays and digitizers and things like that play nice with intel seems to be the thing that would be most inconvenient.

      • http://sean-the-electrofreak.blogspot.com/ Sean the Electrofreak

        Actually, there is a version of Android that will run on x86, and the Dalvik VM should handle Java app compatibility issues.

  • Brett Isbell

    I want the one with the bigger GB’s

    • kidphat

      +1 to infinity.

  • Alexander drzfr3shboialex

    Single-core!!! That’s crazy i’d take a quad-core with a big battery and just buy a spare, there problem solved :D

  • mustybooks

    Got krait and the exynos 4212 to consider as well this year. We shall have an excellent selection of mobile processors. Three cheers for competition!

  • masterpfa

    I’m not too concerned how many cores or what the numbers are or who makes it, all I want is a phone that is noticeably faster than the current batch of Android Smartphones and all other OS’s smartphones currently available.

    As long as I have a totally smooth UI, multitasking, browsing, gaming and greatly improved user experience without any lag and blazingly fast performance with decent battery life to boot (18hrs plus please!), I’ll be happy

  • Nathan D.

    Beside the tegra 3, they are comparing old processor, wait until the other newer ones come then the number should be closer or better but then again these are rumors (so far).

  • classic_hero

    I feel like the current scores should be way higher for quad core processors, idk cant really complain I guess

    • http://androidandme.com Taylor Wimberly

      Yeah i dont think this benchmark is multi-threaded. Higher clock speed equals higher score.

  • lokidokie

    If anything, this solidifies the top spot for Tegra 3.

  • Hall Lo

    Another choice :) Better for us consumers!

  • cxandroid

    To soon to tell, based on an unknown test, however, competition is good.

  • Louis A

    And let the competition begin! And the winners are the consumers.

  • spazby

    The future is bright and exciting…

  • http://genesischess.com/ MJM128

    Now lets see a battery life comparison.

  • Tarwinia

    Taylor is right. Caffeinemark is not multi-threaded. I just tried it out and got better scores on my 1.5ghz single core Flyer than on my 1.2ghz dual core Sensation. As such then ARM cores will most likely perform better in real situations. But then again this is only one benchmark. You need to look at a variety of benchmarks to get a good idea odd an SoCs real potential and even then it doesn’t necessarily translate into real world performance. But IF that benchmark gives a good indication of performance then we’re better off with other processors. I say it because of the power consumption. If those benchmark numbers are legit then I assume the power consumption numbers are legit, which means too power hungry a solution for being a single core chip.

    • dpleus

      Thanks for the clarification. I was wondering how a single core could out preform a quad, and since I wasn’t familiar with the software being used to test the phone, it was nice to get a more in depth explanation. And thanks Taylor for pointing out that Caffeinemark is not a multi-thread capable.

  • nivekkev

    Based on the info that is out at the moment, looks like they still have a little work to do…

  • http://youtube.com/user/jawckamoe Marcus

    Is it just me, or does it look like the higher clock speed is the higher score? Is this benchmark accurate? I don’t really understand…

  • Joe

    I just ran Caffeine Mark benchmark on my stock rom (rooted) Galaxy S2 – Exynos, and got 9255 @ 1.2 GHz, overclocked to 1.4 Ghz = 10672!!. So the Exynos figures are a bit low. I’m on the latest 2.3.6 Firmware – XWKK5, which really has boosted performance, especially in Sunspider 0.9.1, I now get 2015ms @ stock 1.2 Ghz, bye bye iPhone 4S:)

    Long story short, until all these devices are on Android 4.0, these results can be easily skewed by the performance differences of Gingerbread vs Honeycomb vs Ice Cream Sandwich. I bet you Intel was running ICS on their device.

  • supercaliber64

    single core vs. quad, and the single beat it? Sounds like they are using more power to achieve this. I agree with Joe, 2.2 vs. 2.3 vs. 4.0 is huge.

  • Darkseider

    Meh. Single threaded benchmark showing a lower clocked T3 equaling a Medfield in performance. This really isn’t all too impressive. With that taken into account it would be fair to assume that multi-threaded operations would suffer on the Medfield and shine on the T3 and possibly even the T2. Lastly I have my doubts about Intel’s claim to power consumption. The fact that they have to crank it up to 1.6 Ghz JUST to equal a 1.3 Ghz Cortex A9 means this thing is drinking juice. I hope I am wrong because competition is good but the numbers just say otherwise.

    • Darkseider

      OK. Ran it on my G2X OC’ed to 1.2 Ghz and scored a 7916. My Transformer OC’d to 1.424 Ghz scored a 10210 and my Transformer Prime stock 10286. So looks like the Medfield while being a nice single core may suffer greatly if it has to deal with multi-threaded applications.

  • fm123

    Price was not mentioned. Intel is traditionally much higher in price. Current Atom processors cost over $40, and they were rumored to want $50-70 for the Oaktrail Atom used on current Win7 tablets. Can Intel really match the ~$30 and lower price I would expect from ARM solutions? Will they be willing to accept such lower profit margins?

  • french toast

    I think people are missing the point here, although we dont have battery consumption numbers , we can speculate that it uses significantly more power than tegra 3 when it is also on a much more avanced manufacturing process AND is only a single core.

    If you take that benchmark, as Metafor points out it is single threaded, and the atom is clocked 300mhz higher yet tegra 3 matches it…likely consuming far far less power.
    If you run multithreaded benchmarks/apps the tegra is on another level, likely alot more smoother on other things like web browsing as that can be multi threaded also.

    The other chips coming out in a couple of months, namely exynos quad core 4412, duel core 5250, Krait, apple A6, and perhaps the most powerfull of all st errickson nova thor.
    These chips will blow the atom away…

  • X-PAT


  • Moosa Mahsoom

    Soon intel broadwell will be there on smartphones, i wonder what ARM has in stock for that, ARM is years below Intel, definitely.

  • jadin

    Nvidia tegra 2 and 3 are SMP multicore processors, it doesn’t matter if the application is single threaded or not, the nvidia tegra cores share the load equally amongst all cores, and where possible, it will offload tasks to the gpu to be completed. the reason the results don’t seem to be consistent with the core increase in the tegra parts is because the CPU’s dynamically adjust the core frequency to use as little power as possible.

    so, this really is a comparison between a single core processor and multicore processors.

    Remember, this is Intel we are talking about. just look at how well they compare in the desktop market.
    pricing aside, Intel desktop cores are far more efficient than AMD’s.
    i don’t know how AMD are doing now, i haven’t been following tech too much, but, I’m sure the latest i3 is as powerful, if not more powerful than AMD’s quads.

    I would expect nothing less from Intel with their mobile processors.

  • Ben

    Jadin: “it doesn’t matter if the application is single threaded or not, the nvidia tegra cores share the load equally amongst all cores”

    “so, this really is a comparison between a single core processor and multicore processors.”

    Do you mind of i slightly, debunk you claims. The Nvidia Tegra’s shares the load equally amongst all cores, IF:

    a) You are running multiple application at the same time
    b) Running a single application, that is optimized for multi thread.

    The benchmark mentioned above, is a single application, with a single thread. In other words, it will maximize the usage on a single core, and that is where it stops.

    I do not know where you get the idea that hardware can magically split software instruction to multiple cores, when its not optimized for it.

    So, this comparison is actually a Singly Core/Single Thread vs a Multi Core / Single Thread.

    You can say, that 3/4th of the Tegra’s power is not used, as only 1 core of the 4 is stressed.

    This is part of the problem Android is facing right now. The move to dual / quad core, has actually gone so fast, that the software is lagging behind.

    Even today, in the PC world, you still find plenty of software that is single thread. New compilers, that try to optimize the software are helping, by offloading parts to multi thread ( if possible ).

    But, it does not help, if you software’s main functionality is single threaded.

    “and where possible, it will offload tasks to the gpu to be completed” .

    And this has to do exactly with this benchmark? This benchmark is a CPU test, not a GPU test. It will not magically offload threads that are run on a CPU, to the GPU. Those two are totally different animals.

    “Remember, this is Intel we are talking about. just look at how well they compare in the desktop market.”

    And lets look how badly they perform in this mobile market? This is the 3 or 4th attempt at entering this market.

    It took them this long, to finally come up with a solution, that starts to cache up to the old generation ARM. Because this is what we are talking about. The Tegra3 is running a Cortex A9. This is a slightly improved A8. The Tegra3 is actually not a massive improvement. Its just a Tegra2, where they doubled the amount of cores, added NEON instruction again, and clocked a bit higher. That’s it. How old is the Tegra2 design…

    For Nvidia this is just a step over, for the next Tegra4 ( A15 ). Its main focus will be Mid to low level. Even Nvidia told us, they expect the Tegra3 to do, 200$ tablets.

    As of now, most manufactures are moving to Krait / A15. They are already starting to sell Krait Dual Cores. And they even kick the Tegra’s behind, and just keep up with a dual core vs quad core, when all cores are maximized.

    Its not a bad attempt, and it will allow Intel to finally get some footing in this market. But do not expect it to suddenly turn into a 90% market share monster. Unless Intel plays dirty again, by offering low prices, to damage competition. Like they did with AMD. And in the end, the only people that got hurt by this, is we the consumers ( those lower prices where not given to the consumer, just more profit for the system sellers ). And in the end, it crippled AMD, because it lowered there profit, as they where not able to gain a large market share.

    To be honest, Intel’s presence can have a large negative effect on the market, then a positive. As Intel is a Company, with a big War Fund Chest. For a few years, this can be beneficial for people, until one after another ARM manufactures put down the books. Mark my words: 2017, Intel will have destroyed multiple companies.

    And for good reason. As ARM is getting better and better, there is the danger, of ARM systems entering the PC market. HTPC with ARM are already here. Every HTPC with ARM SOC = one less CPU that intel can sell. And if this moves more into the Laptop market…

    The notebook market with Intel and there Atom, is starting to die. Tablets with keyboards are moving into this area. Again, Intels market. They are not dumb. If they want to keep dominating the CPU sales, they need to gain access in this mobile market, before the mobile market enters too much into the PC ( and related ) market.