21
Feb
msm89601

The de facto hardware experts over at AnandTech have spent some time with the newest CPU from Qualcomm: the MSM8960, sporting the new 28nm Krait architecture in dual-core, 1.5GHz form. They've put it through their suite of benchmarks including Linpack, SunSpider, BrowserMark, Vellamo, and Basemark OS. In a nutshell: the MSM8960 absolutely destroys every other phone/CPU on the market right now in every. single. benchmark.

msm89601

An example of what to expect.

The Krait architecture supports 1, 2, and 4 CPUs, and the smaller transistor size (28nm) means that it should outperform most (if not all) comparative 40nm CPUs and offer better battery life. It remains to be seen how well Nvidia's first-to-the-market quad-core Tegra 3 CPU holds up in comparison, too.

Devices packing Krait-powered CPUs should begin shipping sometime in Q2.

[Source: AnandTech]

Aaron Gingrich
Aaron is a geek who has always had a passion for technology. When not working or writing, he can be found spending time with his family, playing a game, or watching a movie.

  • SeCreTEnDinG

    Now I wish I didn't just buy the Galaxy 2.

    • Edd

      Don't worry! Still a great phone! Feel your pain tho...

  • The Genius

    Can't beat quad core though

    • PeterF

      Actually, it can. If the quad core chips are just 4 of the cores that are in the dual core chips then this is better. These two cores are about as strong as 4 of the old cores. Like comparing an old Intel Core 2 Quad Q6600 with a new dual core i5-2410m. The i5 is faster and uses less power, too.

      • Chris

        Umm, you're comparing a CPU released in 2011 to one released in 2006?

        Even a low end Sandy Bridge will outperform a high end Kentsfield.

        We are talking about a SOC that has not been released yet versus one a year old. A more accurate computer analogy would be a Sandy Bridge dual core versus a Nehalem (or Westmere) based quad core.

        • Joshua

          I find it amusing that the top level comment was completely incorrect, the next comment corrected him but got something else wrong, and then the next comment corrected that!

          Sadly, I do not see any errors in your statement to correct.

  • Mike

    And as always... The Qualcomm chips support 1X advanced which allows you to have SVDO on Verizon's 3G network... Which I think is a huge selling point but no one ever mentions it... Like ever! Lol

    Half of my commute in NJ is 3G.. and I live in a 3G area. SVDO on my Thunderbolt is a god send. Its the reason I finally left AT&T (I dealt with the dropped calls because the 3G data was better). Between 4G LTE and SVDO on 3G... Qualcomm chip based phones on Verizon can basically do voice/ data nationwide. No other phones can claim that.

  • Chahk

    I don't care if a phone can benchmark six quadrillion mega-quadrants. As long as manufacturers keep sticking anything less than 3000 mAh batteries on them in the name of thinness, I'll keep passing on.

    • xhazyx

      Well with the 28nm transistor size, they should do better in terms of battery life too

      • devilEYE

        theoretically yes, but in practice will be as usual - 24h on one charge.

        • Pyrotek85

          I'd be happy if we could actually get a real full day of usage out of a smart phone, I'd have no problem charging it every night while I slept. As it is, you can't get anywhere near that if you actually use it to it's full potential.

    • Jon Garrett

      I have a 2500 mAh battery in my new Galaxy Note and I'll tell you right now that its a beast !! Ive had it for 2 days now and it powers its 5.3" screen/1.5Ghz processor all day and all night.

  • xhazyx

    Now imagine a Quad-core Snapdragon manufactured with this Krait architecture..... My computer is scared right now

    • PeterF

      Why is your computer scared? A sandy bridge processor will run laps around any of these ARM things currently out or coming out in the near future.

  • con

    lower power consumption and better battery life or gtfo

  • Marc

    Looks great on paper, just wondering how the real life performance will be.

    The voice in the back of my head is telling me that this is almost too good to be true. The reasoning for this is that Qualcomms greatest supporter, HTC, went with the Tegra 3 for its "flagship" phone for 2012, the HTC Edge/Endeavour/One X over the "Krait". That says a lot to me right there.

    Unless of course, if the Tegra 3 was further along in development and the "Krait" would have delayed the One X. Am I over thinking this?

    • infra_red_dude

      This is not "on paper". These are actual numbers. 8960 compared with Tegra3: http://www.pcmag.com/article2/0,2817,2400409,00.asp

      Tegra3 "quad-core" gets beaten everywhere (well, almost).

    • wvr

      HTC One S/Ville uses S4 and will be launching at the same time as HTC One X. Both are flagship products. One X is getting a lot of hype because of the quad core..

  • xhazyx

    A lot of companies like to tout that they sport the latest and greatest in processor power... So HTCs decision must of been based in just that, the fact that Tegra 3 are actually shipping, and these S4s are just getting ready for manufacturing.

  • Nik

    I would like to see tegra 2 in this list.

    • Golightly

      There are a few on that list.

      My Atrix is currently netting scores of twice the Galaxy Nexus on Quadrant, modded of course. I'd like to see what this new processor could do when tinkered with.

  • slurms mckenzie

    just wait till samsung comes out with theirs, also dont forget about omap 5. no reason to get all excited when the competition hasnt even came out with theirs yet.

    • Adam

      Yeah that makes sense...let's not talk about something because another something similar isn't here to talk about yet. Yes, let's wait and forget Krait was even mentioned.

      Jeeeeezus. Why the heck wouldn't people talk about it? Are you thick in the skull?

  • petikeke

    This list isn't so good...no tegra 3 in it. :S Would like to see a comparison with it.

  • Darkseider

    Me thinks this article from Anandtech is a bunch of hokum. My G2X OC'd to 1.2 gets:

    1089 in Vellamo.
    1802 in Sunspider.
    47.9 in Linpack Single Thread and 79 in multi.
    102564 in Browsermark.
    Basemark OS: N/A (can't get the APK)

    Overall the G2X at 1.2 Ghz compares pretty favorably. The Linpack results are highly suspect though only because of the HUGE gap in results between other notable dual core SoCs not just the Tegra2, OMAP 44xx or Exynos 42xx. Me thinks that aside from a few hundred Mhz and a better power consumption due to die size the Krait's performance is not all that spectacular when compared to other units.

    • dalingrin

      Darkseider,

      I don't understand why the results are "suspect." Read up on the architectural differences before you come to a conclusion. These benchmarks simply reinforce what many have predicted based on the huge architectural change with Krait. That said, of course Linpack isn't representative of real overall performance. It is a single subsystem benchmark.

      This is why I laughed at every article that acted like the devices announced at CES with dual core krait were some type of a let down. There's a reason Intel considers Qualcomm their only competition.

      • Darkseider

        Sorry I still see the scores as suspect regardless of any change. From what I see yes it is an improvement but nothing as drastic as one would hope based on Anandtech's results. His results are also questionable seeing that many folks have already ran the same benchmarks on their devices, like myself, and have shown a HUGE difference in what is published in his article vs. what is actual. With that said I will withhold further judgement until folks have some real time with devices based on the new Krait SoCs to see if they really live up to the hype.

        • Dalingrin

          If you want to remain skeptical that a Cortex A15 like architecture will be faster than Cortex A9, by all means.
          But, I don't find the results "suspect." Anandtech is about as reputable as you can find anywhere. The scores they list are with stock devices. From what I can tell all the scores they posted are what should be expected for these devices running their *stock* configuration.

          Check their architecture overview, they predict Krait to be about 25-30% faster than Cortex A9 at the same clock speed.
          Anandtech themselves state that Vellamo is likely a biased benchmark and to not put much stake in it.

          Look at the browser based benchmarks which are more well rounded CPU benchmarks that feature integer and floating point computations. They show a 25-30% increase over Cortex A9. Almost exactly the general performance increase Anandtech predicted based on what they knew about the architecture.

          Since Linpack is a very focused benchmark, it is always very sensitive to architectural changes. For instance, look at the Linpack performance of the Nexus S vs the Nexus One. Nexus S is a Samsung architecure vs Nexus One being a previous Qualcomm architecture. Does the large disparity make it a suspect result? No, it is reflective of the pipelined FPU in the Qualcomm architecture. Back to Krait's fast Linpack scores, there are a few reasons for this. Krait can issue 3 Neon instructions at a time vs 2 on Snapdragon, the memory bandwidth is double Snadragon's, the cache is not only faster but also about double the size found in Snapdragon, etc. All of this in addition to clock speed increases versus the FPU performance versus Qualcomm's last architecture MSM8660.

    • Darkseider

      Again show me the wow here? Based on your response:

      "Look at the browser based benchmarks which are more well rounded CPU benchmarks that feature integer and floating point computations. They show a 25-30% increase over Cortex A9. Almost exactly the general performance increase Anandtech predicted based on what they knew about the architecture."

      OK let's look at browser based benchmarks.

      1802 in Sunspider
      102564 in Browsermark.

      These are both from a G2X running CM7 and OC'd from 1.0 Ghz to 1.2 Ghz. Where the Krait @ 1.5 Ghz does:

      1532 in Sunspider
      110345 in Browsermark

      Based on this alone and the browser benchmarks showing a more rounded result if I were to OC my G2X to 1.5 Ghz I would be hitting about the same numbers as the Krait. Especially when noting that Vellamo, even according to Anandtech, should be taken with a grain of salt.

      Take it one step further. Even with the faster memory, extra instruction issue and larger cache the system will still be bottlenecked by the speed of its' local storage and MicroSD flash storage. Granted it may load an app slightly faster or perform certain tasks quicker but in real world use again... I don't see it. Until the quad core A15's hit with the A7 as the "Little" then the rest of this year will be just about static in the world of dual and quad core SoCs.

      • Dalingrin

        Again, your G2X running CM7 is not stock. There are several things done in CM7 to improve performance over stock Gingerbread.
        I'm sure after we get through optimizing CM9 the MSM8960 will be that much faster.

        This is the first iteration of Krait which is running at a slower clock rate and with a last generation GPU. The architecture is known to scale to 2.0GHz+.

        The fact that Krait is around 20-30% faster clock for clock than Cortex A9, runs at a significantly reduced voltage and still has room for an integrated radio with LTE means its a big deal for phones. Not only because of the increased performance but because of the better battery life and radio handling.

        Either way my argument was with regards to your use of the word suspect..not so much the merit of Krait versus other future architectures.

  • Darkseider

    Another note. Looks like Vellamo is a single threaded benchmark and is clueless about the cores. So in essence it shows nothing substantial. I can say this because based on testing my G2X and Transformer Prime the difference in speed is:

    913 G2X @ 1 Ghz

    1089 G2X @ 1.2 Ghz

    1407 T-Prime @ 1.4 Ghz

    The difference is approx. 29%. The difference in speed alone is all that is needed for this to show that. So again. Vellamo is nothing special. I would like to see something like Antutu, CF Bench and Nenamark added to the mix.

  • poway guy

    Let me choose how many cores are active. Do i want battery life or processing power? Let me decide.

  • Awesome Sauce

    If these come out Q2 we will prob see this on a sprint phone in Q4 of 2013. Thanks for losing a billion on the iPhone Dan Haase. DB.

  • Alex

    Nice figures, now let's come up with something which makes use of that power.

  • Grweg

    Wow I am really glad I got the Gnex instead of that slow ass Rezound, wow that is a slow phone

    • Tyler C

      You're comparing a phone with Gingerbread and super bloated HTC Sense to a phone running stock ICS....

  • cannon

    Qualcomm CPU has the weakest performance compare with same level CPUs of other brands. (Always!)

    Samsung Exynos>TI OMAP>Nvidia Tegra>Qualcomm Snapdragon.

    Samsung A15 based Exynos will blow S4 ass off.

    • DCMAKER

      and when will that see the light of day?

  • DCMAKER

    Isn't it fishy how the 8660 beats the exynos? i think this test is weighted in qualcoms favor.

Quantcast