Comments Locked

26 Comments

Back to Article

  • douglaswilliams - Thursday, February 23, 2012 - link

    "Javascript performance can be multithreaded at times but most of the benchmarks we run don't scale incredibly well beyond two cores. Making matters worse is the fact that SunSpider performance regressed on the Eee Pad Transformer with the latest update to ICS. I've included the old Honeycomb results as a reference for where things should be. Keep in mind that the Honeycomb browser on the Eee Pad Transformer was very heavily optimized for Tegra 3. It's possible that the same degree of optimizations just aren't present in the ICS version yet."

    This was taken from the Qualcomm Snapdragon S4 (Krait) Preview Part II article. I wonder how much closer the comparison would have been had the browser in Honeycomb been used.

    I don't know if this it totally applicable though because I don't know if BrowsingBench uses Java or if any of the pages loaded had Java content.

    Can someone comment on this to help me understand/learn more?
  • Brian Klug - Thursday, February 23, 2012 - link

    This is the browser in ICS though which is an evolution of the browser in Honeycomb, and ends up using an improved version of V8/Webkit.

    Actually going forward the most threaded browsing experience will be Chrome on Android, which from the testing I've done before is very nicely multithreaded.

    EEMBC BrowsingBench isn't a JavaScript test like Sunspider/Browsermark, just straight loading pages, timing, and making sure they're loaded fully/compliant.

    -Brian
  • gamoniac - Friday, February 24, 2012 - link

    The browsers are multi-threaded but javascript does not support multi-threading until HTML 5's web worker feature. All javascript snippets on a pre-HTML 5 page run in one thread, although the browser could load images/files with multi threads. I doubt these benchmarking tools run in HTML 5, do they?
  • watersb - Thursday, February 23, 2012 - link

    JavaScript is not Java. SunSpider doesn't measure Java performance.

    I couldn't tell from your comment if you were asking about Java as a separate from JavaScript or conflating two.

    I'm typing this on an iPad: Safari on the iPad tablet doesn't support Java at all, right? And while Android native apps are written in Java, they are compiled to a not-quite-Java binary. I don't think the Android WebKit browser supports Java, either.

    So: Javascript performance is one of the things being tested in this demo. But not Java.
  • MantasPakenas - Thursday, February 23, 2012 - link

    Has anyone else noticed that it's actually Honeycomb being used on the TI platform? (back and home buttons give it away). Given that much more development time has been put into optimizing Honeycomb and it's browser (as witnessed by worsened JS performance on Transformer Prime's 4.0.3 browser), could it be on purpose that TI didn't do an apples to apples comparison?

    I wonder what the result would be if Chrome Beta was used and both platforms were running ICS. On my Transformer Prime, Chrome Beta scores way better than stock browser did on Honeycomb (unfortunately, there's no way to compare Chrome on Honeycomb), and it's a better threaded browser.

    So this video is suspicious already. Add the fact that there's no way to tell if Tegra3 is not running PowerSaver mode, which would immensely cripple performance in the scenario being benchmarked, and it's hard to take the video seriously.

    That said, I hope OMAP5 performance will blow Tegra3 out of the water, progress FTW! :)
  • JMC2000 - Thursday, February 23, 2012 - link

    The weird thing is that the TF Prime actually loaded the pages faster than the OMAP5432 device, but was hobbled by horribly slow page transitions.

    The Prime loaded page one faster, fell behind on page 2, then somehow caught up by page 10.
  • hechacker1 - Thursday, February 23, 2012 - link

    It looked like it was getting hung up on networking performance. If you notice, they are both running off a cached local IP server.

    So maybe it's CPU starved to process packet information in time. Or TI's networking is just better.
  • jjj - Thursday, February 23, 2012 - link

    Got to wonder what kind of storage and connectivity TI is using to get such results.
  • Brian Klug - Thursday, February 23, 2012 - link

    EEMBC BrowsingBench works over local network, and you can actually see in the video that they're loading pages over a 192.168.1.x subnet. For example, right here: http://www.youtube.com/watch?feature=player_detail...

    No doubt they're using TI WiLink on their own reference design, and the Transformer Prime is some other similar single spatial stream 802.11n.

    -Brian
  • dagamer34 - Thursday, February 23, 2012 - link

    I would hope TI's platform beats nVidia's, at best it comes out at the end of the year!!!
  • aryonoco - Thursday, February 23, 2012 - link

    The thing is, this is comparing apples and oranges. Tegra 3 was itself delayed but it was finally launched at the very end of 2011 and is in products on the shelves now. OMAP 5 probably wont be on shelves at the very end of this year or early next. There is a year's difference here. If TI wants to be sincere it should compare the 4460 to Tegra 3, not something coming out next year. Im sure Nvidia wont be sitting on its laurels for a whole year while TI launches OMAP 5
  • eddman - Thursday, February 23, 2012 - link

    That's the thing actually. Tegra 3 was never meant to compete with Krait or A15 based chips, but to best A9 and scorpion chips and consume less power at the same time.

    On the other hand, while there are Tegra 3 products already shipping and selling, there are no A15 chips available and no Krait based gadgets AFAIK.
  • infra_red_dude - Thursday, February 23, 2012 - link

    But the point here is that nVidia has nothing better to offer in 2012. Tegra3 has to compete against Exynos 5000 series and Krait this year, whether that's a fair comparison or not.
  • eddman - Thursday, February 23, 2012 - link

    Very true but, I believe Tegra 4 would launch around the same time as other A15 chips form TI and even samsung and also quad kraits (late 2012), so not much to worry about there for nvidia in that respect.

    The real threat are dual-core kraits which will be available much sooner, and are apparently better than Tegra 3 in every aspect, although a shrink to 28 nm might help it to close the gap.
  • Dribble - Friday, February 24, 2012 - link

    They will be in different markets. The Krait using cutting edge 28nm will cost significantly more then tegra3 using cheap and plentiful 40nm.
  • dcollins - Friday, February 24, 2012 - link

    Not necessarily. Tegra3 chips are larger than the MSM8960 and require a separate baseband chip for LTE. Krait is ahead of schedule so yields at 28nm must be pretty good so far. Altogether, I doubt the difference in price will be large enough to significantly effect the price of the phone.
  • eddman - Friday, February 24, 2012 - link

    Where did you read that krait is smaller? I couldn't find the die size anywhere, or maybe I missed it.
  • Black1969ta - Thursday, February 23, 2012 - link

    Kal-El+ is slated to launch Q3ish and Wayne launching by Jan 2013, so it is still possible that nVidia will top Krait, Exynos, and OMAP 5; at the least it will make it much more competitive.
    If Wayne is Delayed like other things this year, then the above three might enjoy dominance, for awhile.
    nVidia has not flashed everyone with their blazing, dominating performance, they have made a name in the SoC market by bringing new technology first (dual-core and Quad-core ie...), not dominating in performance marks.
  • jcompagner - Thursday, February 23, 2012 - link

    Yes exactly we compare here a A9 core against a A15
    And OMAP is again very late, i expect to have dual core A15 before that (Samsung?)
    And in 1 year there will be already Tegra4 which is i guess a A15 with the next gen GPU.

    I am still hoping for Samsung to have a dual core A15 with the latest Mali GPU. The current one still tops the charts when we compare phones even after 10 months...
  • SydneyBlue120d - Thursday, February 23, 2012 - link

    Brian, considering the rumor that the Google Phone 2012 will be built from Motorola, do You think the timeline can be right to be used in it? I really hope the Google Phone running Android 5.0 will use OMAP 5 or Snaprdagon S4 1,7Ghz! Thanks.
  • iampedro - Thursday, February 23, 2012 - link

    Tegra 3 only has one 32 bit channel which limits the processes I hope this has atleast 2 32 bit or 1 64 bit and 1 32 bit.
  • SilentSin - Thursday, February 23, 2012 - link

    Can't really tell from the abstraction in the layout diagram, but is that showing the 3G/4G radios on a separate piece of silicon and not integrated like the MSM8960?
  • scook9 - Thursday, February 23, 2012 - link

    TI has never mated mdoems and CPU on one soc. Really only qualcomm does that currently. Intel and nvidia may be doing it soon as well though.
  • fukat - Thursday, February 23, 2012 - link

    Awesome Job ... The video needs more captions/blowouts ... especially the completion time etc
  • scook9 - Thursday, February 23, 2012 - link

    While I am eager to see A15 based products hit the market and am very optimistic, this came across as marketing garbage to me..... They offered little information into their test settings. For example, what governors were used? The TI tablet is essentially and MDP and like Qualcomm most likely had the fastest governor in use. The Asus Transformer Prime (that is absolutely a prime - you can tell by the Asus skin of the android buttons - I can confirm as I am typing this on my prime....) has 3 out of the box governor settings and on boot defaults to the middle one - aka not the fastest - this does make a tangible difference in use of the tablet.

    My final observation is that unless proven to me, I have no way of knowing they are even running on the same wifi. Call me paranoid, but this could have been made much clearer and credible as a demo.

    And of course, as others have mentioned, the A15 based stuff is going to actually get out to consumers a year later than Tegra 3, at which point I am confident that Nvidia will have an A15 based Tegra part out as well
  • Rits - Monday, February 27, 2012 - link

    This is going into the next Nexus. Something tells me that,

    The OMAP platform has been under-credited for a while and its TI's time to really shine. OMAP has to adopted by more manufacturers, Tegra has been lacklustre always, yet gets the headlines.
    That SGX 544MP2 must beat the Adreno 320 for a start. But then again, there's samsung lurking in the corner, waiting to push the button on something mindblowing again. Its strange how Samsung manages to outdo all these expert chip makers time & again. Its been doing so for the last two generations at least and I expect no different this time around.

Log in

Don't have an account? Sign up now