CPU Performance

The big news with Tegra 3 is that you get four ARM Cortex A9 cores with NEON support instead of just two (sans NEON) in the case of the Tegra 2 or most other smartphone class SoCs. In the short period of time I had to test the tablet I couldn't draw many definitive conclusions but I did come away with some observations.

Linpack showed us healthy gains over Tegra 2 thanks to full NEON support in Tegra 3:

Linpack - Single-threaded

Linpack - Multi-threaded

As expected, finding applications and usage models to task all four cores is pretty difficult. That being said, it's not hard to use the tablet in such a way that you do stress more than two cores. You won't see 100% CPU utilization across all four cores, but there will be a tangible benefit to having more than two. Whether or not the benefit is worth the cost in die area is irrelevant, it only means that NVIDIA (and/or its partners) have to pay more as the price of the end product to you is already pretty much capped.

SunSpider JavaScript Benchmark 0.9.1

Rightware BrowserMark

The bigger benefit I saw to having four cores vs. two is that you're pretty much never CPU limited in anything you do when multitasking. Per core performance can always go up but I found myself bound either by the broken WiFi or NAND speed. In fact, the only thing that would bring the Prime to a halt was if I happened to be doing a lot of writing to NAND over USB. Keyboard and touch interrupts were a low priority at that point, something I hope to see addressed as we are finally entering the era of performance good enough to bring on some I/O crushing multitasking workloads.

Despite having many cores at its disposal, NVIDIA appears to have erred on the side of caution when it comes to power consumption. While I often saw the third and fourth cores fire up when browsing the web or just using the tablet, NVIDIA did a good job of powering them down when their help wasn't needed. Furthermore, NVIDIA also seems to prefer running more cores at lower voltage/frequency settings than fewer cores at a higher point in the v/f curve. This makes sense given the non-linear relationship between voltage and power.

From a die area perspective I'm not entirely sure having four (technically, five) A9 cores is the best way to deliver high performance, but without a new microprocessor architecture it's surely more efficient than just ratcheting up clock speed. I plan on providing a more thorough look at Tegra 3 SoC performance as I spend more time with a fixed Prime, but my initial impressions are that the CPU performance isn't really holding the platform back.

A Lesson in How Not to Launch a Product Tegra 3 GPU: Making Honeycomb Buttery Smooth
POST A COMMENT

204 Comments

View All Comments

  • thunng8 - Thursday, December 01, 2011 - link

    It would be pretty hard to beat the visuals of Infinity Blade II on the ipad2. The game looks amazing. Reply
  • TrackSmart - Thursday, December 01, 2011 - link

    Anand,

    Just a heads-up, as I know you had to get this review out the door quickly: It would be really useful to put the weight of the tablet AND keyboard dock prominently on the first page (maybe in the chart?). This would make it easier to compare the total travel weight to ultraportables and netbooks.

    I'm guessing that we're talking about 2.6 pounds from what I could find online. And you'd end up with some serious battery life (18 hours?) compared to even the longest running Zacate (or Atom) laptops/netbooks. If only the software and hardware capabilities were as good...
    Reply
  • biassj - Thursday, December 01, 2011 - link

    Asus looking good but tablets can't replace x86 PCs yet. I hope Asus launches a 15" Zenbook with 1080 display and maybe even a quad core Ivy bridge. I don't mind if it's thicker but I want something to compete with the rumor Macbook Air 15". Reply
  • HighTech4US - Thursday, December 01, 2011 - link

    Will the ASUS Eee Pad Transformer Prime be retested here when ICS (Ice Cream Sandwich) is released for it?

    My understanding is that there are improved graphics drivers and other improvements in the OS that will improve on the results shown here.
    Reply
  • vvk - Thursday, December 01, 2011 - link

    Hi Anand,

    It seems that the results of the Ipad 2 and Transformer Prime are quite different in GLBenchmark 2.0.3 Egypt (tested by PCworld) vs. what you have found based on the vs. 2.1 of the same benchmark. Could you comment what may be the reason for the differences?

    "The most significant result in our gaming performance tests: The Transformer Prime logged 53 frames per second, the highest frame rate we've seen on the GLBenchmark 2.0.3 Egypt test with no antialiasing. This result topped the Apple iPad 2's previous record of 46 fps, and it just crushed the Android masses we've tested, which averaged 18 fps and topped out at 34 fps (for the 7-inch Acer Iconia Tab A100)."
    https://www.pcworld.com/article/245256/asus_eee_pa...
    Reply
  • vvk - Thursday, December 01, 2011 - link

    Also wanted to add that other sites show much longer battery life for the Transformer. I understand that the battery life depends on use but could be that your unit is a dud in not only in case of WiFi but also battery, so I wonder if you would be willing to repeat the battery test? Reply
  • Anand Lal Shimpi - Thursday, December 01, 2011 - link

    I'm already on it :) Got a new Prime in this morning and battery life is looking much better already.

    Take care,
    Anand
    Reply
  • metafor - Thursday, December 01, 2011 - link

    The tests in the article were done using an off-screen resolution of 1280x720. This allows all devices to be tested at the same resolution -- thus measuring the relative rendering performance of the SoC.

    Looking at the results on GLBenchmark's own website:

    http://www.glbenchmark.com/result.jsp

    Correlate to what is posted in the article.

    The ones posted by users of Tegra-3 devices:

    http://www.glbenchmark.com/phonedetails.jsp?benchm...

    Also correlate with the results of the article. I suspect that version 2.0.3 may stress different parts of the GPU or that PC World got some pretty bad results.
    Reply
  • vvk - Thursday, December 01, 2011 - link

    Well I usually do not play games off-screen :) Anyway, I am still puzzled what is causing the huge difference between the 2.0.3 Egypt vs. 2.1 (assuming both test were done correctly).
    Also while comparing Ipad 2 at theoretical 720p has merits in synthetic testing in practice you can not see more than 1024 by 768 pixels on the Ipad 2.

    I am also wondering if GLBenchmark better than lets say 3dMark for desktop? 3D Mark scores do not always correlate perfectly with real games experience due to optimizations for particular card and also because synthetic testing is not a perfect substitute for real games to start with. So I am not overly concerned about the scores more about what have changed between the GLBenchmark versions.
    Reply
  • metafor - Thursday, December 01, 2011 - link

    Absolutely. But unfortunately few games for mobile if any offer time-tests and there is no equivalent framerate tool like FRAPs to measure framerate in real games.

    So synthetic benchmarks will have to do for now.

    As for 2.0.3 vs 2.1, GLBenchmark's website doesn't give all too much info except that it includes the off-screen mode as well as "high quality" versions.

    But it does mention that 2.1's throughput tests -- which I don't think impact Egypt -- perform warm-up loops in order to cache a lot of its data before measuring GPU throughput.

    Other than that, I'm not really sure what changed.
    Reply

Log in

Don't have an account? Sign up now