Benchmarking Ubuntu Linux On The Google Nexus 10

Written by Michael Larabel in Computers on 26 February 2013 at 02:27 PM EST. Page 1 of 7. 7 Comments.

After spending the better part of the past week running continuous open-source Linux benchmarks on the Exynos5-powered Google Nexus 10, the first extensive benchmark results for the Nexus 10 tablet running the Ubuntu Touch Developer Preview are now available. This performance comparison from Ubuntu on the dual-core ARM Cortex-A15 powered device is compared to numerous other ARMv7 and x86 devices. One of the interesting findings from this new round of ARM Linux testing is that the Google Nexus with its dual-core ARM SoC is competitive with AMD's first-generation Phenom Quad-Core processor for some demanding workloads.

The Google Nexus 7 and Nexus 10 tablets have served as my latest platforms for Linux ARM benchmarking after previously having built PandaBoard clusters, benchmarking the Calxeda ARM servers, been involved with building a solar-powered ARM super-computer, and numerous other ARM Linux performance efforts in recent years. With Ubuntu Linux now coming to more consumer ARM devices, there's new possibilities opened up for benchmarking this low-power hardware thanks to a full Linux stack and tool-chain being available on the devices.

Since last Thursday's release of the Ubuntu Touch Developer Preview for tablets and smart-phones, I've been busy torturing the Nexus 7 and Nexus 10 tablets with a plethora of computationally intensive workloads. I purchased the Nexus 7 and Nexus 10 tablets solely for this Linux benchmarking purpose. In Benchmarking The Google Nexus With Ubuntu I shared some of the shortcomings. The main hurdles I ran through were the Google tablets running uncomfortably warm, the WiFi adapters ultimately going down after hours of use, and the Nexus 10 losing its charge even when using the AC power adapter.

As then shared in Benchmarking Ubuntu Touch Yields Mixed Results, the Nexus 10 with its Exynos 5 Dual SoC (Cortex-A15) has been running great but the Nexus 7 with its NVIDIA Tegra 3 (Cortex-A9) has been much less stable. As of Tuesday morning, I still don't have any complete test results for this Google tablet. Problems have unfortunately plagued this Nexus 7 testing and thus those results have to be delayed to a later article in March. It's some combination of Ubuntu Touch and the Nexus 7 under heavy load as when using the NVIDIA Tegra 3 "Cardhu" reference tablet with Ubuntu Linux, the same tests have no problems running without fail.

I also shared some of my other early Ubuntu findings on the Google Nexus devices within Ubuntu Touch/Tablet Is Using SurfaceFlinger and My Favorite Command For Ubuntu Touch/Tablet. For those hoping to see Ubuntu benchmarks from the Google Nexus 4 with the Qualcomm Snapdragon S4 Pro SoC, see this article note.

With no Nexus 7 results to share today, the Ubuntu Linux x86/ARM performance results being shared today include:

OMAP4460 - PandaBoard ES: One of the common "PandaBoard ES" development boards that features a Texas Instruments OMAP4460 SoC, which bears a dual-core ARM Cortex-A9 1.2GHz processor. There's 1GB of RAM and a 16GB SDHC Class 10 card for storage.

Tegra 3 - Cardhu: NVIDIA's reference Tegra 3 tablet that is similar in specifications to the Nexus 7. The Tegra 3 SoC bears an ARM Cortex-A9 quad-core 1.4GHz processor.

Exynos 5 Dual - Chromebook: Results are available from the Samsung Series 5 Chromebook that normally runs Google's Chrome OS. Ubuntu can be easily loaded on this Chromebook and with its Exynos 5 Dual SoC the performance is very good. This Chromebook is very similar to the Nexus 10 components. The Exynos 5 Dual features a dual-core 1.7GHz ARM Cortex-A15.

Exynos 5 Dual - Nexus 10: The focus of today's testing, the A15-based Google Nexus 10 with Ubuntu! Similar to the Series 5 Chromebook, the Nexus 10 is backed by an Exynos 5 Dual 1.7GHz dual-core Cortex-A15, 16GB of flash-based storage, and 2GB of system memory.

Agena - Phenom 9500: Various x86 systems were also tossed into the mix while running Ubuntu Linux for reference... First up is an old AMD Phenom 9500 "Agena" Quad-Core. This was the first-generation AMD Phenom processor, comprised of four AMD K10 cores, and was released back in 2007. The AMD Phenom 9500 is a 2.2GHz quad-core 64-bit part with a 95 Watt TDP.

Diamondville - Atom 330: The Intel Atom 330 dual-core processor clocked at 1.60GHz with Hyper Threading. The Atom 330 was released in H2'2008 with an 8 Watt TDP, built on a 45nm process, and supports SSE3.

Arrandale - Core i3 330M: The Intel Core i3 330M from the Arrandale/Ironlake days was tested. The Core i3 330M from 2010 is a dual-core part plus Hyper Threading and is clocked at 2.13GHz while having a 35 Watt TDP.

Sandy Bridge - Core i5 2520M: For showing some more modern Intel x86 hardware, a Core i5 2520M "Sandy Bridge" mobile processor was tested into the testing mix as well. The i5-2520M is dual-core with Hyper Threading and has a base frequency of 2.5GHz with a maximum Turbo Frequency of 3.2GHz. Its TDP is also 35 Watts.

This is the selection of hardware that was tested for this article. The latest Ubuntu Linux releases for all the respective platforms were used. This means Ubuntu 13.04, but for the Ubuntu Touch preview on the Nexus 10, the version string was reflected as Ubuntu 12.10. Due to the currently fragmented ARM Linux state with different kernels for different SoCs and the support not being fully unified yet, the tested versions of the Linux kernel were also different between x86 and ARM. The default kernels shipped for each platform target/image on Ubuntu for the given hardware were used.


Automated and reproducible benchmarks are made possible via the Phoronix Test Suite.

There's also obviously some differences with the integrated flash storage and system memory on the different devices that can't be interchanged, but the tests being shown in this article are mostly all CPU focused. A best effort was done to minimize the variables between the system hardware/software. All benchmarking was done in a fully automated and reproducible manner using the open-source multi-platform Phoronix Test Suite benchmarking framework with integration on OpenBenchmarking.org.

Coming up with the Google Nexus 7 (and hopefully Nexus 4) results in March will also be results from the Tegra 2 Trim-Slice, other Intel Atom hardware, and potentially some other systems/devices too. Your feedback is welcome within the forums, emailing me, or @MichaelLarabel on Twitter.


Related Articles