Author Topic: Is Windows 7 Actually Faster Than Ubuntu 10.04? 1 of 2  (Read 1802 times)

Offline javajolt

  • Administrator
  • Hero Member
  • *****
  • Posts: 36193
  • Gender: Male
  • I Do Windows
    • windows10newsinfo.com
Is Windows 7 Actually Faster Than Ubuntu 10.04? 1 of 2
« on: May 03, 2010, 10:53:59 PM »

While Linux has long been talked about as being a faster operating system than Microsoft Windows, in 2010 is this still the case? It seems every time we deliver new benchmarks of the EXT4 file-system it's actually getting slower, recent Linux kernel releases have not been delivering any major performance enhancements for desktop users, the open-source Linux graphics drivers are still no match to the proprietary drivers, and "bloated and huge" is how Linus Torvalds described the Linux kernel last year. This is all while Windows 7 was released last year, which many view as Microsoft's best operating system release to date. Even after using it a fair amount the past few months in preparation for this about-to-be-shared work, it is actually not too bad and is a huge improvement over Windows Vista, but is it really faster than Ubuntu Linux? We have used six uniquely different systems and ran Microsoft Windows 7 Professional x64 and Ubuntu 10.04 LTS x86_64 on each of them with a set of 55 tests (actually, more than 165 if considering that each test is usually run at least three times for accuracy) per installation.

To see whether Linux still is a faster desktop to that of Windows, we have invested a great deal of time and energy in comparing the two vastly different operating systems. In November of last year we announced that the Phoronix Test Suite was coming to Windows 7. The Phoronix Test Suite is our extensible, automated testing framework that we use for conducting our software and hardware benchmarking at Phoronix and is also used by other publications, many software vendors, and practically every major hardware vendor as it can streamline their internal QA, validation, performance optimization, and regression management needs. Read the Phoronix Test Suite web-site for more information on all of its features from verifying statistical accuracy to automated regression testing.


We have been working on the software for years and it already works under Linux, Mac OS X, *BSD, and OpenSolaris operating systems, but Windows support was the last OS milestone there was to reach. It's not been only about being able to compare the performance of Windows to Linux and operating systems, but being able to offer the unique advantages that the Phoronix Test Suite provides to Windows users and while being open-source software -- from centralized test management to automated regression tracking (such as what we do on a daily basis for the Linux kernel) to collaborative testing to being able to autonomously find regressions in a code-base. After taking a sabbatical and investing hundreds of hours in this Windows endeavor, it has now come where we can begin comparing the performance of the modern-day Linux desktop to that of Windows 7 with nearly the same reliability, reproducibility, and ease-of-use that is provided under Linux.

With the benchmarking software in place, for the past few weeks we have been working towards this first article with comparing the performance between Windows 7 x64, Ubuntu 10.04 LTS, and Mac OS X 10.6.3. We did not just benchmark with a single system, but a wide array of hardware to try to get a good idea for where the performance of these vastly different operating systems really are running. As PC gaming is often viewed as one of the last Windows strongholds that are holding back more from switching over to Linux or Mac OS X, that was the first area we began exploring. This is especially important as Linux may soon have the prime opportunity to vastly expand its user-base once Valve's Steam Client and Source Engine come to Linux (it's already running), as there will be more of the popular tier-one titles available and one less reason to be dependent upon Windows. However, as we just had the Mac Mini with NVIDIA graphics for testing as a Mac with ATI graphics was not sent by AMD in time, we are holding off on publishing our initial Mac OS X 10.6.3 vs. Ubuntu 10.04 LTS vs. Windows 7 benchmarks. The first of these three-way OS benchmarks may be published as soon as later this week.


We had performed clean installations of Windows 7 Professional x64 and Ubuntu 10.04 LTS x86_64 one at a time onto each of these six test systems. Each operating system was left with their defaults and all standard Phoronix test practices were enforced. In regards to graphics drivers, we installed the latest proprietary graphics drivers for each operating system where applicable, which included ATI Catalyst 10.4 for the Radeon/FirePro hardware and on the NVIDIA side the 197.54 WHQL Windows 7 x64 driver (and the 197.16 notebook driver was needed for the ION setup) and on the Linux side the NVIDIA driver was 195.36.15.

Ubuntu 10.04 LTS was used as the Linux distribution to represent the modern-day Linux desktop since after all it is the most popular desktop distribution available and thus will have the largest relevance on the most users. It is also commonly the first Linux distribution that many Windows users end up trying. As previous benchmarks at Phoronix have also shown, the performance between Ubuntu, Fedora, openSUSE, and other tier-one distributions are not often too different at least with the gaming / OpenGL performance when equipped with similar packages.

With this first article looking at the OpenGL / graphics / gaming performance between the operating systems
we used the Nexuiz, OpenArena, Urban Terror, Warsow, Lightsmark, Unigine Sanctuary, Unigine Tropics, Unigine Heaven, X-Plane 9, and SPECViewPerf 10.0 benchmarks. The Phoronix Test Suite on Windows, Linux, and in most cases Mac OS X natively supports all of these test profiles too. With each of the games we ran them full-screen at 800 x 600, 1024 x 768, 1280 x 1024, 1400 x 1050, 1680 x 1050, and 1920 x 1080. For displaying the results, we have a line graph that shows how each of the systems and operating systems scaled at the different resolutions and an accompanying table to show the actual result numbers and to provide a more concise way of comparing the results. In addition, note that for any test where the system running Microsoft Windows was the winner is highlighted red and any time Ubuntu won, the number is highlighted in green. The process of generating these graph visuals were also provided automatically by the Phoronix Test Suite thanks to new features in the 2.6 "Lyngen" code-base.


Below is a listing (from the Phoronix Test Suite too) of the different hardware components that comprised each computer system. Also, note the identifier associated with each test system, as that is how the results are displayed in the graphs (i.e. the system bearing the Intel Core i3 CPU/graphics is called "CI3" and the AMD Opteron workstation is called "OPT").

ION:
Processor
: Intel Atom CPU 330 @ 1.60GHz (Total Cores: 4), Motherboard: ASRock AMCP7AION-HT, Chipset: nVidia MCP79, Memory: 2GB, Disk: 320GB Seagate ST9320325AS, Graphics: nVidia ION VGA 512MB (450MHz), Audio: Realtek ALC889A, Monitor: DELL S2409W

PX4:
Processor: AMD Phenom 9500 Quad-Core @ 2.20GHz (Total Cores: 4), Motherboard: ECS A790GXM-A v1.0, Chipset: AMD RS780 + SB700/SB800, Memory: 4GB, Disk: 250GB Seagate ST3250310AS, Graphics: ATI Radeon HD 5770 1024MB, Audio: Realtek ALC888, Monitor: DELL S2409W

AX3:
Processor: AMD Athlon II X3 425 @ 3.21GHz (Total Cores: 3), Motherboard: MSI 890GXM-G65 (MS-7642) v1.0, Chipset: AMD RS780 Alternate, Memory: 4GB, Disk: 250GB Seagate ST3250310AS, Graphics: ATI Radeon HD 4670 512MB, Monitor: DELL P2210H

CI3:
Processor: Intel Core i3 CPU 530 @ 3.32GHz (Total Cores: 4), Motherboard: ECS H55H-M v1.0, Chipset: Intel Core, Memory: 2GB, Disk: 64GB OCZ-VERTEX, Graphics: Intel Core IGP 256MB, Audio: VIA VT1708S, Monitor: DELL P2210H

CI7:
Processor: Intel Core i7 CPU 920 @ 3.60GHz (Total Cores: 8), Motherboard: ASRock X58 Super, Chipset: Intel 5520/5500/X58 + ICH10R, Memory: 3GB, Disk: 320GB Seagate ST3320620AS, Graphics: nVidia G92 [GeForce 9800 GTX] 512MB (675/1100MHz), Audio: Realtek ALC889A, Monitor: DELL P2210H

OPT:
Processor: 2 x Quad-Core AMD Opteron 2384 @ 2.70GHz (Total Cores: 8), Motherboard: TYAN S2932/S2932-E, Chipset: nVidia MCP55, Memory: 4GB, Disk: 165GB Western Digital WDC WD1600YS-01S, Graphics: ATI FirePro V8800 2048MB, Audio: ATI R6xx HDMI, Monitor: DELL P2210H


As you can see, across these six test systems used (not counting the Mac(s) to be used in future articles) there is quite a range of hardware with a mix of Intel and AMD processors as well as Intel, ATI/AMD, and NVIDIA graphics.

Any questions about our test procedures or benchmarking software can be directed to the forums. Without further ado, let us see how Ubuntu's 10.04 LTS release can run against Microsoft's proprietary Windows 7 operating system.


Right from the start the Nexuiz test was interesting. This multi-platform, open-source game that runs atop the DarkPlaces engine did the best at all resolutions on some systems with Windows 7 and other systems with Ubuntu 10.04. The NVIDIA ION NetTop with an Atom 330 processor was faster on Ubuntu 10.04 LTS when run at 800 x 600 and 1024 x 768, but when pushed to 1280 x 1024 and higher it turned to Microsoft's advantage by a small margin. In Microsoft's advantage was the AMD Phenom 9500 system (PX4) with the ATI Radeon HD 5770 graphics card where it narrowly beat out Ubuntu 10.04, but by two frames per second or less in each case. Meanwhile, the Athlon II X3 system (AX3) with Radeon HD 4670 graphics was slightly faster at each resolution, which is showing there is more or less feature parity between the Linux and Windows Catalyst drivers at least with this intensive open-source game. Results for the 1400 x 1050 resolution are not available as under Microsoft Windows 7 it would end up mode-setting to 800 x 600 rather than the requested resolution at this stepping for whatever reason with both the ATI and NVIDIA drivers.


Definitely in Microsoft's favor was the Intel Core i3 system (CI3) with its integrated Clarkdale graphics. As we reported early, the Intel Linux graphics driver is the real loser in these tests where it is easily and always clobbered by the Intel Windows. While the Intel Linux graphics driver is open-source and has made great strides over the past quarters with kernel mode-setting, a new 2D acceleration architecture (UXA), in-kernel memory management (GEM), a migration to DRI2, and other improvements, its performance is a joke. The Clarkdale system was playable at 800 x 600 through 1280 x 1024 with Nexuiz, but even at 800 x 600, the open-source Intel graphics driver used by Ubuntu 10.04 LTS put out just six frames per second.

When running at 1920 x 1080, the Clarkdale i3 530 graphics were 11.35 times faster under Windows 7 than in Linux! Ubuntu 10.04 LTS is using a stabilized Intel graphics driver stack and not the very latest bleeding-edge DRM/Mesa, but still, it is unlikely we will see Intel's Linux graphics driver recoup any of these major performance losses anytime soon as they are not even committed to switching to the Gallium3D architecture that is faster than the classic Mesa DRI drivers.

Meanwhile, the Intel Core i7 system with the NVIDIA GeForce 9800GTX running Ubuntu 10.04 beat-out Windows 7 each time (ignore the highlighting on this system's run, it was a coloring bug in PTS) and by a rather significant margin. Lastly, the lone workstation system in this Windows vs. Linux testing was about even with this OpenGL game.


Except for a few cases where Ubuntu 10.04 LTS was the clear winner (namely the Intel Core i7 system again), Windows 7 won more often. This may be due to OpenArena being better optimized to run on the Microsoft operating system with its ioquake3 engine, but this is what you get with the "out of the box" experience provided by this open-source game. Many of the systems were faster by more than twenty frames per second with Windows 7 than Ubuntu 10.04.


While Windows 7 was faster near consistently with this CPU-limited game, the system that produced the most interesting results was the AMD Opteron workstation. With the octal-core Opteron setup that boasted AMD's brand-new FirePro V8800 graphics card that was released last month, its performance was severely hampered under Ubuntu Linux. At 1680 x 1050, for example, the Opteron workstation running Microsoft Windows 7 was 3.37x faster than when it was running Ubuntu 10.04 with the equivalent proprietary driver. The FirePro V8800 may be a brand new card, but it should be well supported on Linux considering that is where many AMD customers use the hardware. When running at 1920 x 1080, the FirePro V8800 setup with Ubuntu was 3.11x faster than when it was run at 1680 x 1050, and even faster than when the game was run at 800 x 600. It seems there may be a few bugs hampering the new FirePro card under Linux with select consumer OpenGL applications.


The results from Urban Terror, another ioquake3 game, were most surprising. With every system and at every single resolution, Windows 7 was faster than Ubuntu 10.04 and by usually a wide margin. Within the Windows version of Urban Terror, occasionally there would be a texture that would turn white or other graphical artifact, but like OpenArena, it heavily favored running under a Windows environment.


All wins for Windows 7 x64.


Next, we have the results from Warsow, which is another open-source game but it uses the Qfusion engine. The results here are quite mixed, but still most frequently we are seeing Windows 7 taking the gold. At 800 x 600 where the performance would be heavily CPU limited, Ubuntu 10.04 had the best results except for when using the crippled Intel Linux driver. Strangely, the AMD Opteron + FirePro workstation could not run this game without locking-up under Windows 7, while Ubuntu 10.04 could just fine, but for the lacking of Windows results its numbers were removed.



Lightsmark is a cross-platform OpenGL demo that has been rather good for benchmarking since 2008 and supports more features than many of the open-source games thanks to its real-time global illumination, penumbra shadows, and other lighting features. However, as sad as it may be, Lightsmark 2008 on Windows was unimaginably faster -- with 100+ frames per second differences. Ubuntu 10.04 only came out ahead once and that was with the Athlon II X3 system and its Radeon HD 4670 graphics when running at 1920 x 1080 where it had an odd performance spike similar to the FirePro setup.


Worth noting is that similar to the OpenArena numbers, the dual AMD Opteron "Shanghai" setup with its FirePro V8800 graphics was rather sluggish with its proprietary ATI driver until it was running at 1920 x 1080, the same resolution as the X.Org Server. It is looking like this may indeed be a Linux-specific bug within the proprietary AMD graphics driver stack when running at the X.Org Server's standard mode.

source:phoronix 
continued