Windows 7 Apps in Linux: Wine; Cameyo/ThinApp; VM vs. Native

I was in the process of transitioning from Windows 7 to Linux. I had identified a substantial number of Windows programs for which I would be seeking Linux alternatives. As described in a separate post, I did not find good Linux replacements for some of those programs, and therefore planned to continue to use those Windows programs. The question was, in what form would I continue to use them?

There were several possibilities. I wanted to reduce my dependence on Microsoft, and anyway I expected a virtual machine (VM) to yield worse performance than a program running directly on a native installation, so my first choice was to run those Windows programs in Linux using Wine. But I was not sure that Wine could accommodate and store all of the additions and adjustments that a person might make when installing a program, adding plugins, doing updates, and so forth. For some programs, it seemed it might be better to created a virtualized (i.e., portable) version of the program, with all those tweaks built in, and then run that package on Wine. For purposes of virtualizing software, I had the free Cameyo virtualizer, and had also downloaded the 60-day free trial of VMware’s ThinApp.

If Wine did not work well with either the program itself or with a virtualized package of that program, then the alternative would be to run that program, or virtualized package of it, in a Windows VM. Finally, there would be the least convenient but most familiar option of running the program in Windows natively, in a dual-boot system.

Contents

Testing Environment
Adobe Acrobat Pro 9
Adobe Premiere Elements 12
Cool Edit 2000
CyberLink PowerDirector 12 Ultimate
Microsoft Office 2003
Conclusions.

Testing Environment

The following sections of this post describe what I encountered with each of several different programs. In most cases, I compared the program’s basic functioning in my native Windows 7 x64 installation against its functioning in Wine and in a Windows VM. In the two latter settings, I considered both the program and its virtualized counterpart.

The programs tested here would not necessarily be those that would matter most to others. For example, Cool Edit 2000 and Microsoft Office 2003 were ancient, but still useful for my purposes. The latter was an especially good candidate for virtualization, in that its full installation included a variety of tweaks, add-ins, and updates. Regardless, I felt that this set of programs offered a fair sense of how various installation settings would compare.

Note that I was not a professional tester, was not funded for this investigation, and was not always impressed by theoretical benchmark testing. More power to the funded professionals using recognized theoretical benchmarks; they had much to offer. My goal was more in the direction of obtaining a rough, real-world sense of how these options would compare for my purposes.

My native (dual-boot) Windows 7 installation was not a clean machine; it was the cluttered Windows 7 system that I had been using for some time. It was possible that a clean machine would have yielded better performance. But I did not think that was going to matter for present purposes. The installed sizes of programs installed in that native system could be higher than the sizes that would appear after a new installation. I had been using those installations for some time. They may have included cached files not added to the other installations.

The computer running these tests used an Intel Core i7-4790 CPU with 16GB RAM. I used a fresh, clean Windows VM for each VM test. I tested Windows 7 VMs for all programs, and Windows XP VMs for several. Each VM ran in VirtualBox 5.0 on Linux, and had 4GB RAM with PAE enabled, two CPU cores, and 128MB of video RAM allocated to it. The Linux and Windows operating systems and the VMs were all located on solid state drives (SSDs). The Cameyo and ThinApp packages were all 32-bit packages, created in barebones Windows XP x32 VMs. No doubt some programs would have run faster if they had been virtualized in 64-bit Windows 7. But it appeared that they would have been less likely to run at all.

As the following discussion indicates, my testing tended to look at the questions of whether the program ran at all in the specified setting; whether it seemed to be functioning acceptably, within my sense of how it normally ran — that is, whether its menu options seemed to be working normally; and whether it seemed quick or slow. In most cases, I timed its performance on a specific task. Except where otherwise noted, after having been run once to work through initial (e.g., license) questions, programs would start within one or two seconds. Obvious or redundant information has been trimmed out. Times are reported as 0:00 (i.e., minutes:seconds).

Some of the information provided below was collected before I thought of producing the present post. This yielded some variations that would not be present in a more carefully prepared research project. Also, unlike the native Windows 7 and the virtualized (Cameyo and ThinApp) installations, the installations in the VMs were limited to the basic program installer — that is, tweaks and updates were not included.

It appeared that both Cameyo and ThinApp packages could probably be shrunk to some extent, for those who knew how to do that, and that shrinking might improve their functionality. For me, the prospect of learning how to edit and improve a virtualized package was secondary to the question of whether the packager was capable of producing a working package. (Note: some of the results reported here were reported previously in various posts on Cameyo, ThinApp, and Wine.)

Adobe Acrobat Pro 9

Tasks tested:

  1. Combine 30 text JPGs into a single PDF, using File > Create PDF > Merge Files into a Single PDF.
  2. OCR that combined PDF, using the default options within Document > OCR Text Recognition. Test by searching for a few words; report any problems.

Native Windows 7 Installation

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 2.1GB (at C:\Program Files (x86)\Adobe\Acrobat 9.0).
  • Task performance: (1) 0:08. (2) 0:21.

Linux Wine Installation: Program Installer

  • Did installation or packaging appear to succeed? No.
  • Comments: This was what the CrossOver database results would predict for Acrobat 9. (See other post.) The CrossOver database said some other versions (especially Acrobat 8) worked pretty well in Wine. (See also the WineHQ database page.) I tried Acrobat 8. It installed and seemed to run, but brief use produced several error messages that may or may not have stemmed from an activation issue.

Linux Wine Installation: Cameyo Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 1.8GB.
  • Task performance: Unexpected “Cannot convert file type to PDF” error when attempting to convert JPGs. File > Open did not work.

Linux Wine Installation: ThinApp Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 3.9GB.
  • Comments: Program failed to run.

Win7 x64 VM Installation: Program Installer

  • Installed size: 1.9GB.
  • Task performance: (1) 0:10. (2) 0:22.

Win7 x64 VM Installation: Cameyo Package

  • Task performance: Aborted shortly after startup with licensing error message.
  • Comments: See other post.

Win7 x64 VM Installation: ThinApp Package

  • Task performance: (1) 1:35. (2) 0:51.
  • Comments: Excellent look and feel. This package was several weeks newer than the Cameyo package. It was not clear whether the Acrobat license would expire in ThinApp as it had apparently done in Cameyo.

Windows XP x32 VM Installation: Program Installer

  • Installed size: 1.9GB.
  • Task performance: (1) 0:12. (2) 0:25.

Windows XP x32 VM Installation: ThinApp Package

  • Task performance: (1) 1:42. (2) 1:01.

Impressions

  • Installation success and performance: my tests of Acrobat set the theme for most of the other tests (below), and most of the outcomes were predictable. As a general rule, the Windows 7 native installation worked in every case, and was the fastest; the Cameyo and ThinApp packages were least likely to work, and would typically be the slowest options in any setting; programs were more likely to work in a Windows VM than in Linux via Wine, but were faster in Wine than in a VM if they did work there; programs were more likely to run, and to run somewhat faster, in Windows 7 VMs than in Windows XP VMs; thus a Windows 7 VM would tend to be the best alternative to the native Windows 7 installation. The following discussions focus on exceptions to these general guidelines.
  • Installed sizes: with Acrobat and some others, the Cameyo package was nearly as small as, if not smaller than, the native Windows 7 installation, whereas the ThinApp packages could be twice as large as the Cameyo packages, or even larger. It was not always clear whether the problems experienced by Cameyo packages were due to excessive shrinkage. In the case of Acrobat, most installations came in around 2GB, whereas the ThinApp package was a bloated 3.8GB. This could presumably be reduced, with the investment of time, effort, and perhaps some trial and error to learn how to edit ThinApp packages.
  • For Acrobat, ThinApp produced a seemingly functional package whose activation might or might not persist over an extended period. Where the ThinApp and Cameyo packages failed, it was not clear whether tweaking these packages would resolve the issues I observed, whether there would be other issues, and whether the packagers had flagged those issues when assembling the package. The ThinApp packages tended to be much larger, and yet did not necessarily run any better, though they may have offered more editing options over the long term.
  • Appearance: all Acrobat sessions in Linux had graphic quality inferior to that of Acrobat in Windows.
  • Regardless of whether Acrobat was installed or was run from a virtualized package, there appeared to be potential issues of appearance, functionality, and licensing for Acrobat in Linux. Further exploration would be required to determine whether those issues could be overcome.
  • Of the installation approaches tested here, the only one that worked immediately and had been working for more than a month was in the native Windows 7 system accessible via dual-boot.
  • In the case of Acrobat, within the Windows 7 and XP VMs, the installed version ran much faster than the ThinApp packaged version. This difference would be more noticeable with large projects. The ThinApp version felt very responsive when browsing and editing the 30-page PDF produced in the test.
  • Depending on the task, the VM installations ranged from being slightly to visibly slower than the native Windows 7 installation.
  • The ThinApp package completed the first task roughly 7% slower, and the second task 20% slower, when it was running in the Windows XP VM than when it was running in the Windows 7 x64 VM. Although the measurements were not precise enough to say for sure (and of course these two tests would not necessarily be indicative of the program as a whole), it looked like Acrobat, installed in the VMs, might run 10-15% faster in the Win7 VM than in the WinXP VM overall.

Recommendations

  • The foregoing results suggested that installation in a Windows 7 VM would be the best alternative — and not a bad alternative — to a native Windows 7 installation of Acrobat 9.
  • If a bit of tinkering with Acrobat 8 in Wine on the Linux desktop was sufficient to make it work, then that would probably be faster than Acrobat 9 in a VM.
  • If Acrobat’s license remained functional in ThinApp after a period of a few months, then it would be feasible if slow to run the ThinApp package in a Windows 7 VM. If the license sputtered out, it was possible that the Cameyo package could be made to run in a VM.
  • There did not appear to be much prospect of success for Acrobat 9 via Wine.

Adobe Premiere Elements 12

Tasks tested:

  1. Import and finish conforming a 500MB .mp4 video (The Hunchback of Notre Dame).
  2. Export the first three minutes of that video in NTSC DVD Standard format (720 x 480 MPEG-2).

Native Windows 7 Installation

  • Installed size: 1.0GB.
  • Task performance: (1) 0:44. (2) 0:27.

Linux Wine Installation: Program Installer

  • Did installation or packaging appear to succeed? No.
  • Comments: The WineHQ AppDB reported no success in efforts to run earlier versions of APE on Wine. The executables files in the installer did not respond to my attempts to open them in Wine. This failure was consistent with all four entries in the WineHQ and CrossOver databases. Various searches failed to produce countervailing success stories.

Linux Wine Installation: Cameyo Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 1.5GB.
  • Comments: Program failed to run.

Linux Wine Installation: ThinApp Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 5.9GB for all files; 2.9GB for the essential executable files.
  • Comments: Package failed to run.

Win7 x64 VM Installation: Program Installer

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 940MB.
  • Task performance: (1) 0:23. (2) 0:28.
  • Comments: Attempting task no. 2 repeatedly froze the entire Linux system, such that I had to use the computer’s power button to restart. Disabling 2D and 3D acceleration in the VM enabled APE to complete task no. 2, after producing a message: “We have detected generic Microsoft’s display driver on this machine. To get a better and faster playback performance, please update your display driver.”

Win7 x64 VM Installation: Cameyo Package

  • Comments: Package failed to run, with or without 2D and 3D acceleration.

Win7 x64 VM Installation: ThinApp Package

  • Comments: “Unable to load DLL” error at startup, resulting in “The program can’t start because MSVCR80.dll is missing from your computer.” Similar error regarding LogUtils.dll. Similar results without 2D and 3D acceleration.

Windows XP x32 VM Installation: Program Installer

  • Installed size: 750MB.
  • Task performance: (1) 0:36. (2) 0:38.
  • Comments:  Took six seconds to start. Same result as in Win7 VM (above): system freeze. On the second try, disabling 2D and 3D acceleration in the VM slowed performance on task (1) to 0:39, but task (2) was able to complete.

Windows XP x32 VM Installation: Cameyo Package

  • Comments: Unable to start: “FATAL: APIRET_DOTNET_REQUIRED” error. Disabling acceleration made no difference.

Windows XP x32 VM Installation: ThinApp Package

  • Comments: With or without acceleration, unable to start: “Configuration error. Please uninstall and reinstall the product. . . . Error: 16.”

Impressions

  • Aside from the native Windows installation, the VM installations were the only approaches that worked with APE. There did not appear to be any realistic prospect of getting APE to work via Wine.
  • In Windows 7 and XP VMs alike, it was necessary to disable video acceleration to prevent system crashes.
  • The Win7 VM installation was twice as fast on task 1 as the native Windows 7 installation. The WinXP VM installation was 50% slower than the Win7 VM installation on task 1, yet it was still faster than the native installation. That slow native performance may have been a fluke, or this may have been a result of complete access to two CPU cores.

Recommendations

  • The Windows 7 VM was the best alternative to the native Win7 installation. Pending further experience with the effects of disabling video acceleration, it appeared that the Win7 VM might actually be superior to the native installation.
  • As long as my system would run at least two VMs simultaneously, and especially if other VM programs benefited from video acceleration, or if APE’s performance was much better when it had CPU cores all to itself, it might be best to put APE in its own Win7 VM.
  • It was possible that APE would only run well in the native dual-boot Win7 installation. In that case, it would make sense to look again at alternative Linux video editing programs.

Cool Edit 2000

Tasks tested:

  1. Convert a 20-minute .wav file to .mp3 with my usual settings.

Native Windows 7 Installation

  • Installed size: 13.2MB.
  • Task performance: 0:59.

Linux Wine Installation: Program Installer

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 13.7MB.
  • Task performance: 0:54.
  • Comments: It appeared that at least basic program functions (e.g., Play, Convert Sample Type) were operational. After installing the program, I used Wine to install the registration key. It looked like other tweaks would work as well (e.g., replacing the stock scripts at /home/[username]/.wine/drive_c/Program Files/Cool2000/scripts with my own scripts).

Linux Wine Installation: Cameyo Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 19.2MB.
  • Comments: Program failed to run.

Linux Wine Installation: ThinApp Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 24.4MB.
  • Comments: Program failed to run.

Win7 x64 VM Installation: Program Installer

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 13.0MB.
  • Task performance: 0:55.

Win7 x64 VM Installation: Cameyo Package

  • Task performance: 0:54.

Win7 x64 VM Installation: ThinApp Package

  • Task performance: 0:55.

Windows XP x32 VM Installation: Program Installer

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 13.0MB.
  • Task performance: 0:56.

Windows XP x32 VM Installation: Cameyo Package

  • Task performance: 0:54.
  • Comments: Oddly, only the trial installation was loaded, even though the package was created in a WinXP VM and ran as a full installation in other settings (above).

Windows XP x32 VM Installation: ThinApp Package

  • Task performance: 0:56.

Impressions and Recommendations

  • Cool Edit was atypical among the programs examined here: it worked in all settings except for the attempt to run its virtualized versions via Wine, and its performances were almost identical in most settings. These results may have been due to the fact that it is a small, legacy program: it may be especially amenable to Wine and to virtualization.
  • If my learning about Wine advanced, such that I could include this program’s licensing and plugins, it would seem most advisable to install it on the Linux desktop. Otherwise, in a Win7 VM, licensing and the advantages of virtualization would argue for running the Cameyo package.

CyberLink PowerDirector 12 Ultimate

Tasks tested:

  1. Play portions of a one-minute .avi file while observing before-and-after split-screen effect of image stabilizer feature with default settings.
  2. Save resulting stabilized video as Windows Media Video 9, 720 x 480 / 30p (1.6 Mbps).

Native Windows 7 Installation

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 3.3GB.
  • Task performance: (1) OK. (2) 1:45.

Linux Wine Installation: Program Installer

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: Not available.
  • Comments: Program failed to run. There seemed to have been very few efforts to run PowerDirector in Wine. There was virtually no information about PowerDirector in the Wine or CrossOver databases.

Linux Wine Installation: Cameyo Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 2.3GB.
  • Comments: Program failed to run.

Linux Wine Installation: ThinApp Package

  • Comments: Did not attempt to create a package in ThinApp.

Win7 x64 VM Installation: Program Installer

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 3.1GB.
  • Task performance: (1) Playback ran without hesitation in real time. (2) 2:35.

Win7 x64 VM Installation: Cameyo Package

  • Time to start: 0:04.
  • Comments: Was not able to set up stabilization: attempts repeatedly crashed the program.

Impressions and Recommendations

  • The Windows 7 VM installation was slower than, but an acceptable alternative (and the only working alternative) to the native Windows 7 installation.

Microsoft Office 2003

Tasks tested:

  1. Verify that key programs open and look OK: Access, Excel, PowerPoint, Publisher, Word.
  2. Excel: save a spreadsheet containing many calculations.
  3. Word: search and replace many occurrences of “Monday” with “Tuesday.”
  4. Repeat the Excel and Word tests using LibreOffice in the same setting.

Native Windows 7 Installation

  • Did installation or packaging appear to succeed? Yes.
  • Task performance: (1) OK. (2) 0:08. (3) 0:10.
  • LibreOffice comparison: (2) LibreOffice Calc crashed twice when attempting to open the spreadsheet. (3) LibreOffice Writer claimed, “This is not a valid WinWord97 file.”
  • Comments: These performance values were for Office 2016, which was what I had installed on the Windows 7 dual-boot system. Office 2003 apparently tended to produce faster benchmark scores than some newer versions (e.g., Office 2007, Office 2010). Office 2016 programs generally did not open nearly as quickly as Office 2003 programs. In this instance, Word 2016 took 1:18 to open this file and another 0:11 to open the Search and Replace dialog. Word 2016 also produced “Not Responding” when I tried to close it.

Linux Wine Installation: Program Installer

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 293MB.
  • Task performance: (1) OK. (2) 0:39. (3) 1:06.
  • LibreOffice comparison: (2) 0:20. (3) Repeatedly failed to open the Word document; crashed every time.
  • Comments: Appearance was good. Among the entries for Office 2007 or 2010 in the WineHQ database at this writing, results for the 64-bit version of Office 2010 were consistently terrible. This was not surprising; I had gathered that Wine did much better at the 32-bit level. All reported results for Office 2007 were at the Gold or Platinum levels, as was one of the five reported results for 32-bit Office 2010. When people reporting those results emphasized the 32-bit Wine prefix, I believed they were calling for just the steps that I was now treating as standard procedure when installing Wine. They also called for setting riched20.dll to “native, builtin.” One reviewer recommended doing that and also adding msxml60 through Winetricks.

Linux Wine Installation: Cameyo Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 707MB.
  • Comments: Package failed to run. See other post; see also another post regarding Cameyo with Microsoft Office 2010.

Linux Wine Installation: ThinApp Package

  • Did installation or packaging appear to succeed? Yes.
  • Package size: 1.6GB.
  • Task performance: Unable to open any Office 2003 program.

Win7 x64 VM Installation: Program Installer

  • Did installation or packaging appear to succeed? Yes.
  • Installed size: 272MB.
  • Task performance: (1) OK. (2) 0:07. (3) 0:59.
  • Comments: Programs open almost instantly — very snappy feel. Reinstalled and re-ran tests to be sure the task performance numbers were correct.

Win7 x64 VM Installation: Cameyo Package

  • Task performance: (1) OK. (2) 0:04. (3) 0:55.
  • LibreOffice comparison (using LibreOffice Portable): It appeared LibreOffice was unable to load either the Excel or Word files. After several minutes of inactivity, I tried closing LibreOffice Calc and Writer, but ultimately had to reboot the VM to shut them down. Similar results in a repeat try.

Win7 x64 VM Installation: ThinApp Package

  • Task performance: (1) Slow startup (~8 seconds) in a few cases, but all ran. (2) 0:11. (3) 1:49.

Windows XP VM Installation: Program Installer

  • Installed size: 272MB.
  • Task performance: (1) OK. (2) 0:11. (3) 1:37.

Windows XP VM Installation: Cameyo Package

  • Programs failed to run.

Windows XP VM Installation: ThinApp Package

  • Task performance: (1) Programs took 12-15 seconds to start. (2) 0:11. (3) Failed to respond to command the first time; “Not Responding” for ~10 seconds. Second time: hourglass without progress indication; completed at 2:26.

Impressions

  • In these tests, Excel 2016 in the native Windows 7 installation was no faster than Excel 2003 installed in the Win7 VM, but Word 2016 was significantly faster than Word 2003.
  • These tests provided grounds to believe that Office 2003 might be more responsive, for ordinary office tasks, than Office 2016. But for some tasks, the latter would have many features and capabilities not available to the former.
  • The Excel task ran much faster, and the Word task ran slightly faster, in the Win7 VM than in Wine on the Linux desktop.
  • In the Win7 VM, both the Cameyo and ThinApp packages worked. Cameyo was actually faster than the installed Office 2003.
  • Neither Cameyo nor ThinApp worked in the WinXP VM. The WinXP installation was significantly slower than the Win7 VM.
  • I had been mentally reserving the option of using LibreOffice as an alternative to Microsoft Office. Indeed, I had used LibreOffice with good results on a number of occasions. For whatever reason, though, with these two test files, LibreOffice turned in a surprisingly poor performance.

Recommendations

  • Rather than depend on LibreOffice, it seemed best to retain access to both Office 2003 and Office 2016. The latter, already installed in the dual-boot native Windows 7 system, might also or instead be installed in the Win7 VM.
  • Given my desire to move toward a less Windows-based system, it appeared reasonable to install Office 2003 on the Linux desktop via Wine.
  • If Office 2003 via Wine ran into problems, I would want Office 2003 available in a Win7 VM. Given the complexity of the Office 2003 installation, with various updates and tweaks, I would prefer to use the Cameyo package in lieu of a regular Office 2003 installation in that VM.
  • I had heard that the presence of an installed program could interfere with its virtualized counterpart. If I installed Office 2016 in my primary Windows 7 VM, then I would probably offer Office 2003 via Cameyo in an alternate Win7 VM.

Conclusions

This post explored my options for running some key Windows 7 programs in Linux. What I learned about the five programs considered here — Adobe Acrobat and Premiere Elements, Cool Edit 2000, CyberLink PowerDirector, and Microsoft Office — would be informative for other Windows programs as well.

In the configuration shaping up here, it appeared that I would be able to use a Windows 7 VM to replace the native dual-boot Windows 7 installation on my computer. Or, more accurately, two Win7 VMs. One would contain installations of Adobe Acrobat, CyberLink PowerDirector, and Microsoft Office 2016, and might also offer Cool Edit 2000 in a Cameyo package. Another, with video acceleration disabled, would contain an installation of Adobe Premiere Elements, and would also offer Microsoft Office 2003 in a Cameyo package.

It appeared, in addition, that I would be able to run some programs in Wine on the Linux desktop. I was already able to run Office 2003 and Cool Edit 2000 there, though for both I would probably need to learn how to use Wine to add updates and tweaks. If I wanted to spend the time, I might also try getting Acrobat 8 to work in Wine.

Those appeared to be the conclusions from this preliminary exploration. I would need additional experience with these programs in these contexts to see whether they were performing as hoped.

Advertisements
This entry was posted in Uncategorized and tagged , , , , , , , , , , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s