Initial Performance
In keeping with the previous review of Win4Lin Pro, Tom's readers were exposed to an industry-respected benchmark suite called SYSmark by BAPCo. Owing to DirectX compatibility issues under Win4Lin, SYSmark instrumentation is essentially useless, because DirectX is listed as a product requirement according to Win4Lin documentation. To further complicate matters, benchmark results are skewed between native and virtual contexts owing to differences between emulated and actual clock cycles. We contacted Jim Curtin, CEO of Win4Lin, to find out whether any in-house benchmarks demonstrated an objective performance comparison, and the information we received from their engineers confirms this information.
Instead, they suggested two subjective performance tests to produce ballpark estimates, as follows:
- Total Time to Installation - from start to finish, calculate the amount of time Windows takes to install under both emulated and native contexts.
- Total Time to Desktop - from a cold (first) boot, record the amount of time Windows takes to deliver you to the desktop
Time To Installation
We did exactly that, and the results appear in the table below. The time frame originates from the moment the Windows installer displays a progress meter during the textual pre-installation process (when it is essentially copying files).
| Install Progress | Emulated Install | Native Install |
|---|---|---|
| 10% | 3:10 | 1:20 |
| 25% | 6:42 | 1:43 |
| 40% | 10:37 | 2:30 |
| 50% | 16:28 | 3:04 |
| 100% | 32:03 | 5:45 |
| PRODUCT KEY | 35:40 | 10:03 |
| COMPLETION | 55:08 | 25:01 |
Several interesting things occur: under emulation, the installation process slows to a crawl around the 30% mark or thereabouts. At first, the Windows installer speeds along up through 24% of the installation process, then slows down appreciably thereafter. 51% into installation, the emulated installation again slows perceptibly from its previous slowdown. The native installation idles when it hits 50% for around three minutes but continues to outpace the emulated process considerably. Unfortunately, these results don't favor Win4Lin Pro, and many factors contribute to this difference - for starters, Windows isn't the only OS occupying space and exerting control over system resources when operating under emulation.
Startup Time
That said, we then turned to start-up time from a cold boot to safe delivery to the desktop. Since installation is a one-time deal, it makes little sense in the context of daily usage and operation. Here are those results:
| Progress | Emulated Boot | Native Boot |
|---|---|---|
| Time to Desktop (1st) | 2:18 | 1:15 |
| Time to Desktop (2nd) | 1:09 | 1:24 |
| Time to Desktop (3rd) | 0:53 | 1:03 |
The third table illustrates time to open PowerPoint presentation files of various sizes. The demonstration files used are obtained from the Space Symposium Web site. Both presentations under the first heading, Human and Robotic Space Exploration of the Solar System, and the presentation file by Jim Geringer (29 MB) are most suitable.
Open Powerpoint File
| PowerPoint Open File | Emulated Open | Native Open | ||||
|---|---|---|---|---|---|---|
| File Size | 1st | 2nd | 3rd | 1st | 2nd | 3rd |
| 3 MB | 0:08 | 0:09 | 0:08 | 0:03 | 0:02 | 0:01 |
| 10 MB | 0:06 | 0:07 | 0:09 | 0:02 | 0:02 | 0:01 |
| 29 MB | 0:27 | 0:26 | 0:26 | 0:09 | 0:05 | 0:04 |
Though these are fuzzy measurements of operational performance, the numbers show no surprises. Start time begins with the initial opening (first click) of a presentation file and finishes the moment PowerPoint becomes usable. Each file is selected in no particular order, with no single file opened twice in succession. On the native side, performance is clearly faster with some measurements (in the 3rd column) coming in under a full second. Initially, performance seems to improve with each open despite the non-deterministic opening and closing of the individual files.
