The PixInsight Benchmark

Something we can't stress enough is that the PixInsight benchmark is not meant as a competition among computers. Our purpose is to provide PixInsight users with robust measures of quality for their user experience on the PixInsight platform using different hardware. The benchmark has been designed to evaluate the efficiency of a modern machine working on intermediate and advanced image processing tasks with PixInsight. To this purpose we provide three integer performance indexes:

• Total performance index. This is a global estimate that attempts to include all of the factors that condition the performance of a computer system running PixInsight.
• CPU performance index. This is an estimate of pure CPU performance. This mostly includes processor performance, but also other important factors such as bus and memory bandwidth, and motherboard/chipset capabilities.
• Swap performance index. This is an estimate of the total I/O transfer speed for management and storage of image swap files.

In all cases, larger values denote better performance. The highest recorded total performance indexes are in the 10,000 - 12,000 range as of writing this document (May of 2014).

It is important to note that the total performance index is not the sum—neither numerically nor functionally—of the CPU and swap indexes. Actually, it is a much more complex estimate that includes less-than-obvious factors such as the way different operating systems implement parallel execution or memory allocation/deallocation and disk cache management, or how the amount of RAM available affects performance by conditioning these factors.

The swap performance index, along with the measured total transfer rate in GiB per second, includes both read and write transfer rates in a single estimate (this may change in future benchmark versions). The swap performance index is much more important than what may seem at first sight. A fast CPU on a machine with slow disks can provide a poor user experience in terms of the amount of time necessary to write and read working data before actually processing the images. As expected, solid-state disks are the absolute winners, as you can check by inspecting the benchmark results. The best swap performance is achieved when PixInsight is configured to use several physical hard disk drives for parallel swap I/O operations. There are some good examples of this in the benchmarks, where two or more solid state disks, or combinations of RAM disks and SSD drives, have been used in parallel to achieve spectacular transfer rates in the multi-GiB per second range. Leaving these extreme cases alone, the benchmarks demonstrate that even with several low-cost SATA drives configured for parallel swap storage your PixInsight experience can be substantially improved.

The PixInsight Benchmark script is a JavaScript implementation of the PixInsight benchmark. It requires PixInsight Core version 1.8.1.1092 or later, and forms part of the standard PixInsight distribution on all supported platforms (distributed as a regular update). To run the benchmark on your machine, you just have to execute the script (select SCRIPT > Benchmarks > PixInsight Benchmark from the main menu), accept a welcome screen, and follow a few simple instructions:

• First of all, thank you so much for submitting your benchmark results. With your contribution, we'll manage to make the PixInsight Benchmark a valuable resource for all present, prospective and future PixInsight users. The data that you submit also help us diagnose the performance of our software on different platforms and hardware, which is important to optimize our development plans.
• Benchmark submission is an anonymous process. The PixInsight Benchmark script does not collect any personal information that could be used to identify you.
• Before performing benchmarks, make sure you don't have other applications running, including background processes such as operating system or application updates. If you have an antivirus software installed (specifically relevant on Windows), also make sure that your antivirus application is not going to start a routinary disk scan or a similar maintenance task, as this would ruin your benchmark results.
• If you have configured PixInsight to use a reduced number of logical processors (i.e. less than the number of available logical processors on your machine), or if you have altered the default parallel execution settings via preferences (or using the 'parallel' command from the console), please don't submit your benchmark to our server.
• To perform the benchmark process, click the Run Benchmark button. If it isn't already available, the script will download the benchmark input image from this server. It's a combined linear RGB image of M74 acquired with the 1.23 meter Carl Zeiss telescope at Calar Alto Observatory in Southern Spain, which occupies about 11 MB. This image is part of the official PixInsight gallery. Once the image has been downloaded (or loaded from the local filesystem if available), it will be validated and the benchmark process will start. A benchmark may take a while to complete, from just 10 seconds to several minutes, depending on your computer's performance, so please be patient.
• Leave your computer alone during benchmark execution. For the best results, don't touch it at all; don't even move the mouse cursor during the benchmark process.
• Once the benchmark has finished, a new image window will be shown. This 32-bit floating point image is the result of the benchmark process, and includes a text line with your benchmark performance indexes. You may want to save this image for your reference.
• You can—should—perform several benchmarks, not just one. To submit your benchmark data, you must perform at least three benchmarks. This reduces the probability of sending occasional poor results, which improves the reliability of the whole benchmark system. Keep in mind that a single instance of the benchmark process can be affected by background tasks that all operating systems need to execute from time to time, such as disk caching and indexing operations, or update checks. By running several benchmarks you reduce the probability of these background operations having a severe impact on your benchmark results.
• To submit your benchmark, click the Submit button. You will be asked to provide optional information about your swap disks, your motherboard, and your computer, along with additional comments. These items are optional, but if you can provide them, your benchmark data will be more useful for us and for the rest of users. Of particular importance is the information that you can provide about the hard disk(s) used for PixInsight swap files. The script provides detailed information about the purpose of these optional items.
• The script allows you to generate a report of all the benchmarks performed, which you can save as a plain text file. When you submit your data, the script automatically generates a report with the data that have been sent to our server. We suggest you save this report for future reference. You can use your benchmark's serial number to locate it on this page.