Welcome to SPEC's interactive results engine, a companion service to SPEC's 'osgresults' tool that enables anyone to interactively browse through the thousands of results available at SPEC.org.
To begin, use one of the buttons above to chose a benchmark, this will load all of the results for that benchmark that have been published at SPEC.org. Once all the results have been loaded this view will be updated with options to select and view these results. For help, the active parts of this interface all have "tooltips" with more information, hover over an element or a control to get the tooltip to appear. Go ahead, pick a dataset and go exploring. If you get stuck anywhere just reload this web page to return to a fresh start and then you can read the help text below.
Most of SPEC's results are available in one or more of these datasets; one dataset per benchmark. Each dataset contains entries for every result published at SPEC.org, and depending upon the benchmark there may be several or many different columns of information included in the dataset. One of the last columns is the 'Disclosures' field, which includes links to the full disclosure reports for each result with all of the detailed information about that result. [Note: a few of these datasets are very large, especially the datasets covering "CPU2006" each of which have around ten thousand of results included. Depending upon your system and your network connection the larger datasets may be very slow to work with. It may help to start with one of the smaller datasets.]
Filters can be used to trim unwanted results from the current dataset, e.g. "'Processor' not match 'xeon'" (matches supports regular expressions and ignores case), "'License' equals '6' OR 'Tested by' matches 'Sun'" (two conditions can be combined with either AND or OR), "'# Cores' more than '1'" (simple comparisons) as well as "'Test Date' at least 'Jan-2017' AND 'Test Date' at most 'Dec-2017'" (inclusive comparisons are supported to make ranges easier). A filter can be disabled by unsetting or clearing one or more parts of the filter definition. Also when a dataset is first loaded no results are filtered, so loading a new dataset or reloading a current dataset will clear all filters.
These results then can be displayed in three different ways.
The 'Table' is the default display. There are controls to select which columns to include in the table. Also clicking on a column header sorts the table by that column's values. There is a button for a smaller font size to shrink the table display, which can be helpful when looking at datasets with many columns selected. Formatting large tables is very compute intensive for most browsers, so to avoid responsiveness issues the tables are truncated after a set number of rows (though this limit can be adjusted to suit, or can be set to an enormous value if a complete table with many thousands of rows is desired).
The 'Dump' display generates simple CSV allowing copy-and-paste into other analysis tools. The current filter settings will be used to generate a block of CSV text. As with the 'Table' display, columns can be selected individually to be included/excluded from the CSV. [While lines of text are much simpler than tables, most browsers will still struggle even with plain text when the content length reaches high into the thousands of lines, so some browsers may become unresponsive while formatting a full dump of the larger datasets. In such cases, to avoid the browser overhead, use the link to "Download full dataset" to save a local copy of the data and then use a local data analysis tool, e.g. Excel, Tableau, R, Python, ...]
The 'Chart' display provides a means of doing simple visualizations of the dataset. Select a column to use for the "X" (horizontal) axis, and another column to use for the "Y" (vertical) axis, and a scatterplot is displayed. A third column can be selected to use as labels for the points on the chart; with some color schemes the values in this column will determine the color of the points on the chart (different color schemes can help show patterns or trends in the values). Hovering over any point in the chart will generate a pop-up display of that point's label value and will also change the color of all points that share the same label value. Any numeric or date field can be charted (text fields work only as labels, not as axis values).
By default the columns selected for these charts shows the first available result field (or at least the first available numberic field) over the 'Published' field (date a result was published at SPEC.org). This provides a quick overview of the history of that benchmark. Change the parameters to generate a chart that matches your interests. These charts are for quick exploration of what the data looks like. Each dataset looks different, and the interesting views of one dataset might show very little with another dataset. For deeper visual analyses, we recommend using the 'Dump' feature to get a local copy of the full data to load into your own favorite analysis tools.
Displays generated in this tool can be saved for sharing or for keeping a "snapshot" of interesting observations. The leading word in the header above the table, or chart or dump, is a link that can be used to save a static copy of the current display. Exported tables are saved as HTML (simple, single file copy of the current table), exported dumps are saved as CSV files (and should load easily into Excel or other tools), exported charts are saved as an SVG file (for raster formats, e.g. PNG, use your device's screenshot feature). At the lower right-hand corner of the "controls" area there is a button that can be used to hide all of the controls with a simple header so that snapshots look cleaner, click on any part of the resulting header to restore the controls.
In the lower left corner of the "controls" area, there is a "link to current settings" which encodes all of the current user settings in arguments at the end of the link. This link can be used to open a new window with the same information and same settings as the current view. This link can by copied and shared with others to enable them to open this app right at the current view. Or this link can be saved to provide a way to come back to the current state.