The "Most Benchmarked" page is messy now.

[Noksak]

The list is not sorted correctly (see picture). And some entries appear with 0.0%.

[David (PassMark)]

We'll get it fixed up in the next couple of days.
(I think the percent values are from PerformanceTest V10 submission, but the bar length is computed from the PT9 submissions).

[Noksak]

Okay, thank you. But why not combine both version statistics together as they all indicate popularity (numbers of benchmark run)?

FYI, I think each bar and its percentage are quite correlated, those entries are just sorted incorrectly.

[Richard (PassMark)]

This should now be corrected.

The chart data was correct, it just wasn't sorted properly. The chart does take into account both sets of data from V9 and V10 for the last 30 days. When we combined the results (samples from V10 and samples from V9), we forgot to resort it by the total samples and it was originally sorted by V10 total samples only. So while the results may be in the top 200 for V10, after adding V9 sample count in the calculation of popularity, the 0.0% was likely due to a rounding issue.

[Noksak]

Well, the sorting problem has been fixed.

There is room for improvement. The lower part of the list has so many entries with same percentage (e.g., 0.1%), which is impossible to determine the potential different between them. Adding a digit should be fine.

[David (PassMark)]

We believe the sorting is correct, even if the result isn't printed to 2 decimal places.
So units higher up the list are more popular.
Not sure if it matters if popularity of a CPU is 0.08% or 0.06%. ?

Was better in the past when there was less CPU models were on the market. It's a bit surreal now that even the most popular CPU model only have 3.6% market share. Maybe this would make sense if there were a dozen different CPU vendors. But there are really only 2 in the x86 space.