No announcement yet.

Annoying new (filtered) single-threaded charts

  • Filter
  • Time
  • Show
Clear All
new posts

  • Annoying new (filtered) single-threaded charts

    Thank you for the charts and the valuable data you provide.

    I understand the recent changes in PT10, which to the dismay of many, has thrown off the results; and although I was also initially unhappy, after reading your posts (particularly on single-threaded results), I've come to appreciate the reason for the change, particularly since other software has moved-on (such as newer compiler optimisations, and utilising newer instructions such as AVX/2).

    Still quite wary as to why Intel is once again on the top of your single-threaded charts, despite for example, AMD's third-gen Ryzen 3950x usually outperforming Intel, particularly in other single threaded tests in Cinebench and POV-Ray. Very strange to see the i7-9700 / 9700F, i7-8086k, i7-8700k, and even the i7-7700k, all doing better than the Ryzen 3950x?!

    Regardless, the reason for this post is to voice my dislike of how you have by default, filtered the single-threaded charts to separate desktop / laptop / server parts.
    Although this is useful, I wish you left the combined chart, but introduced filtering options. Or, at least have the option to combine all?

    It's sometimes really useful to see how well certain laptop chips do compared to the competition's desktop parts.

    Many thanks.

  • #2
    The top 50ish single threaded results are very tightly grouped. So close that a user wouldn't actually notice any performance difference in real life (internet speeds and disk speeds would have a much much bigger impact on how fast a machine feels).

    It is also an indication of just how small the progress has been on single threaded performance improvements over the years.

    So even a movement of a couple of percent means a major re-ordering in the single threaded charts.

    I think a few CPUs like the 7700K will drop a few percent over the next few weeks as we get more results for PT10.

    A lot of people quote Cinebench and POV-Ray as the reference for single threaded performance. But these two applications are a pretty narrow sample from the entire market of software applications. How many people actually use either application in real life? I don't know what the results would be, but a better test would be to see how the single threaded tasks in Excel, Javascript, Photoshop, Apcahe, PHP, (and a few other common apps do). Of course this kind of testing is hard to do well so no one does it. It's kind of lazy to say our results are wrong because they don't match Cinebench. Cinebench & PerformanceTest are just a guide to performance. They aren't (and never can be) exact numbers that represent the performance of a CPU under all possible conditions.

    We've started to split up a bunch of our pages into tabs. Some pages were so large that iPhones (running Safari) were crashing when they hit the page.

    Other pages recently broken up or filtered where.
    The socket page
    The CPU list page
    The mega list

    If you play with the filters & sorting on the mega list page, you can still get the entire single threaded list on one screen.