CPU comparison in games - 0.1% fps please

[total]

Hi guys, you do a great job, and there's still some room for further improvement.

You show only average fps and 1% low, and it's also important to show if there were any freezes/microstutters, 1% does not really show that, that's why 0.1% low is important for a gamer to have a more comfortable gaming experience.

Please, when testing new CPUs (Zen 2 for example), add 0.1% low to the results so that people know which CPUs have no or at least less of those annoying microfreezes/stutters!

Thank you in advance!

 

[neeyik]

This is a good topic for discussion, as in how much statistical data is worth displaying in reviews. Let's say one has a healthy sample size for each test, such that the statistics displayed are reasonably robust - here's what one could usefully do for central tendency and dispersion:

• mean
• median
• 1st and 3rd quartiles
• 1% and 99% percentiles

Then, of course, is another question on how this can be displayed and at what point does the amount of statistics shown be detrimental to ease of reading. I personally like the 1% and mean in a bar chart approach, but I also like fill rate charts, box and whisker displays, and histograms.

 

[total]

There are many posts about *likely* microstutters and thus less comfortable gameplay when your CPU load % is close to 100, in CPU-demanding games (e.g. AC Odyssey, Shadow of the Tomb Raider (DX12)) i5-8600K/9600K or even i7-9700K is often almost 100% loaded, and they use it as an argument against CPUs with less threads.

i7-7700K and i5-8400 lost to 2700X at 1% fps in BF5, but 1% fps doesn't really indicate if there were frametime spikes, unless there were many enough of them.

Of course it would be much better to rather add the frametime graphs separate for each of the main CPUs to do a comparison of the cheaper or less-threaded CPUs (i5-9400F, i5-9600K, R5 2600(X)/3600(X)) and a comparison for more expensive ones (8700K, 9700K, 9900K, 2700X, 3800X, 3900X). I know it will take more time to do that, but it will also add more value to those benchmarks results, because having frametime spikes makes your gaming experience worse, and it's really great to just come to techspot.com and see those graphs and know which CPU has enough power to handle a game without annoying stutters and which CPU has less of those frametime spikes.