Ryzen APU Benchmarks

Every now and then, I update my Ryzen spreadsheet for Bang per Buck comparisons (personal curiosity).

The past couple of times, I've noticed that the Ryzen 3 2200G is out-pacing the Ryzen 5 2400G in one of the benchmarks (the second image):






Is this a benchmarking issue, or do APUs inherently create 'dodgy' benchmarks, given that we can't know which processors are using integrated or dedicated gfx in the tests?

Just curious as to the likely causes of this. btw, the 'Bang per Buck' is just a number based on perf vs price. Nothing worth quantifying beyond that. I find it handy to see who's up and down. Oh, and prices are UK GBP, and are the lowest from my own selected vendors, based on personal experience.
 
Solution
Thanks chaps. Given that both have 6-7K samples, I thought outliers would have been sidelined.

One thing I noticed when re-visiting the UserBench site - http://cpu.userbenchmark.com/Compare/AMD-Ryzen-3-2200G-vs-AMD-Ryzen-5-2400G/m441832vsm433194



Notice that the 2200G is given 4% better effective speed, but the specific test differences don't seem to support that. Very odd. Maybe the average 2200G user tries to boost their system in other ways (e.g. faster RAM), than the average 2400G user, and gets better bench results, as the gfx is linked to the CPU, and relies on RAM speed (which might come back to the UB bench tests not isolating CPU testing as much as Passmark, perhaps)...

WildCard999

Titan
Moderator
I just wanted to add that not everyone who's benching via Userbenchmark has updated iGPU drivers or the CPU/iGPU overclocked which is going to create quite significant differences. Then there's also the performance difference of people using DDR4 2133 to DDR4 3200.
 
Thanks chaps. Given that both have 6-7K samples, I thought outliers would have been sidelined.

One thing I noticed when re-visiting the UserBench site - http://cpu.userbenchmark.com/Compare/AMD-Ryzen-3-2200G-vs-AMD-Ryzen-5-2400G/m441832vsm433194



Notice that the 2200G is given 4% better effective speed, but the specific test differences don't seem to support that. Very odd. Maybe the average 2200G user tries to boost their system in other ways (e.g. faster RAM), than the average 2400G user, and gets better bench results, as the gfx is linked to the CPU, and relies on RAM speed (which might come back to the UB bench tests not isolating CPU testing as much as Passmark, perhaps).

We'll probably never know. :D
 
Solution