Manual benchmarking can be done by learning the route. I benchmark Cyberpunk 2077 all the time using the same route. When using this method, you will learn the route. You will almost certainly run it time and time again until you can repeat it consistently. The thing is, while there are variances from run to run in terms of where the cars are and things like that, it doesn't generally impact the data much if at all. That is, I can run it on a given hardware configuration and follow the same route 5 times and the data will be consistent between all 5 runs. The differences you might see from run to run don't actually alter the data. So long as you start from the same save, giving you identical weather conditions etc., it works fine. Reviewers have literally been using this method for more than two decades now. It works. Is it ideal? Not necessarily.
The accuracy of this method depends on a lot of factors. How consistent is the reviewer? How consistent is the game's performance? Some games by their very nature aren't terribly consistent or have anomalies in the data caused by streaming, loading instances areas etc. Destiny 2 does the latter which creates weird results in the data sets. It's why I don't use it anymore. That and I prefer the canned benchmarks as they are more hands off and I can run those in the background while doing other tasks.
There are pros and cons to it. In game benchmarks are not always representational of the performance you would actually get playing the game. The big benefit is that they are hands off and take user error out of the equation almost entirely and the numbers are certainly consistent with your hardware and game settings being the only variables. However, manual runs are what you'd see actually playing the game because you are actually running the game. The benchmarks are often configured by the manufacturer to paint the game in the best light possible, so while you might get great scores in the benchmark, it doesn't mean that your actual experience on that same hardware will be as good.
As a reviewer, I benchmark all kinds of shit all the time. However, I normally just max out my games on my personal system and if they run like crap, I figure out what setting I can dial back with the most performance impact, but the least visual downgrade. If I can upgrade my hardware, I will.
Ok thanks guys. I’m not a reviewer. I’m upgrading from a 3770k to a 12900 and using my 3080 to the new rig. Obviously, I’ll have a huge difference in performance but seeing the actual numbers will be interesting.
I have zero faith at YouTube Games footage in the form of benchmarks.
99% of them they are fake, and their origin from Africa and other third world countries, authors unknown, no name no address.
Its a scam so them to collect few cash from YouTube views.