PC hardware reviewer benchmarking Attilla need advice from community

[Dropinbodies]

Hi everyone, I have a gaming CPU comparison article I'm writing for CustomPCReview, From actually playing the game (Something a lot of other hardware reviewers don't do) I noticed the in game benchmark doesn't seem representative of the actual gameplay. I get about 60-70FPS in the benchmark and a range of 40-70 FPS in game (sometimes even worse when bringing the camera closer on units fighting) I wanted to know if the commuinity had a better way of doing it, possibly a game replay or something very repeatable.

Sorry for the poor grammar It's very late and I'm quite tired lol.

[RGA]

- Siege battle with city walls and lots and lots of ladders/ siege towers
- 40 vs 40 with 100% melee unit
- Set the city on fire with onagers
- Mad rush
- Zoom in close as men fight against men on the battlements
...
- Profit

[Huberto]

I find it hard to believe that you're getting 40-70 fps, at least while zoomed in on a large battle. I have decent hardware that's overclocked and I get 15 fps.

[Dropinbodies]

- Siege battle with city walls and lots and lots of ladders/ siege towers
- 40 vs 40 with 100% melee unit
- Set the city on fire with onagers
- Mad rush
- Zoom in close as men fight against men on the battlements
...
- Profit

I can use replays to try to get it very close to the same correct?

I find it hard to believe that you're getting 40-70 fps, at least while zoomed in on a large battle. I have decent hardware that's overclocked and I get 15 fps.

I said even worse when zoomed in, I can still somewhat micro when zoomed out but it gets a bit odd.

[RGA]

I can use replays to try to get it very close to the same correct?

Sadly no. One of the most CPU-taxing task is AI. City battle is notoriously laggy because of complex path finding algorithm (other may call it sub-optimal, it's personal taste ). Thus replay tends to have better FPS than actual gameplay.

I find that siege battle is sufficiently reproducible. If you are attacking and use the same approach, chances are that AI would behave pretty the same.