My way for performance testing (Guide)

There is my methods, how I test real performance on devices. I always use that if I doing tests for @ptr1337

What you need?
If you expect tools, special software... I will disappoint you. We want real world performance in real a scenario.

1. Select fast games, what you really know. (It mean, seriously fast - every fps drop will strongly hit your gameplay)... I recommend games where you have +200 hours. I use https://xonotic.org/ and Trackmania.

  • Where fps usually went down and up.
  • Aspect what's can change fps value. (Maps, number of enemies, graphics settings)
    Cool tip: Remeber usuall fps value for unique weapons, vehicles parts of story - cinematic scenes)

2. Same environment for device (Avoid to interferences from thermal throttling)

  • Same or similar temperature at your room.
  • Same place for device (More important than room temp) - airflow is differently if you change place where you use device.

3. Use only 3 scenario for testing, you can't correctly memorize more.

  • Small, medium and large map where you will observe fps value. Don't use more maps!

4. Hard core conditions.

  • Set your game on graphical settings where you have 30fps... everything smooth for you, if fps went up or down... you know immediately. (Mmm game is better or worse), you can't see so much changes if you play on 60+ fps.
    Apart from me... I have nystagmus - rapid involuntary movements of the eyes. Smooth movements start for me around 50 fps.
  • Play with sniper rifle... lower or unstable fps is hell for this.
4 Likes

Games are just one possibility of a real world scenario.
There are others, like e.g. compiling, compressing, hashing, rendering etc.
If you only test games, you will also only get results for games :wink:

3 Likes

Well it's true, you can apply similar principles to another scenario.

I think the only reason games are used to test hardware more often than software is that it's where fps matters. If your computer is slightly slower to render a video, or compile a kernel, the end result is still the same - a usable file/package. In games, it's really all about that live performance. I guess certain heavy 3D software like Blender can yield varied experience under certain amount of processing power, but beyond that, speed is just not that important in professional/everyday computing as it is in entertainment, and when it is it is much less noticeable in real time by the user. That's why benchmarks are better for assessing performance in those kind of tasks. If you are gonna use a computer for gaming, I agree, you should test how it performs in games!

2 Likes

Does it really matter whether a game runs with 156 or 159 fps?
The end result is also the same. A fully playable game.
(and yes, I'm a gamer as well :wink: )

Everyday, okay. But professional, certainly not. It's exactly in the professional domain where performance matters most. Time is money after all :slight_smile:

4 Likes

No denying that, but is an audio engineer really going to sit and time how long it takes them to bounce a 16-channel, 10 minute track and compare it among different kernels? Probably would be considered a waste of time. Gamers however, seem very keen to spend that time testing. Maybe because it's still a game either way.

Probably not, but apparently for some people 49 vs. 52 makes a difference...

1 Like

Well, no, they have better things to do than to compare kernels :slight_smile:
But imagine big companies like the GAFAs, they want their huge systems to bring the best performance (or rather best performance per Watt).

In the end, it comes down to what you do with your computer.
If gaming is the most important task, then of course it makes sense to benchmark and tune the system into that direction.
But I think that most people will use their computer for plenty of different tasks.

3 Likes

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.