We all know how important the minimum framerate is in producing a good gameplay experience. Please could this be measured in future Guru3D reviews? The average framerate on it's own does not tell the whole story. A hypothetical game could run at an average 100fps but if it drops down to 2fps on occasions, it could result in a horrible gaming experience. Comparing 2-way to 3-way SLI is another area where minimum framerate is important. 3-way SLI tends to have higher minimum framerates which decreases or eliminates the micro stutter seen by SLI users.
Hey guys, the problem I have with min FPS is that they can be interpreted the wrong way. If you play 15 minutes at 80 FPS and then denote a min framerate of 4 FPS that happened on say two occurances for a second then that paints a subjective story. Anything can be the cause of a framerate drop -> utility polling the framerate - HDD loads up something - background tool does something - email checks - windows starts up something. But obviously also the game itself. So the minimum framerate is IMHO too sensitive to external factors, as such we don't do anything with it. If we do note down the min FPS .. that 4 FPS is the only thing that people will see and form their opinion on. The average framerate really is the most consistent and objective way to show performance as it paints a generic picture of overall performance.
I think it's still worth recording, with a simple quick "explanation" if you will of what you experienced. Let's say there was a minimum of 4 for a tenth of a second that shows up in the bench. Just explain that you never saw it therefore it was most likely in error. I can't think of many cases where the minimum FPS would even be close to that low with modern hardware, and the number does help compare different CPU's especially and evaluate CPU bottlenecks.
Thats a good idea but I think a diagram of time vs fps will be even better. With a detailed enough diagram everyone will know if the min fps for a particular game are actually relevant or not.
You would think so, but in reality people see a minimum of 5 fps for 1 second and they freak out. Doesn't matter how good the description is, if people don't actually read it then its pointless.
If the drops are constant, even for 1 sec, its a problem, if not its just a random anomaly. People here at Guru are quite tech savvy and they'll know the difference imho
Games that stream a lot from disk will have high fps drops when other apps push the i/o up. Hence why people experience a huge difference when they cache to ram or got the game on ssd. Keep in mind that a lot of people only got 1 hdd for everything.
Min. framerate is the FIRST thing i look for in reviews!. It will be even better with a time-table so we can see were the ''dips'' occure and some explanation with it ( i.e. transition from a level, loading texture's, game engine being borked etc.) When i see a good min fps over a period of time during a heavy action scene or whatever i know that card delivers and is a must buy!.
I agree it's still most definitely worth doing. I love the Guru3d reviews, but always have to go to additional websites to see the minimum framerate on games Running the games off SSDs would help to eliminate some of the issues that Hilbert mentioned above My games are installed on a 256GB RAID0 volume of Corsair Performance Pro SSDs. Windows is also on it's own SSD too. I appreciate not everyone has this set up, but the hardcore amongst us probably have something similar. And this is an enthusiast website after all! Plus as long as Hilbert explains erratic FPS drops, we can draw our own conclusions too. Many thanks
Minimum frame-rate may be misleading. Minimum sustained should be somewhat better. Techreport gives Time-spent-beyond 50,33,16ms
What tech report are currently doing with frame times is really quite beneficial and moves performance testing on. Average fps isn't sufficient to say if a game runs smoothly on a piece of hardware. Graphing out the range of frame times certainly starts to show the problems.
I agree. Benchmarks should be time based, with a graph. Appreciate this is problematic, but perhaps should be considered for the lowest common denominator of specifications/resolutions/settings.
The problem I see with tech report's method is that it only takes one tiny little area of low fps to skew the average fps results for the entire test. Although I guess a graph/timeline would show readers the full picture.....
