GravityMark v1.87 GPU Benchmark

[:D:D]

Ryzen 7 5700X + RTX 3060 Ti: https://gravitymark.tellusim.com/report/?id=634e37499290e7c3d57cd6b6f851549440e1c330
Min / Max 29.8 / 280.9

While minimum shows 29.8 the graph shows all FPS above 100?

 

[SchumannFrequency]

While minimum shows 29.8 the graph shows all FPS above 100?

It shows the frame rate per second, where it makes sense that you don't see the minimum and maximum.
It probably calculates with a simple formula how many fps were generated per second.
In a smaller time interval you would see that the min/max values are indeed correct.

Determining the maximum frame rate (equation):

Product Documentation - NI

www.ni.com

 

[:D:D]

Would have thought FPS would have read x number of frames each second and graphed accordingly. IMO using 1 / single frame rate for FPS seems dumb and should just be shown as shortest frame time. Max FPS seems to be equal on graph and report although I only looked at a couple of results. If minimum FPS is instead supposed to be something like 1% minimum then should be stated as such.

Running a third party tool such as Mangohud under Linux shows some frame times much higher and lower than reported FPS. Average FPS is close though.

 

[Klemc]

What is the folder directory where result is saved, when pressed SAVE button ?

 

[Ruru]

 

[:D:D]

What is the folder directory where result is saved, when pressed SAVE button ?

IIRC save takes a screenshot and saves under 'Documents'. If you click 'send' a link is saved there too or you could check GravityMark\.GravityMark.log under PC <username> folder.

 

[Klemc]

7800X3D
B650E-E
KF560C32RSK2-32 - XMP SOC 1.07v

 

[SchumannFrequency]

IMO using 1 / single frame rate for FPS seems dumb and should just be shown as shortest frame time.

Technically, the term used is 'Min/Max'. It does not say min fps or max fps.
You can always contact the developer and ask how he calculates it.

1 / time for single slowest (or fastest) frame is a very useful statistic
To give an unrealistic example, suppose it takes half a second to calculate one frame and then it generates 98 frames in 0.5 second.
Then your fps is going to be 99. It seems that it runs smoothly if you would only look at the fps for that one second, but in reality it did not run smoothly.

Running a third party tool such as Mangohud under Linux shows some frame times much higher and lower than reported FPS.

This gives the impression that one of the two tools may not be reliable.
If you look at the source code of both tools you can probably see where the differences lie.

This reminds me of the following article: https://vermaden.wordpress.com/2022/07/12/desktop-environments-resource-usage-comparison/
Looking at the article you quickly come to the conclusion that each tool seems to calculate RAM usage differently.

There is software in circulation anyway that is just not reliable.
Many programming languages that are used are not optimal for the task for which they are used.
No super small percentage of programmers are weak in math, and many programmers are only really good at one or two programming languages (or not really strong in any programming language)

I am not saying this applies to this situation, but it is one possibility.

 

[:D:D]

Have you tried the inbuilt -times setting? This saves frame times of the run.

Here's a run I did with -times option. https://gravitymark.tellusim.com/report/?id=f0146a29f086caaca19f36ac046008298ebd4e1b

If we take the number of frames each second we get this

And if we take 1 / each single frame this. For me the third time frame always long, around 0.2 to 0.4 seconds! We also get some short ones as can be seen.