Hey all,

I was curious as to how to optimize the performance of MGE and Distant Lands, so being the test engineer that I am, I decided to run a semi-thorough set of tests on the various parameters of Distant Land on MGE and record the results.

https://spreadsheets.google.com/ccc?key=t8BBP3vmSdB8oNROiZ_Sizw&hl=en&authkey=CIfkiOYM#gid=3 Graphs, Screenshots, and raw test values.

The big take-aways from this were:

• Normal Maps at their highest settings are a huge waste of memory (almost 300MB) for almost no improvement in visual fidelity. Anything above 2048 (or arguably 1024) is wasting hundreds of MB of video memory.
  
• Reduction of Draw Distance is hands-down the biggest improver of FPS and makes the most meaningful impact on GPU Load. Reducing World Mesh detail helps as well, but only slightly.
  
• Reducing World Texture and Normal Map resolution size resulted in zero improvement of my FPS or GPU Load. I thought this was surprising.

Anyway, I hope this is useful to someone, possibly the MGE devs or just someone trying to squeeze out more performance from their rig and would like to see what has an effect on what. Enjoy.

(cross-posting to PES as well)

Well that is what I've needed! Some good benchmarks are always helpfull!

This is brilliant work Chesko! Are You planing to do some more performance test of other revisions? Newest scatter cloud svn 186 and XE for example?

Thanks JageriuS! I haven't looked at svn186 at all, if they rewrote some significant portion of the way Distant Lands works then I may. If not I may test the performance of atmospheric scattering (if only for my own curiosity).

XE I plan to test more thoroughly since it's such an overhaul. And maybe this time I will have a more controlled PC environment and not make boneheaded mistakes taking screenshots, opening programs, etc.

Do you have a link to svn186? I don't see it in the Beta files on their Sourceforge page.

It's towards the bottom of the OP in the http://www.gamesas.com/index.php?/topic/1150836-morrowind-graphics-extender-mge/.

Thank you sir. Installing now.

Very interesting. Thanks for all these.

Thanks for these, I find it rather interesting to compare the visual differences between all of the settings, should be added to the MGE thread as a comparison. As should the results of your tests, do you plan on measuring the exact hit of shaders?

Interesting stuff, thanks for taking the time to do all that testing :thumbsup:

Looks like the CPU is the bigger bottleneck to me.

This would make sense if you think about it, since as you increase the draw distance, the number of draw calls will increase dramatically, plus the impact of visibility testing of statics etc.

It would be nice if the test tried to limit VRAM consumption under 512 MB and selected draw distances were 6,8,10,12,14,16. Mesh reduction has a huge impact on performance in my tests for example which can't be seen in results.