Preliminary Performance Improvement Results

ironically the multithreading test gave me markedly worse performance: dropping from a stuttery-if-playable 20fps in kaom's dream on minimum settings to a blisteringly fast 6fps standing still, without enemies onscreen in the same location.

"

AndIMustMask wrote:

ironically the multithreading test gave me markedly worse performance: dropping from a stuttery-if-playable 20fps in kaom's dream on minimum settings to a blisteringly fast 6fps standing still, without enemies onscreen in the same location.

Sounds like you are GPU bound. In the post where they announced this patch they said those who were GPU bound should not expect increase.

will there be some info here about the coming improvements like the 64 bit client? or will this info stay exclusively on reddit?

i'm still trying to figure out for myself how the 64bit client's ability to cache more assets will interact with the caching mechanism you just introduced that dynamically loads assets to use less memory?

the only advantage i see is that hdd users who run with --noasync will profit from the increased buffering ability and pay for it with even longer loading times...

Good. Now fix the XP penalty on red maps and I might have a reason to start playing again.

Its weird, yesterday when using this it felt better, today it feels like its back to the same.
Did ggg change or remove the fix, maybe until 2.4?

I am running this game on 2 cores 4 threads intel i3 3110m and these are the fps benchmarks i made running dried lake and builds that can squeeze out the most of my laptop, without completely freezing

This one is with my selfcast Arc character 6.5 casts per second


This one is with my Mjolner Elementalist using Lightning strike as main proc skill


As you can see the multithreading gives me higher Min/Max fps but lower average with the Arc character, and overall better performance in the second case. Dips still exist but are smoother, the fps is not all over the place.
No crashes or bugs so far.

Conclusion: Good job GGG.

Side note: Im running --noasync.I also tried running just --experimental-frame-optimizations with async loading and the significant improvements are just in town and load screens are even faster but other than that overall gameplay is still unplayable for me.(hitting 0 fps with these two builds especially the mjolner one is completely unplayable)

Running an i3 6100 - this has two cores, but 4 threads, would I see an improvement using this CPU?

"

Note that these averages include all gameplay. The improvements are more noticeable (and represent larger percentage increases) under extreme combat than when idling or running around.

To more accurately judge the improvements, we should examine the variance of the frame rate rather than just the average frame rate. We'll keep collecting data and will share anything interesting we find!

4% user here and I found that I have faster recovery from critical frame drops than without. Even at single digit frame rates, I can at least control character and get recovered quickly. Great job !

Havent crashed once (YET) but not only do I not see any boost, quite the opposite :(
barely playable on a not-so-laggy build (Windripper nado) with like 10 fps avg in maps WITH the beta line. 20-30 without... Tested on just my laptop though, just came home to my PC soo its time to patch, paste the line here too and see how this one turns out.