Original Post — Direct link
Hi everyone,

Below is our next quarterly game performance preview. This time, we are starting with a recap of Update 27 performance improvement effectiveness and next steps, followed by a preview of what’s coming for Update 28 in November. As with our last game performance preview, this includes details from our Engineering team.

Thank you, and if you haven’t yet had a chance, we also encourage you to check out the Update 28 Combat Preview here.


Update 27 Retro

Cyrodiil AoE Ability Testing: The AoE ability performance tests in Cyrodiil are wrapping up, with the final round of tests concluding on October 19. Out of the first three weeks of tests, the global cooldown test resulted in the greatest performance gains when compared to the others (roughly 25% increase in server performance). However, there is still work to be done as the performance gains, though substantial, are not enough to make the overall experience as smooth as we would like it to be. We will be running more combinations of tests right after this last round completes, and providing a more comprehensive summary of this round of tests. Stay tuned for more info soon!

Here are a couple general notes about the AoE ability testing:
[list]
[*] Our ideal server FPS is less than 30ms. Anything over 50ms and players will see noticeable delays in abilities firing. The global cooldown test brought server FPS in high-intensity situations down almost 25%, but it still spikes above 50 in large battles.
[*] Population remained consistent during peak hours for these tests. (i.e. – On PC EU, Ravenwatch was still pop-capped for all three Alliances)
[/list]
Trial & Dungeon Performance on the Server: We had some issues with configuration settings when we first launched these initial changes that are now resolved. In the past few weeks, things have been stable and running smoothly. We identified a few more adjustments that we need to make, and those should be in for Update 29.

Database Improvements – Activity Finder: We are happy with how well these changes have gone thus far. Getting into dungeon groups is a much quicker and more reliable experience than it has been in the past. There are still a few edge-case bugs we’re tracking, but overall things have vastly improved.

Intermittent Load Screens: The fixes we implemented to address intermittent load screen issues have performed remarkably well. There has been a significant drop in intermittent load screens across the board, in all zones. For example – in Cyrodiil (which was one of the worst offenders) we have seen a 95+% decrease in intermittent load screens randomly popping up in the zone. (We went from approx. 50k a day down to under 3k a day).

Overall Client Stability (critical memory system & Havok physics engine hardening): We have seen an improvement in overall stability due to these initiatives. We have not fixed all the crashes and still have work to do, but our crash rate on consoles has decreased dramatically since U27 launch.


Update 28 Preview
For Update 28, we have the following initiatives coming:

[list]
[*] Large-Scale Combat Performance Improvements: [list]
[*] These include optimizations for how the client updates visual effects and should result in a higher framerate when participating in mass-scale combat with lots of abilities being cast (client FPS improvements).[/list]
[*] Frustum Culling & Render Thread Optimizations:[list]
[*] Improved framerate in situations where there are many individual objects visible, such as crowded areas with many players or within player housing that has a high volume of placed furnishings.
[*] This was done by reducing the workload on the rendering system by caching data and eliminating unnecessary commands.
[*] Note: These optimizations are console ONLY.
[/list][/list]

The Update 28 game performance work noted above is live on the PTS right now and will go into the game in November.

Thank you all for your patience waiting for this quarter’s game performance preview, and we hope this helped illustrate how the Update 27 work went and what we’re working on for Update 28.

over 3 years ago - ZOS_GinaBruno - Direct link
FYI: Just corrected a typo in the post. Originally one of the bullet points stated the global cooldown test brought server FPS in high-intensity situations down almost 45% - this should've said 25%, and has been corrected in the original post.
over 3 years ago - ZOS_GinaBruno - Direct link
Kingslayer513;c-6988323

ZOS_GinaBruno;c-6988315

FYI: Just corrected a typo in the post. Originally one of the bullet points stated the global cooldown test brought server FPS in high-intensity situations down almost 45% - this should've said 25%, and has been corrected in the original post.



Pretty sure that should say "ping" or "latency" instead of FPS. FPS makes no sense in that context and you don't measure FPS in milliseconds.



In this case, we're using FPS to refer to server frames.