Loading...
  OR  Zero-K Name:    Password:   

Low FPS on High End System

31 posts, 2720 views
Post comment
Filter:    Player:  
Page of 2 (31 records)
sort
4 years ago
Hey guys, curiously I just had a 2 v 3 match on a 22 x 12 map and had suprisingly low fps in late game, especially considering it was only hovering around 50% usage of both RAM, CPU and GPU.

System specs:
i5 9600K 6 core ( I can see the game is utilising all 6 cores)
1080Ti
32GB DDR4 RAM
Game installed on SSD

I run live monitoring of all my hardware throughout the game and it's not getting close to maxing out any of these, yet I'm getting down to 20 and even 15FPS, what gives?

Any way to improve this?

+3 / -0
Firepluk
ZK been feeling not its best recently - I as well observe huge lag spikes and FPS drops to 1-2 FPS for a second or more in big team games... Most of the big games(8v8+, 16x16+) become unplayable like after 10-15min
I heard CZrankpsaniac and at least 3 other people complaining of low fps and game unresponsiveness.

GTX 1060, 32gb, i7 2600k, windowz. On linux this problem is even worse(proprietary)
+0 / -0
Link replays and give times. We have been updating the engine recently so maybe it is worth making a regression benchmark. It could also be due to meta changes as I have seen more spammy units like Tremor and Nimbus.
+0 / -0
Firepluk
Surprisingly me and a couple of others had below 30 FPS almost right from the start of this game(first seconds)
http://zero-k.info/Battles/Detail/748803
Gets much worse as game progresses further(<10)

I would not count -much- on replays only though, because it's often a different story performance wise -watching- replay and -playing- a game(network I guess)
+1 / -0

4 years ago
AUrankLord_Squirty What operating system are you using?

If you are using Linux, make sure you are loading the right driver.
+0 / -0


4 years ago
If your connection is fine then I don't expect there to be much of an effect due to networking. If your connection is not fine then you may get stuttering and catchup due to missing out on information from the server. This is fairly simple to rule out. I expect the main difference between playing and spectating to be drawing the UI and selection shapes associated with selecting units.

If you use any non-default widgets then you may have poor performance in packed games because Deluxe Player List is terrible.

I have a hunch that performance is more related to how much fighting is happening than it used to be. Tremor is a culprit as it spams projectiles and you can see a bit of a jump in terraform CPU usage when a few of them are smoothing an area. A few months ago I made the projectiles in ZK be tracked by lua a lot less, which improved performance, but I had to revert the change because it broke buried Stinger LOS reveal. I've just implemented LOS reveal in a better way so can reapply the projectile non-tracking, so we'll see if there is a change next update.

https://github.com/ZeroK-RTS/Zero-K/commit/57e271cf3d5a518c055a8eb1adc764ae4436394c
+2 / -0
Firepluk
I've tried watching replays of big team games multiple times - I almost never ever get those horrible lag spikes which I have during actual playtime and average FPS is almost always higher.
It's either network related(lot's of player in such games have their CPU/Ping in the red zone on player list) or it's indeed selection or even LOS related(when spectating happens LOS calcs aren't done)
P.S. my network should be fine, fiber optics 100mbit, channel always idle when playing zk(ping to zero-k.info ~60msec)
+1 / -0


4 years ago
Watch a replay, do /specfullview on yourself, and then randomly select units and give orders.

quote:
when spectating happens LOS calcs aren't done
This is false.
+0 / -0
Firepluk
Did just this on replay, I start with 100+ FPS contrary to ~30-40 FPS start when I played this game and it never goes down below 30, average FPS is like 2 times higher than that in real game...
there are slight drops when I randomly mess with user input but nothing horrible
+0 / -0
Do /cheat and /team X to get onto your original team. Then /luaui reload to make lua none the wiser.
+0 / -0
4 years ago
Thanks for the replies guys, to answer some questions;

OS is Windows 10

AM playing AI only matches so network speed shouldn't be an issue at all, and my connection is good , couldn't tell you actual speeds but everything i download is 10MB+/s (just in case the game still interacts with online servers and network could still somehow be an issue)

What I'm more interested in is how it can be slowing down when nothing appears to be bottlenecking or reaching full capacity? Can anyone riddle me that one?
+0 / -0
I'am having huge fps drops too, since maybe one or two weeks ago? For example... I've noticed fps drops (more like 1-2 second freezes) while missiles from four Catapults exploded on Fields of Isis (10+ player game).

4690K@4.6GHz, 16GB DDR3@2133MHz, 1060 3GB, W10
+1 / -0
Firepluk
On this big team battle I had FPS drops to 0.3-0.5 FPS...
http://zero-k.info/Battles/Detail/749313

When I or swappan scouted - game almost stopped due to insane FPS drop(17-18+min mark and forward)
swappan complained he could not control his units... honestly neither could I...

Most of players had red CPU/ping after midgame on playerlist
+0 / -0


4 years ago
Everyone here should run a benchmark. I heard a rumor that the problems started 2-3 weeks ago, so I have updated the benchmark to compare the current engine with the engine that was in use from June 20 to July 3.

A game update would be required to deploy the benchmark to everyone, so for now unzip the standalone version and follow the instructions here: https://github.com/ZeroK-RTS/Zero-K/issues/3668
+1 / -0
I too suffer in almsost every game by a sudden and huge 3 second lag spike every 2-7 minutes
Edit: I'll run a benchmark once I go to bed.
Edit2: I've had this problem for more than 3 weeks. Like an entire year or so.
+0 / -0
This is looking good. The two benchmarks that I have received so far indicate a detectable regression between 104-1251 and 104-1305 for some hardware.

Here is the data: https://docs.google.com/spreadsheets/d/1JBBEW-rSY4oWJSEWge4fJbPtBSFMTyVB99xW1jvodWw/edit?usp=sharing

I've created Benchmarkv3a to probe the region between 104-1251 and 104-1305. As before, unzip it to your install and run a benchmark through Help -> Benchmark. https://github.com/ZeroK-RTS/Zero-K/issues/3668
  • Basic benchmark bisects on 104-1278
  • Quick compare compares a multitude of versions between 104-1251 and 104-1305.
  • Long compare is the overnight version of quick compare.
+0 / -0
4 years ago
My benchmark's been running for over 12 hours now, and it's supposed to be 7 hours long, what do
+0 / -0


4 years ago
What does running look like? It should be starting and stopping new Spring windows. If nothing appears to be happening then a download broke. Otherwise, perhaps you computer is particularly slow.
+0 / -0
4 years ago
just the lobby, nothing else. I'll grab the benchmark v3
+0 / -0
I've made another version. This one test a commit that kloot (an engine dev) says may be responsible and also tests a setting. https://github.com/ZeroK-RTS/Zero-K/issues/3668

When you run a benchmark it should start new Spring instances in a new window and run a high load test case. It can take some minutes to download the engines required for benchmarking when they are run for the first time. If nothing happens for 10 minutes then there may be an issue with the unzip directories or the engine downloads.

Check that the replays referenced in benchmarks/config.lua are in the demos folder. They come in the zip. Also check that the required engines are in engine/<platform><bits>.
+0 / -0
Page of 2 (31 records)