so, quick update, while fixing up with a different wine version did not fix the issue, surprisingly fixing up mesa drivers did. I can warn you, its gonna be a pain in the ass to get everything right. but yeah, updating to mesa 23.3-current from the git repo made the 100% cpu issues go away(at least after testing it for a couple hours)
1 Like
quick followup. I used this script to facilitate building:
It is an old script, parts of it dont work any more. you will need to add your distro(Debian 12 in my case) in the branch and set gcc to the appropriate version and llvm needed to be version 15 in my case. the script will install all dependencies - if you are on debian, it will fail trying to install llvm with debug symbols - just remove the line in the script trying to install it.
it also tries to build the swr-renderers, which dont exist any more, so fix COMMON_OPTIONS accordingly.
finally, git checkout will fail because the script sets a branch depth to checkout, which fails on my system with too many arguments to git, so i removed the depth.
After building everything, mesa-tests failed(from the installer script), however after a reboot of the system i was sucessfully able to run eve with the new dev-version of mesa.
to do so, I used the following wrapper script:
cat trymesa.sh
#!/bin/sh
export MESA32=$HOME/mesa-git/installdir/build-i386-opt/install
export MESA64=$HOME/mesa-git/installdir/build-amd64-opt/install
#export DXVK_HUD=full
export LD_LIBRARY_PATH=$MESA32/lib:$MESA64/lib:$LD_LIBRARY_PATH
export LIBGL_DRIVERS_PATH=$MESA32/lib/dri:$MESA64/lib/dri
export VK_ICD_FILENAMES=$MESA32/share/vulkan/icd.d/intel_icd.i686.json:$MESA64/share/vulkan/icd.d/intel_icd.x86_64.json
export D3D_MODULE_PATH=$MESA32/lib/d3d/d3dadapter9.so.1:$MESA64/lib/d3d/d3dadapter9.so.1
echo $VK_ICD_FILENAMES
exec ā$@ā
I invoke this script like this:
WINEPREFIX=/home/brightsky/Games/eve-online ./trymesa.sh /opt/wine-staging/bin/wine Games/eve-online/drive_c/EVE/eve.exe
Note that you will have to change the directories accordingly(the mesa script builds to ~mesa-git on default, so mostly you should only need to adjust your username in /home/$u/
to see whether you are actually using mesa 23, uncomment the line saying DXVK_HUD=full in the script, and you will get an overlay with fps, driver version and some stats.
While at it, I am now also running eve directly, aka not via lutris anymore. saves a few megs of memory.
I hope this helps someone, fly safe and without 100% cpu load
final note: I do only have an intel iris gpu, so i removed all the build stuff for the radeon driver from this script. this might also explain why i only load the icds for intel. be aware that you need 32 and 64 bit versions to run eve because the launcher is still a 32bit pe executable(WHY THE F CCP?)*
!!!WARNING!!! DO NOT DO THIS UNLESS YOU KNOW YOUR WAY AROUND YOUR SYSTEM - THE PROCEDURES OUTLINED ABOVE ARE DONE IN A WAY THAT SHOULD NOT PERMANENTLY CHANGE ANYTHING IN YOUR SYSTEM, BUT SHOULD YOU CHOOSE TO MESS WITH YOUR MESA INSTALLATION YOU CAN BREAK X11(GUI) BEYOND AN EASY REPAIR
1 Like
Nice investigative work. 
Iām trialing Mesa 23.3 via this PPA on my Linux Mint box: Updated Open Graphics Drivers - since 2011! : Oibaf
Ill update if I run into performance issues again.,
hmm, still having issues running more than one client. Single client performance is great. Im at a bit of a loss now.
Seems like in my case falling back to Proton 6 fixed issues. Playing for 15 minutes and CPU usage is under 3% and not rising to 6-8-10-13-freeze.
P.S. CCP support replied that according to logs i sent thereās no issue with Eve client but with Wine (not sure it is).
P.S.2: i was wrong, its rising anyway but not so fast, sorry.
It looks like there is a new fix in the latest version of Wine-GE in Lutrix and Proton Experimental for some CPU+GPU issues: Release Wine-GE-Proton8-12 Released Ā· GloriousEggroll/wine-ge-custom Ā· GitHub
Honestly it seems like CCP did sth to fix that in general. After updating to mesa-current with stock(repository) wine-staging 8.10 it was better throughout the last week but the CPU crap still happened every now and then. I was out moon mining the last two days however and it was just fine. I didnāt change or update anything on my system, so I dunnoā¦ Keep us posted whether wine proton-ge fixes it for good
Ehā¦ I did some testing this afternoon. I was able to run two clients for about 40 minutes before starting to hit unplayable levels of fps/performance. Which is an improvement over what I was getting.
Its like there is some garbage collection not happening somewhere while changing systems/ jumping gates that eventually grinds everything to a halt.
Iām using Steam + Proton Experimental + Ryzen 3 CPU + nVidia graphics. EVE runs fine solo, but when I join a big fleet (200+) CPU goes up to 100% and stays there. Itās been a problem ever since the Viridian expansion. Before that, I could multibox and everything with fleets.
Also enable DirectX 11 in launcher and enable āDownload Full Clientā. It prevents some jitters solo.
Steam+proton experimental.
I see performance degradation when doing PI, after going into planet view for 5-7 times the fps will go to ā ā ā ā . This issue started 3-4 weeks ago.
very likely - I do run up to 11 clients in parallel where 10 mostly stay in one system (industry/cyno) and one account that continuously moves around. The moving account has growing memory over time without any recognizable limit, since I have 128GB it can use up ā¦ highest I ever had it grow was to roughly 25 GB after approx. 10hrs before I had to restart as it became unplayable.
One thing that still grinds the whole client to a halt and sometimes leads to freezes/crashes ā¦ is industry on scale. Loading 25.000 blueprints (the limit of the client - its a fixed limit) for a structure just freezes the client for up to multiple minutes. And the same happens every time a client watches those blueprints in the industry window and some other client/player adds/removes/uses blueprints. This also affects contracting large amounts of blueprints, trading them, etcā¦ The FPS just hits somewhere between 0 and 1 and everything freezes.
1 Like
Anyone have Eve working with 7900 XTX GPU?
I have Eve working on one of my PCs that is running Zorin (Ubuntu 22) and has a 6900 XT GPU instead. This one is a fresh install and the CCP logo pops up on Steam, then it never does anything else. I tried Ubuntu 22 as well, and itās doing the same thing. I suspect the GPU is the issue, but would be great to hear if anyone has encountered this issue before.
UPDATE: I got it working! I uninstalled Steam from Ubuntu Software and installed from the .deb file from steampowered.com.
2 Likes
Update: I replaced the Ryzen 3 (1300x) with a Ryzen 7 (5700x). EVE is running a lot better, with no 100% CPU issue. I havenāt tested large fleet situations yet but so far so good.
There is a possibility Steam updated Proton Experimental, or there was a flatpak update that somehow fixed the issue. Will update if the issue comes back.
FYIā¦ CCP as always messing up CPU usage on EVE clientsā¦ I used to run 5 clients with my Ryzen5 3600ā¦ with a cheap RX460 4GB (the one single slot without power plug) and 32gb RAMā¦ now I upgraded the GPUā¦ same polaris family, but the best in that classā¦ the RX480, same VRAM, upgraded RAM to 48GBā¦ and nowā¦ guess whatā¦ with a better machineā¦ I can only run 3 clients (whereas before I could run 5 with no temperature issues on the CPU)ā¦ now I run 3 and Im close to 70C. Makes no senseā¦ other than CCP messing cpu client usage.
Im still running with Steam, with Proton 8.0. Itās the same install I had with the cheaper GPUā¦ so you would think I would be able to run maybe 6 clientsā¦ but no, now is 3.
Donāt know what CCP changed recently but the current plex offer made me open the New Eden Store which breaks the proton layer or something because my fps will go down to <1 and the only way to recover is by restarting the client.
Tried proton 7 and latest steam experimental, also reinstalled Eve but i get the same results.
Before opening the store
and after
The new beta launcher went into effectā¦ and it wont loadā¦ so, Im no longer being able to play ;O (my last working EVE was using steam with proton-8 and working everything āfineā (except the high cpu usage on each client).
This is the result from a strace on my main client. Looks like a lot of CPU calls going to āfutexā. Maybe a clue here.
strace -c -p 51765
strace: Process 51765 attached
^Cstrace: Process 51765 detached
% time seconds usecs/call calls errors syscall
76.15 37.078009 22 1666108 243484 futex
8.77 4.270163 864 4937 pselect6
8.14 3.965452 29 134967 sched_yield
1.60 0.781410 7 98247 read
1.23 0.597935 4 121508 rt_sigprocmask
1.08 0.526064 5 98157 79409 recvmsg
0.83 0.404074 40 9927 poll
0.67 0.326769 7 43246 write
0.36 0.175848 14 11926 openat
0.32 0.155712 7 21211 writev
0.21 0.100301 8 12063 close
0.17 0.083560 7 11544 newfstatat
0.16 0.075522 18 4065 3593 futex_waitv
0.09 0.043612 13 3122 834 stat
0.06 0.028885 132 218 munmap
0.06 0.028426 5 4741 lseek
0.03 0.013471 20 653 653 getxattr
0.02 0.010197 13 774 getdents64
0.01 0.004756 7 653 lstat
0.01 0.003823 9 400 192 ioctl
0.01 0.003781 71 53 pread64
0.01 0.003321 13 255 fstatfs
0.01 0.003152 5 551 mprotect
0.01 0.003147 11 268 mmap
0.00 0.000982 17 56 clone3
0.00 0.000674 84 8 brk
0.00 0.000627 3 160 fstat
0.00 0.000609 10 56 sendmsg
0.00 0.000600 10 56 pipe2
0.00 0.000438 5 85 fcntl
0.00 0.000278 2 106 106 fgetxattr
0.00 0.000095 6 15 recvfrom
0.00 0.000050 1 29 times
0.00 0.000047 3 15 getrandom
0.00 0.000003 0 4 fchdir
0.00 0.000000 0 2 rt_sigreturn
100.00 48.691793 21 2250186 328271 total
Ive had some positive results in Lutrus turning off fsync and just using esync. Ill keep everyone updated.
using ubuntu
base wine
wine --version
wine-8.12 (Staging)
with dxvk 2.2
I reverted to nvidia server 525
xserver-xorg-video-nvidia-525/lunar-security,lunar-updates,now 525.125.06-0ubuntu0.23.04.1 amd64 [installed,automatic]
because server 535 had random crashes.
with vsync off on jita undock:
it seems like anti-aliasing:high and shadow:high uses a lot of CPU (I hear the fan) so I turn it off.
with max settings on jita undock
I remove logs to file because on my previous rig it produces HIGH load after a while, like the log was in-memory and rewritten every 2s instead of flushing the new entries. Made the game literally unplayable afterā¦ 1h ?
1 Like
Noticing similar issues recently when multiboxing (i run 10 clients). CPU usage and temperature extremely high even for clients that are idling in station or with gfx turned off (ctrl-shift-f9).
Iām having some success throttling fps to 30 with mangohud. On steam āset launcher optionā:
MANGOHUD_CONFIG=no_display,fps_limit=30 mangohud %command%
My CPU usage and temperature are lower now.
1 Like