csgo-osx-linux
csgo-osx-linux copied to clipboard
Long map load times after map change
Your system information
CPU: AMD Ryzen 5 3600 6-Core Processor
OS: Windows 10 LTSC 21H2 (64 bit)
GPU:
Model: AMD Radeon RX 6600 XT
Driver version: 31.0.12019.5007
Date of release: 8 10 2022
RAM: 16310 MB
- Have you checked for system updates?: [Yes/No] Yes, latest GPU driver. No system updates avail.
Please describe your issue in as much detail as possible:
Don't know when, but an issue appeared with long map loading times. I loads just fine the first time you run the game, but on second change it hangs on "Getting server info". System responsive, but csgo process is in "not respoding state". I've measured loading times: its 1:48 till "Getting server info" changes, total load time: 2:04. In fact, you join the game when 1 round is already started.
There are several reports of this happening on PCs with 6xxx radeon cards, and not happening on previous lineups: https://www.reddit.com/r/csgo/comments/kv0q2d/csgo_stuck_on_retrieving_server_info_after/ https://community.amd.com/t5/graphics/rx-6600-xt-1-minute-long-load-times-in-csgo/m-p/485238
Common solution has been adding "-disable_d3d9ex" to launch options, which is un-optimal by many means (makes alt-tab a suffering, etc). I see people waiting for a fix from AMD driver team, but maybe something can be done by CSGO devs to tackle this issue?
Steps for reproducing this issue:
- Run game normally.
- Join any casual match.
- Play till the end of map.
- Second map will not load properly for a long time.
Actually, I've found out that you don't even need to connect to community server.
Just:
map de_dust_2
Game loads fine.
disconnect
map de_inferno
Game gets stuck on getting server info.
I have the same issue and have done a lot of googling. As stated it seems to be AMD 6xxx cards and the fix is to use "-disable_d3d9ex" and to set "Model / Texture detail" to "Low". Unfortunately, using the workaround causes occasional crashes so it's not idea either.
I am using a 6600. I do not have these issues in Linux though.
Any progress or recommendations on this one? It's quite an annoying bug.
Most relevant discussion here: https://community.amd.com/t5/graphics/amd-csgo-loading-bug-on-current-graphics-hardware-drivers/m-p/473670/page/3
Also I confirm the crashes with -disable_d3d9ex option, so it cant be accepted as a solution.
New AMD driver - 23.2.2. The issue is present.
Seriously, Valve, you can tackle this issue by directly contacting AMD. Its almost 2 years with a certain game being broken.
New build, confirm. I have the same issue. Although I believe this is a AMD issue. Not Valve :(
Build: 5700X 6700XT 64GB RAM 1TB NVME
Same issue as everyone else. Tried so many things. Thinking about moving to Linux. But looks like the same issue exists on Linux as well.
Issue persists. Recently bought a Sapphire RX6700 coupled with R5 5600X, 16GB of RAM and 1 TB SSD. Adding "-disable_d3d9ex" to custom commands makes the game crash.
I was shocked when I saw that this problem was a thing since almost two years now, this is ridiculous.
Guys at this topic found a working solution - enable ReLive in drivers settings.
It stops the card from sleeping during map change and makes the change fast again.
Please report if it works for you, for me - it did.
@braintweak Enabling ReLive seems to have worked (Sometimes), when it doesn't doing CTRL+Shift+E to start a record, seems to wake up the GPU, or Alt-Tabbing. Which is ridiculous. But I'll take it I guess. Since it seems this bug will never get fixed at this point.
@mspencerl87 I really hope the bug is not present in CS2, so yeah, fixing it is a low priority now.
I found out that this Issue might be related to Low VRAM-Clockrates during Loading. When "forcing" the VRAM-Clocks to be higher during loading the game loads just as it should. That's also the reason why lots of users don't might have this. Using higher Screen-Refreshrates cause GPUs to increase their VRAM Clockrate and therefore the game loads normally. You can see your clockrates (on LINUX) by typing cat /sys/kernel/debug/dri/0/amdgpu_pm_info in the terminal. If you have like 96Mhz on MCLK you might suffer from that issue.
@CaptainCoward Not sure if that's the case. I have a 144MHz refresh monitor on Windows. I've tried various clock settings. To me it could be VRAM related. But I'm not sure. As the only work around I've found is before a map loads. Using AMD Re-Live to start recording to keep the GPU active long enough for the map to load. Then once Loaded I stop recording. This works every time. While its quite annoying. It will suffice, as I don't think this will ever be fixed.
I'm curious if is issue exists on CS2, I don't have beta access to test.
Using AMD Re-Live to start recording to keep the GPU active long enough for the map to load. Then once Loaded I stop recording.
You don't need to record manually, just enable instant replay for the lowest amount of time. That's enough to keep the card active.
@braintweak does it lower performance the other 99% of time? As opposed to only recording during map chagnes?
@mspencerl87 never experienced any performance degradation due to instant replay enabled, but it's on windows. Don't know how it's on Linux.
Not an issue on current cs2 build.