rpcs3 icon indicating copy to clipboard operation
rpcs3 copied to clipboard

Test Drive Unlimited 2 [BLUS30527]: Freezes at title screen [Intel CPU only]

Open Shoegzer opened this issue 3 years ago • 12 comments

This game always freezes the first time RPCS3 builds the shader cache. After the first time, it progresses to the Eden title screen at least, but then it freezes no matter how many further attempts are made. Curiously, RPCS3's main menu toggles from "Pause" to "Play" upon freezing.

This issue was also partially discussed in #10876 but moved here for better tracking. As @xddxd reported, the game has been known to test past this point on Windows systems.

Test information

  • OS: Linux Mint 20.2 / kernel 5.11
  • CPU: Intel i7-10870H
  • GPU: Nvidia GTX 3070 / driver 470.63
  • Non-default settings tested: Driver wakeup delay set to 600, 1000 and 2000us
  • Commit tested: ddec5d6908938d7509120b73d8e92f7598cd9e33
  • Logfile: RPCS3.log.zip

Shoegzer avatar Sep 17 '21 16:09 Shoegzer

Try booting the game with Intel HD Graphics as the adapter.

xddxd avatar Sep 17 '21 17:09 xddxd

Thanks. For some reason nothing shows at all after booting with Intel HD Graphics (won't progress to title screen sequence, stuck at FPS 0.00). I tried clearing all caches and setting driver wakeup delay to 1000 but with the same results.

Shoegzer avatar Sep 17 '21 17:09 Shoegzer

It's a linux bug. Happens here as well, European version, AMD GPU, while it works fine on Windows. Happens regardless if on Vulkan or OpenGL. RPCS3.log

xddxd avatar Sep 17 '21 18:09 xddxd

Update: Tested this again after a while. Same issue it seems, though I noticed that now rpcs3 crashes rather than freezes after the Eden splash screen. I looked through the log and can't find anything out of the ordinary, has me wondering what linux-specific thing this title is doing that others aren't? At least we know it's not GPU-specific given @xddxd sees it happening on AMD too. I did catch this in the logs, perhaps a clue:

·W 0:00:42.174553 {PPU[0x100000a] Thread (RenderThread) [libsre: 0x019138a0]} sys_semaphore: sys_semaphore_create(sem_id=*0xd00bdba0, attr=*0xd00bdba8, initial_val=0, max_val=1)
·W 0:00:42.174559 {RSX [0x21009e0]} RSX: Could not gather textures into an atlas; using CPU fallback...
·W 0:00:42.174567 {RSX [0x21009e0]} RSX: Cache miss at address 0xC2D7C000. This is gonna hurt...

Here are specs that have differed since first test and an updated log in case it helps:

  • OS: Linux Mint 21.1 / kernel 5.15
  • GPU driver: 525.85
  • Commit tested: 0beda6fa8943608c73ae3ad7f5457a444ffc47ea
  • Logfile: RPCS3.log.zip

Shoegzer avatar Mar 18 '23 19:03 Shoegzer

Update: Changed title based on newer observations - it seems more likely this is related to CPU and not OS. @xddxd can you help confirm?

Shoegzer avatar May 27 '23 16:05 Shoegzer

Update: Tested against latest master (commit 8772219492118f538e4b6b97e5449c4f3fc66aca) as a few PRs that help with booting were recently merged, but unfortunately it seems to have regressed somewhat (won't even boot to the Eden title screen now, even after all caches were removed). Would be great if someone could confirm this as an Intel CPU issue as well.

Shoegzer avatar Aug 25 '23 02:08 Shoegzer

After debugging with GDB the game returns a segmentation fault with some functions:

crash gdb.txt

rpcs3 log:

RPCS3.log

configuration:

https://wiki.rpcs3.net/index.php?title=Test_Drive_Unlimited_2

tested with both oGL and VK

Fabxx avatar Oct 22 '23 14:10 Fabxx

Thanks @Fabxx I hope your GDB log helps. I also noticed you're using an Intel CPU similar to mine (i7-10700 vs. i7-10870H).

Shoegzer avatar Oct 22 '23 17:10 Shoegzer

Update: further evidence of this can be found in #14837, which involves an AMD CPU. Tests involving an Intel CPU have not progressed to the point reported there.

Shoegzer avatar Nov 25 '23 23:11 Shoegzer

RPCS3.log

Posting a new log based on the lastes version. Menu graphics appear to be corrupted as of now.

Fabxx avatar Feb 11 '24 15:02 Fabxx

I don't own an intel CPU, so I'll unassign for now. It also points to a potential CPU issue, so it makes sense to let the CPU guys take a look.

kd-11 avatar Feb 28 '24 21:02 kd-11