| 22 Mar 2025 |
 Sean | let me double check though | 01:52:57 |
 olivia | I thought the gpu dropping wasn't happening anymore, and the only visible problem is lower fps and input delay? | 01:53:18 |
| olivia | * I thought the gpu dropping wasn't happening anymore (after removing the overclocking), and the only visible problem is lower fps and input delay? | 01:53:32 |
| Sean | well I guess it’s not dropping. that’s just what it looks like | 01:53:41 |
| Sean | no like it’s stuttering to 2 FPS for 1/2 second every 3 seconds | 01:54:10 |
| olivia | the reason to check whether the seh/libunwind messages occur elsewhere is that if they do occur elsewhere with similar frequency, then it's evidence against the hypothesis that they are related to the visible fps/stuttering problems, which occur only in this one place | 01:54:36 |
| Sean | okay right | 01:55:02 |
| olivia | but if they don't, then it points to them being reltaed | 01:55:05 |
| olivia | * but if they don't, then it points to them being related | 01:55:07 |
| Sean | well the stutters correlated with the logs that were in the middle of the semaphore warnings | 01:55:39 |
| Sean | yeah other areas are fine | 01:56:47 |
| olivia | oh... that's potentially significant | 01:57:00 |
| olivia | the "PSO compatibility hash mismatch." logs, or the ones about allocations? | 01:57:32 |
| olivia | the hash mismatch ones probably trigger shader recompilation, so it would make sense if those caused stuttering | 01:58:12 |
| Sean | I think the hash mismatch. it was kind of hard to tell because the entire graphics stack stuttered | 01:58:25 |
| Sean | oh maybe yeah | 01:58:53 |
| olivia | if that's the case, it puts it back into being a likely graphics problem, but unfortunately not an easily solvable one | 01:59:52 |
| Sean | oh okay | 02:00:42 |
| olivia | lemme look at vkd3d-proton for a bit to confirm | 02:01:48 |
| Sean | if you want something that will crash in the same way. the monster hunter wilds benchmark crashes in the final scene like this | 02:02:58 |
| olivia | I can possibly try that | 02:03:21 |
| olivia | I don't have any amd hardware to test with though lol | 02:03:27 |
| olivia | but possible it is reproducible elsewhere | 02:03:35 |
| Sean | oh okay, maybe not worth your time then | 02:03:41 |
| Sean | but yeah | 02:03:43 |
| olivia | /* Verify the expected PSO state that was used. This must match, or we have to fail compilation as per API spec. */
if (compat->state_desc_compat_hash != pso_compat->compat.state_desc_compat_hash)
{
if (vkd3d_config_flags & VKD3D_CONFIG_FLAG_PIPELINE_LIBRARY_LOG)
INFO("PSO compatibility hash mismatch.\n");
else
WARN("PSO compatibility hash mismatch.\n");
return E_INVALIDARG;
}
oh lol, this is not what I was expecting | 02:04:10 |
| Sean | oh so they're just not to spec? | 02:04:38 |
| Sean | and so the driver fails like it should | 02:04:46 |
| olivia | yeah, I'm wondering if this is possibly an application problem | 02:05:11 |
| olivia | I don't know d3d very well at all | 02:05:20 |
