search

RPCS3 / rpcs3

PlayStation 3 emulator and debugger
https://rpcs3.net/
GNU General Public License v2.0
15.39k stars 1.91k forks source link

[Ridge Racer 7] Driving behavior of "main car" goes nuts after ~50% of the first lap in CPU-only-cars races (Attract mode race + AV-Player races (replays not affected!)) #9650

Open axbmcuser opened 3 years ago

axbmcuser commented 3 years ago

Quick summary

[Ridge Racer 7] Driving behavior of "main car" goes nuts after ~50% of the first lap in CPU-only-cars races (Attract mode race + AV-Player races (replays not affected!))

Details

The "main car" stops driving correctly after about half of the first lap und starts driving against the walls and stuff. All the other "not main cars" (also CPU-controlled) seem to be not affected - so they drive on successfully, but the camera of course remains focusing the "main car" which has gone nuts. (All courses affected, 100% reproducible in my tests (happens every time without exceptions). The issue starts at the exact same point in the tracks every single time (same corner for example))

I noticed the above for some time now and just had the time to document it with a quick video. I suspect this is not a typical "desync-issue" people with (often) insufficient CPU desribe, but maybe something else. If this is already known, i'm sorry. Just want to help. :-)

See video + log for details

Besides noticing the described issue i only play "Arcade mode" which works awesome with locked 60fps + locked frametimes on all courses. Many thanks for that! It's a much greater experience then on real PS3 due to the better visual quality. :-)

RPCS3's log

https://www.dropbox.com/s/m1c4i1llglzbesx/RPCS3.log.gz?dl=1

Video documentation

https://www.dropbox.com/s/oyftba8dl72ol9a/2021.01.24-16.37-2.m4v?dl=0 (@ timcode: 1:30 minutes)

Please provide comparison with real PS3.

Does not happen on real PS3 (tested with identical disc-data).

Please provide your system configuration:

Whatcookie commented 3 years ago

It's probably an issue with floating point accuracy, you should see the same kind of problem if you recorded a replay on real hardware, and then tried to play back the replay on RPCS3.

axbmcuser commented 3 years ago

@Whatcookie Thanks for you reply. Anything we can do about it? Or do you suspect it to be a bigger, maybe conceptional problem which (maybe?) can't really be fixed? Of course i already experimented by enabling "accurate xfloat" and the two "accurate"-options under "Advanced" with same results.

As described, the stuff that works (Arcade mode) runs great and frametime-stable and my 12T-CPU does run with low workload and it looks like it can handle the game quite good. I suspect if your hunch is correct the issue won't be fixed just by upgrading from a 12T to a 16T CPU or something like that - correct? Regarding replays: As mentioned, my replays are not affected by the issue in my tests. Would it still be helpful if i got a replay from one of my real PS3s to work on RPCS3 to work to to see if your suspicion regarding transferred replays from real hardware failing is correct? Or did you maybe just overread the info "replays are not affected"?

Thanks again :-)

Whatcookie commented 3 years ago

The issue should be solvable, but the problem is that it's hard to solve, and the solution will be slow.

If you could grab a replay from real hardware, that would be helpful.

axbmcuser commented 3 years ago

Just tested 2 replays with "clean races" from real hardware and tried them on RPCS3:

I can confirm your suspicion! My "real hardware clean races" are shown in exactly the same broken "main car goes nuts"-way when they are played on RPCS3.

I guess now i (maybe) understand the technical idea behind it some more: Replays recorded on RPCS3 itself play fine on RPCS3 because replay "recording" and "playback" accuracy is identical. The "Attract mode race" and "AV-Player"-races(not own replays) are both - so to speak - "replays recorded by the developers on real hardware + baked into the game" and therefore fail the same way as any replay recorded not on RPCS3 itself would fail. (?)

Back2Life888 commented 3 years ago

same thing with RRPSP and Henrik's PPSSPP.. Its a CPU calculation problem.. This will take a while for this to be fixed because Henrik has trouble with getting the bug fixed.. RPCS3 devs will probably encounter the same thing of having trouble with fixing it..

axbmcuser commented 3 years ago

Good to know, thanks. Could you please post the exact corresponding issue/PR/thread for PPSSPP? Maybe that'd help to avoid wasting time on non working approaches while trying to solve it for RPCS3.

isJuhn commented 3 years ago

The PPSSPP issue is https://github.com/hrydgard/ppsspp/issues/2990

benderscruffy commented 1 year ago

RPCS3 v0.0.28-15409-fd6829f7 Alpha | master still an issue