Star Ocean 3 - Till the End of Time [SLUS 20488] (U)
#1
status=5;
version=1.5.0;
console=PS2;
crc=23A97857;
img_t=empty;
img_f=empty;
wikiurl=Link;
Code:
empty
The beast is shown @ www.x-homepage.nl
Reply

Sponsored links

#2
PC Specs:
CPU: Intel Core i5-3570K @ 4.2GHz
GPU: Nvidia GTX 560ti 1GB
Windows 10 x64

Build Description:
PCSX2 1.5.0-2388 - Windows

BIOS Used:
USA v02.00(14/06/2004) Console

EE/VU Clamp modes:
Normal/Normal

EE/VU Rounding:
Chop/Chop

GSdx Settings:
OpenGL (Hardware), 4x Native

GSdx HW Hacks:
Half-pixel Offset: Special (Texture), Round Sprite: Half

Speedhacks Used:
Preset 2 - Safe (faster), fast CDVD

Gamefixes Used:
VU Add Hack (Automatic Gamefixes)

Amount of testing done (little/medium/much/completed-game):
Completed (ish... didn't do 4D difficulty and only 69% trophies, ~200hours)


Performance:
60fps using 3x native res on the 560ti. Tested mostly with 4x Native which would slow down when buffer effects are happening.
When using the Hardware renderers on GSdx need to make sure "CRC Hack level" is set to either Automatic or higher than Minimum, it does skip shadows that are way too slow otherwise however.


Comments:
PCSX2:
Widescreen mode is enabled by keeping the Triangle and Cross buttons pressed while the game is booted and then selecting the 16:9 option, this also enables Progresisve mode.
"Automatic Gamefixes" enables the "VU add hack" necessary for booting the game.


GSdx:
"Half-pixel Offset: Special (Texture)" HW hack works for keeping the buffer effects aligned correctly when using upscaling without other side effects.
"Round Sprite: Half" HW hack works well for the vertical lines that appear when changing areas while using upscaling but only fixes it on OpenGL (Hardware).
The effect when going into battle seems to only show properly with Software renderers.
Some shadows are skipped by CRC hacks due to performance hit and some extra glitches when upscaling. They show correctly on Software renderers. OpenGL (Hardware) with native resolution also shows the shadows fine but the speed hit is too great and there's extra glitches if upscaling, Direct3D11 (Hardware) has some of the shadows being misaligned on top of the issues with OpenGL.

I mostly tested with Direct3D11 (Hardware) due to Vsync issues but would recommend OpenGL (Hardware) when upsclaing since it does a better job with accuracy (shadows and vertical lines).


Lilipad:
It is necessary to set up an extra key for Circle button with sensitivity set at 0.500 for the Dragon Flute puzzle at Ruins of Barr. This puzzle requires the use of half-pressure inputs to progress thru the area.
You could also try this method if you have a DS3 controller with sensitive pressure buttons for the proper way to progress.
To be fair, even on the PS2 and with the proper controller, people were often lost on how to pass this part anyway.


This happened a couple times but couldn't reproduce it, might have just been the game Tongue2    


Status
Fully Playable


Attached Files Thumbnail(s)
                       

Core i5 3570k -- Geforce GTX 670  --  Windows 7 x64
Reply
#3
I ported the palette code from opengl to d3d11 on v1.5.0-dev-2395 which improve the shadows so opengl and d3d11 should look the same.
CPU: I7-4770 3.9GHZ
Motherboard: Asrock B85M - DGS
RAM: Hyper X Savage 2x8GB 1.6GHZ CL9
GPU: GTX1070 8GB GDDR5
OS: Windows 10 Pro 64bit
Reply
#4
Just tried it with 1.5.0-2413, it seems to work but only for the shadows that get skipped with crc hacks. The uhhh NPCs only have misaligned shadows on D3D11 but fine on OGL

OpenGL with crc hack level none
   

Direct3D11 hacks none 1.5.0-2413
   

Direct3D11 hacks none 1.5.0-2388
   

Direct3D11 hacks auto (same in both cases)
   

could be a different issue?
Core i5 3570k -- Geforce GTX 670  --  Windows 7 x64
Reply
#5
Yeah that's how it works right now on GL and D3D.

There's an explanation why the crc hack is needed/the slowdown issue.

Code:
// The game emulate a stencil buffer with the alpha channel of the RT
// The operation of the stencil is selected with the palette
// For example -1 wrap will be [240, 16, 32, 48 ....]
// i.e. p[A>>4] = (A - 16) % 256
//
// The fastest and accurate solution will be to replace this pseudo stencil
// by a dedicated GPU draw call
// 1/ Use future GPU capabilities to do a "kind" of SW blending
// 2/ Use a real stencil/atomic image, and then compute the RT alpha value
//
// Both of those solutions will increase code complexity (and only avoid upscaling
// glitches)
CPU: I7-4770 3.9GHZ
Motherboard: Asrock B85M - DGS
RAM: Hyper X Savage 2x8GB 1.6GHZ CL9
GPU: GTX1070 8GB GDDR5
OS: Windows 10 Pro 64bit
Reply




Users browsing this thread: 1 Guest(s)