(03-21-2015, 11:09 PM)Tino Wrote: [ -> ]Master dos not support predictive fifo.
I was referring to question 1
In master branch, when dual core is enabled, does it make sense to check "Synchronize GPU thread" AND set Deterministic dual core to "fake-completion" in order to optimize stability and speed?
updated the latest folder with a new version. @kibypuff can you test if this fixes the memory problem you reported?
(03-21-2015, 11:50 PM)Archdevil456 Wrote: [ -> ] (03-21-2015, 11:09 PM)Tino Wrote: [ -> ]Master dos not support predictive fifo.
I was referring to question 1
In master branch, when dual core is enabled, does it make sense to check "Synchronize GPU thread" AND set Deterministic dual core to "fake-completion" in order to optimize stability and speed?
you will have to test it for yourself the performace stability will depend on your hard and you alone will know what is the best configuration for you.
(03-22-2015, 08:05 AM)Tino Wrote: [ -> ]updated the latest folder with a new version. can you test if this fixes the memory problem?
So, in this build you've adjusted the cache flushing interval / frequency to be twice as fast as before.
This "fixes" the memory leak, but still doesn't help with the excessive VRAM usage (1500 MB of VRAM used at 6xIR vs. 200MB in master)
That's just too much for a GPU with 1GB of VRAM.
There are many GPUs that are perfectly capable of running Dolphin at 6xIR, but have only 512MB of VRAM. With Ishiiruka's new cache, those GPUs are now rendered "obsolete".
I'd like to see the return of the standard caching method and this new cache provided as an option (disabled by default).
The normal caching method that's used in the master buids is good enough and suitable for all GPUs (even those with limited VRAM).
That is really weird in my machinne at 8x afyer 30 minutes the maximun value i get is 800MB and the mean values is 600Mb
(03-22-2015, 11:44 AM)Tino Wrote: [ -> ]That is really weird. At 8x, after 30 minutes the maximun value i get is 800MB and the mean values is 600Mb
There's a ~1500MB spike @ 6xIR in dkcr l8-6
My problems with textures were solved. And it seems another problem was solved too: sometimes (after 30 or 40 minutes of playing) Xenoblade Chronicles started to lose FPS - from the stable 30 FPS to 15-20, with slowdown of music and stutters of sound. But now I'm played over 1 hour and there's nothing with these problems. I will check further...
Thanks for the fast solution of problems!
There seems to be a bug in the latest Ishiiruka build (probably recently introduced, I'm not sure) not present in master:
In the screenshot attached (Tiger Woods PGA Tour 12, Wii) the shadows are noticeably misaligned (DirectX 11).
(03-23-2015, 05:01 AM)Archdevil456 Wrote: [ -> ]There seems to be a bug in the latest Ishiiruka build (probably recently introduced, I'm not sure) not present in master:
In the screenshot attached (Tiger Woods PGA Tour 12, Wii) the shadows are noticeably misaligned (DirectX 11).
can you please compare if the result is the same in ogl and dx9?
(03-23-2015, 06:45 AM)Tino Wrote: [ -> ] (03-23-2015, 05:01 AM)Archdevil456 Wrote: [ -> ]There seems to be a bug in the latest Ishiiruka build (probably recently introduced, I'm not sure) not present in master:
In the screenshot attached (Tiger Woods PGA Tour 12, Wii) the shadows are noticeably misaligned (DirectX 11).
can you please compare if the result is the same in ogl and dx9?
Same problem with DirectX 9 and OpenGL backend. (noteworthy: Creating a savestate and immediately loading it solves the problem for the current hole but as soon as the the next hole starts the problem reappears. Moreover not all holes are affected.)