Do you use any other performance affecting features in RetroArch for example Run Ahead and Frame Delay?
You might have to adjust those if you do. Since I started using my new Slot Mask presets I’ve found that I had to lower my Frame Delay settings quite a bit in some instances.
I just assumed that the Slot Mask might have been more performance intensive.
Since it’s a relatively recent discovery that possibly coincides with your new performance issues maybe it could be related and also indicate a slight performance regression somewhere.
These days I’m looking at Windows updates and Windows itself as possible sources for performance inconsistencies.
I use a Geforce GTX 1070 which is supposed to be comparable in performance to a GTX 980. I haven’t used the glass preset in a while but I can test it and see.
Have you considered using MSI AfterBurner to overclock your GPU?
You know many if not most of my current presets don’t require the “MBZ__0__SMOOTH-ADV_GDV.slangp” level presets to run like back in the olden days.
That is because most of them don’t use ScaleFX nor prescaling. Very few also require MDAPT.
I just continued to use the SMOOTH-ADV base preset for ease of development and flexibility as well as to allow users to toggle the features as they see fit.
So you can probably replace the current Glass Preset that’s under performing with a lighter version and most of the presets should continue to look the same while having improved performance.
This might be cheaper than purchasing a new graphics card.
Yes, what Cyber said about the base preset he uses is what is slowing things down. All of his presets use the SMOOTH-ADV as the base so if you swap this out for an ADV or STD glass preset it will run much faster.
The SMOOTH-ADV have a 3x upscale in the middle of the chain and have scalefx integrated
So the Zoom Crop basically does this kind of overscan cropping.
If we did have known overscan amounts for specific tvs and manufacturers it could be worth it to have a few predefined zoom crop percentages available.
I was just thinking of ways in which I can improve things and I remembered that @Nesguy uses a custom resolution in order to get more pixels at 1080p. It makes sense. I just wanted to try that at 4K using Mega Bezel to see if I can further improve quality.
This is something that we can start to research. For now I’ll probably start with @Nesguy’s overscan percentage. I think he actually has a formula he uses to calculate how much overscan in lines of resolution are appropriate. I’ll refer to it. I’m just putting the idea out here straight from my thoughts one time before I go about experimenting later.
If implemented correctly this might reduce the need to have as many different Game Presets just to eliminate black borders and junk from many games.
If anything I might actually prefer playing games which had black bars on certain TVs with the same black bars in HSM Mega Bezel Reflection Shader. For example Street Fighter II Turbo Hyper Fighting for SNES. This might also help with preserving accurate aspect ratios in games because cropping off from the sides or top/bottom like I currently do can distort the aspect ratio.
Having some real-world references in this regard built right into the shader might take things to an even higher level.
This is what I’m talking about and would like to be able to do in HSM Mega Bezel Reflection Shader as well as getting a few predefined zoom crop percentages:
In this example @Nesguy speaks about specific resolutions not percentages. I hope I will be able to implement things as precisely as this while using percentages.
I can probably look at the resolution info while adjusting the percentage as a guide (or use a calculator). I just hope it will look and work the same way even if there’s a slight rounding error due to the use of percentages instead of discrete numbers of lines.
People have asked time and time again what kind of hardware they need to setup a great emulation system, while others have asked what graphics card they would need to run HSM Mega Bezel Reflection Shader.
I’ve always dissuaded them from anything SBC or Android related and have instead tried to steer them towards a minimum Ryzen 4000 series APU or better based on some users’ reported success.
Well, I’ve always wondered if I could run these on my phone after GDV-Mini was integrated and I just did so successfully.
Device: Samsung Galaxy A71
Core: Nestopia UE
Driver: Vulkan
Audio Latency: 64ms
Frame Delay: 0
Auto Frame Delay: On
Run-Ahead to Reduce Latency: On
Number of Frames to Run-Ahead: 2 Frames
Use Second Instance for Run Ahead: On
Shader: MBZ__5__POTATO__GDV-MINI.slangp
Display Resolution: 2400 x 1080
The Super Mario test ran at 200~ on Mesen with runahead at 2, with it disabled, it went above 400, on FCEUMM it went beyond 2000. So I believe the test part could use more info so you’ll have more accurate reports like using a specific core and disabling/enabling latency settings, etc.
True, at least CPU and GPU would probably be good enough, and maybe even using stock RA settings, as there are some tweaks the user change that can affect performance.
Yes, with the MBZ__3__STD.slangp preset, no other setting changed from its stock config.
Keep in mind my PC is at the low tier range as of today, basically an i7-2600 and a GTX-1660ti.
I tried running it on my Android, and it either keeps freezing/crashing, or it runs so unbelievably slow that I almost forgot my CPU is a Snapdragon 865.
Maybe it’s because my device prefers GL over Vulkan usually.
So far, the “Potato” presets are the only ones to run at all, with the exception of a couple of “Glass” presets amongst the higher ones which run very slowly.
Well it is a phone after all. Thank God for small mercies.
This is great news that we can actually have the Bezel and these presets running on such relatively low performance and power devices albeit at the cost of reflections.
I think that’s a frontier I might like to tackle next with my preset pack.
