FlyTheMaddogx Announcement

[shamrockflyer 647]

Seeing this makes me a little concerned for MSFS 2020. I know its off topic, but how will this sort of issue be dealt with in MSFS as I'm sure the level of detail will increase over time with addon aircraft.

Are Asobo using DX12 in MSFS2020 or not? 

[Jazz 1,046]

20 minutes ago, shamrockflyer said:

Seeing this makes me a little concerned for MSFS 2020. I know its off topic, but how will this sort of issue be dealt with in MSFS as I'm sure the level of detail will increase over time with addon aircraft.

Are Asobo using DX12 in MSFS2020 or not? 

Not.

At first anyway. They will migrate later down the line.

[PurdueKev 159]

On 5/7/2020 at 1:25 PM, Skywolf said:

I wouldn't be surprised that even 11gb is not enough once one moves the sliders to the right.

You don't even have to move your sliders to the right to see the issues.  I have a 12GB card and I have been seeing these "memory overflow crashes," if you want to call them that, with very modest settings.

The pattern for me seems to be when I have a pop-up open (like for the F1 GTN). I am experimenting with the timeout detection settings in the registry which seem to have an impact, so it certainly points to exactly what Maddog is describing in terms of the way the program and the GPU are interacting in v5.  I think everyone's expectations - including mine, and maybe including L-M's - that existing add-ons could just be ported over with minimal effort was fantasy-land.  Either L-M needs to address it in their code, or all of the developers are going to have to put  lot more more effort into re-working their products than they have admitted to or realized.  Good on Maddog for acknowledging this.

The idea that "you need more VRAM" is unrealistic flight simulators, because no one is making mainstream GPU's with more brute force memory capacity (my 12GB is pretty much the max right now), just like no one ended up making ever faster clock speed CPU's once they go to a certain point.  The developers are going to have to get more efficient in their coding and architecture, because there is no hardware on the horizon that will magically solve this problem.

[micstatic 4,313]

1 hour ago, PurdueKev said:

You don't even have to move your sliders to the right to see the issues.  I have a 12GB card and I have been seeing these "memory overflow crashes," if you want to call them that, with very modest settings.

The pattern for me seems to be when I have a pop-up open (like for the F1 GTN). I am experimenting with the timeout detection settings in the registry which seem to have an impact, so it certainly points to exactly what Maddog is describing in terms of the way the program and the GPU are interacting in v5.  I think everyone's expectations - including mine, and maybe including L-M's - that existing add-ons could just be ported over with minimal effort was fantasy-land.  Either L-M needs to address it in their code, or all of the developers are going to have to put  lot more more effort into re-working their products than they have admitted to or realized.  Good on Maddog for acknowledging this.

The idea that "you need more VRAM" is unrealistic flight simulators, because no one is making mainstream GPU's with more brute force memory capacity (my 12GB is pretty much the max right now), just like no one ended up making ever faster clock speed CPU's once they go to a certain point.  The developers are going to have to get more efficient in their coding and architecture, because there is no hardware on the horizon that will magically solve this problem.

i totally agree with your points.  Though let's be honest.  It's very possible we see the 3080ti at 16g+

[jalbino59 114]

I haven't switched just yet waiting to see how this plays out, but I wonder if it helps at all not loading the entire scenery library like we used to do in 32bit to avoid address space issues.  Could DX12 be allocating space for all scenery to allow for faster loading, or do these OOMs occur with clean installs also?