The 2 box solution

The 2 box solution.

I have been in technology since 1972. This is before anything like the 8008 chip was in civilian hands.

I have had flight simulators around the house since about 1980 or whenever SubLogic got started.

Landing the Dreamliner is hard because it seems like the stick isn’t sampled very often. This is a classic “long delay” control loop fault.

When I look at all the things MSFS2020 is trying to do and how painful and marginal some of the current solutions are, I ask myself a simple question… The same question that aerospace professionals might ask: What major functions can run in parallel in 2 separate boxes; 2 separate disk-motherboard-IO system units?

I think the solution is obvious…

Box 1. Maintains the flight model and samples control inputs frequently. The geographic coordinates and the pitch, yaw, roll output of the aircraft flight model are then passed to…

Box 2. This box renders the aircraft and world and handles all network traffic.

The current state of MSFS2020 tells us that for the game to run really well in a single box (one Intel mother board and one graphics system), several iterations of optimizing code are ahead of the developers. Large products like this are usually “demoed” in easy, script like environments and getting the thing running at all is usually when investors start pulling resources and money back out of the project. Then it stays in script form forever.

I do not expect that this team will be allowed to or will be able to move from demo script toward machine language level speed.

Therefore, if a version of this flight simulator is ever certified for the flight schools, I predict it will be running in twin boxes with the world rendering engine box receiving aircraft current location info across a simple serial line from the flight model box.

They do not have to be equal boxes. The flight model function may not need more than a standard modern Intel I7 chip set.

A really clean flight model that is highly responsive to student and instructor inputs is what the flight schools want.

Instruction can still proceed even if there are funky world rendering issues and delays in drawing swans on lakes.

2 Likes

I would be very willing to use something along this line too.

If it would allow us to drive triples in 4k/1440 @ 50fps, with an option of having a monitor in front of us with the dashboard (as a minimum) I would be happy to fork over the cash. For the people that drive the big planes, they would need a ceiling monitor too.

Sign me up. I paid off my Master Card this month and am itching to step it up! Yes, I have GAS…

A lot of people running $3500 single boxes are still dissatisfied with performance. Especially those wanting to run VR.

I think MSFS2020 is not a computer challenge, it is a “network” challenge. The major cloud service providers do not try to do what they do with a single super processor.

Computing acceleration from force over mass can run in parallel with mission execution (FMC), with modeling complex onboard systems, with staging up scenery and finally with presenting scenery. There we have 5 interconnected boxes, each handling its responsibilities easily. This is how to get hundreds of cycles per second.

I would not be surprised if the major government and aerospace entities are, as we speak, asking the developers to prepare to “modularize” MSFS2020 so that great immersive performance doesn’t need hardware or software miracles.

Look back at the “semi-parallel” data processing techniques that came out of the Manhattan project. They are written up in lots of histories.

Modularization probably already exists in the software. People code their activities and pickup/leave results for other activities. Wires between boxes work just as well as shared memory locations.

1 Like

The workload of the sim (by nature) is currently not able to even utilize all the threads on top end (productivity) consumer CPU’s. If the scenario you posted in your opening post were possible, they would also be able to do the same thing on a single CPU.

So while it’s nice to theorize about 2 box solutions, I feel there would be no benefit whatsoever, probably more like a deficit, since there will be overhead in letting the two communicate to eachother.