Lockheed Martin Prepar3d

[Neil Hewitt]

Have people been following this at all? Prepar3d is Lockheed Martin's fork of the Microsoft ESP source code. So it's basically FSX for commercial simulation. In the last few weeks their site has begun to come alive and it seems they're close to shipping an inital version. No word on pricing, I imagine given the audience that they're pitching for that it will be business-level pricing rather than home-user-level pricing, but even so we're probably talking hundreds rather than thousands for basic licensing.

Word has it that they've been working on the core engine to optimise performance, move work onto the GPU etc. They apparently have some ex-ACES people on staff working on the code.

This must be worth keeping an eye on. Not only will Prepar3d be backwards-compatible with FSX add-ins, but given its target audience you can expect it to become more open and programmable and suitable for a cockpit-building audience, unlike Microsoft Flight which looks like it's going to target the casual gamer.

Could Lockheed Martin, rather than Microsoft, end up being the salvation of the home cockpit hobby?

One to keep a close eye on, I think.

http://www.prepar3d.com/

[JNicol]

Hi Neil,

I thought I should pipe up here as I am the guy responsible for Prepar3D commercialization. I have been a member of this site for years under another user name and I guess I wanted to keep that one under wraps... I have built sims at home for a few years (helicopter, desktop Cessna and a truck sim for my son), built them for a living (Bell 412, A320, tank sims etc), have a copy of Mike's sim instrument book and so on. Am I passionate about home cockpit building.....? yep.

We are going to be announcing the price point for the client and the Developer Network within the next two weeks and I wish I could be more specific on that. The surprise will undoubtedly come for developers participating in the development program.

I can tell you that I think it was the first day I joined the project, I said that the number 1, 2 and 3 items that will need to be addressed will be performance. (Actually I probably said that during my job interview as well!) We have been working on the core engine and we are looking at updating it from DirectX9 and getting cycles onto the GPU at some stage. It is a hard road, so I wouldn't count on major performance increases just yet, but there will be tweaks. A million lines of code is not something that we can get through easily! The development team are world-class though, they are working on it and we have an exciting roadmap for the product!! Yes, we also have a couple of guys from the old ACES Studio as well and I will leave it to them to introduce themselves through the website blogs when they come up for air.

Some of the new features include sensors, multi-user enhancements (i.e. more than two users sharing a cockpit), brand new user interface, upgrade to Visual Studio 10, OpenGL to DirectX support and others that I am hoping to share in the next week. (*hint* anyone like submarines?) I also asked Pete Dowson if he would support Prepar3D with FSUIPC and he has integrated this support in the very latest version (which will also be available as a download on the Prepar3D website as well).

All in all as long as you are not using Prepar3D for entertainment purposes, cockpit builders should be happy. At least you know that there is a MyCockpit mole on the inside....

I would encourage interested people to get onto the site and tell us what they want and if there are any annoying things that would make life easier that we could fix or put into a future build. We really intend opening up the doors and getting the community involved with us to make it even better. It is our intent to make regular updates and patches (er...feature enhancements) to the product and not do a release every 3 or so years, so there will be more opportunity to get features in and updates out.

Regards,
John

Note that the views and opinions expressed in this post are my own and do not necessarily represent the views and opinions of Lockheed Martin.

[Neil Hewitt]

Thank you, John, for the great news. Nice to hear it from the horses mouth, at it were. I would argue that most non-pro builders are going to derive some degree of entertainment from their flying , but it's clear that from a company like Lockheed it's not going to be targeted at your casual simmer - the guys putting together commercial sims for real-world training have presumably got to be your real customers. For a significant and ongoing improvement over FSX I would certainly be willing to invest a reasonable amount over time, and I'm sure many here would too. The backwards-compatibility is clearly going to be a major draw compared to, say, moving to X-Plane or the Aerosoft product (assuming it happens). Having FSUIPC on board is a big win.

I'm a software developer by profession, so I appreciate that fixing the performance issue is far from straightforward. I dread to think what a genuine multi-core rearchitecting of a monolithically single-threaded codebase that old would look like as a task. You'll have to do at least some concurrent tasking to take advantage of GPU offloading, of course. Rather you than me <g>. Actually, scratch that. I would jump at the chance if I thought I was even vaguely good enough. I'll just have to stick to .NET

I look forward to finding out lots more as the weeks go by.

[JNicol]

Well, I also hope that we can attract the development community to the platform that have had previous experience with FS9 and FSX. There will be lots of opportunity to reach our target commercial market with the add-ons and utilities that make the base platform the powerful engine it is. There are lots of examples that I am seeing now, including the use of payware aircraft and scenery add-ons by commercial users that want to enhance the experience. I definitely believe that there are opportunities for people to contribute and be part of this at many different levels.

I hope, for example that we can start a Search and Rescue group and get the community to contribute payware/freeware aircraft, missions, challenges, flight plans and so on. An Emergency Preparedness and Response group, a Humanitarian Assistance group, the list goes on. There are flight schools and colleges that want Prepar3D, so I see lots of opportunity for enhanced payware aircraft with realistic start-up and shut down procedures, local scenery add-ons for those guys and again tutorials, missions, tools etc. Lots of opportunity I think for hardware providers as well as we move the goal post from entertainment into commercial areas that may be interesting for folks. People in this forum have great ideas and talents putting hardware and software together so I have no doubt that there will be some great ideas coming up!

Just some random thoughts.

[JNicol]

Hi Neil,

I apologize in advance to the forum admins and members, this isn't supposed to be an ad. Just letting you and everyone know that I have just posted the pricing and availability for Prepar3D on the website. www.prepar3d.com.

The price for the client application is US$499. The price for joining the developer network starts at $9.95 per month. Note that anyone developing add-ons, missions, flight plans and so on for Prepar3D can join the developer network. When a solution is delivered for an end-user, they will pay the $499 for the commercial runtime license. For $9.95 per month, developers will have access to two (2) copies of the Prepar3D client application, plus access to other Prepar3D apps. If anyone has any questions, please feel free to go to the web site. Again apologies if this is not an appropriate medium to get the message across. (This is a world-wide exclusive for myCockpit!)

Regards,
John

[Matt Olieman]

John, welcome to MyCockpit and thanks for the update of Prepar3d.

I (we) look forward to hearing more of Prepar3d progress and any other updates you may have to offer us.

Again, thanks for the "world-wide exclusive."
Matt Olieman

[Neil Hewitt]

Thanks, John - that's very interesting. And not as pricey as I had expected

Some builders end up using more than 2 copies of the sim software at a time, though - while today's powerful PCs do a lot, in a five-projector system, say, it's hard to avoid having more than 2 PCs running the simulator; you just can't get the performance out of a single client and there certainly isn't a Quintuple-Head 2 Go. What do developers who need more than the 2 copies of the Prepar3d client do? Do they need to acquire commercial licenses for any development clients over the 2 limit? Or will there be bolt-on pricing for the developer network?

I will certainly be looking at joining up and getting a good look at the first release of the client, particularly to see where it has evolved over FSX/ESP.

[JNicol]

Hi Neil,

You raise a couple of good points and hopefully I can address these. The new ATI Eyefinity cards can have up to six monitors out. There was a limitation with ESP that had a maximum of 4 monitors out before it would crash. This has been fixed in P3D, so that using something like an Eyefinity, you can have 5 or 6 monitors/projectors out. We have also put in a new feature to address performance issues where you can run the main simulator PC with only an instrument panel and no out-the-window visuals.

Simulation systems such as Project Magenta in a typical client/server configuration will allow instrument panels to be run on separate machines with a copy of FSX being used for each out-the-window view. For typical development environments I can see that two copies of the software will be sufficient for most applications, however I acknowledge there will be some commercial applications that will require additional copies for development. I would see that these more complex environments may have more than one developer working on them?

Saying that, I will look at this for sure and look at pricing of additional development copies over and above the two. Remember that we are talking about $10 per month for the first two to allow smarter people than me to develop apps for P3D.

When I built my A320 sim I had a single channel with multiple projectors out on the single video card, so only needed one copy of FSX and had other software for the instruments. My Bell 412 was the same. If I was going to upgrade the basic systems, then of course I would look at a PC per channel for greater performance.

Perhaps people reading this could give me an idea of their typical configurations for multiple copies. That would certainly help me to ensure that the developer community is going to be served.

John

[Neil Hewitt]

Good point about EyeFinity, although I don't think too many people here have gone down that route yet, based on posts I've seen. The issue with FSX has always been that it's CPU-bound. Even today's fastest CPUs, overclocked, cannot satisfy the desire for high framerates when flying in a very complex scenery area, with heavy weather, lots of AI traffic, add-ons like PMDG, and sliders maxed (or even very high). The graphics hardware is almost never an issue these days because GPUs are so much faster than they were when FSX was released. But everything that is rendered on-screen has to be calculated by the CPU on what amounts to a single thread. Microsoft was able to push texture loading onto separate threads, but AFAIK that was about it. Unless your team has been able to substantially re-engineer the core of Prepar3d to be inherently multi-threaded in a comparatively short time - which, with all due respect to them, I doubt; we'd be talking about absolutely fundamental changes to a million-line codebase - then performance is limited by what can be done on a single thread. No doubt there are areas within that thread where performance gains can be made, and it looks like you've achieved them based on what you say, but I would be dubious indeed about trying to run a three- or five-display system using EyeFinity on a single PC, which would almost certainly need multiple visual windows active, and getting decent performance landing at JFK, say.

All the client/server stuff you can buy that uses FSUIPC is brilliant - and takes the load off the FS PC - but some of the really CPU-consuming stuff like autogen and AI traffic cannot be networked out in any way (at least AFAIK). Being able to run a copy of the client with no visuals and no CPU hit so that all the power could be used for other stuff would let you dedicate one PC to visuals with no control hardware, sound hardware etc, and then run panels, hardware adapters, controls, sound etc on a non-visuals PC; I could see that working even with a multi-projector system, but I'd still be expecting to see the frame rate drop through the floor with all the above stuff I talked about turned on.

I guess what would be great, over time, would be for you to be able to split the functions of the client up and run them on multiple PCs networked, and indeed in multiple processes on a single fast PC. This would closer approach the kind of architecture you see on Level D systems. In particular, having a visuals-only client that didn't run in slew mode but in a dedicated outside-view mode, with other modules feeding it all the data it needs including the position of AI planes etc, would let you get great visual rendering performance out of todays fast PCs.

You could take the sim engine as it stands and make it possible to turn off chunks of the overall pipeline. So for example, one engine process could be only generating the sound, one could be calculating the positions and activity of AI aircraft and vehicles, one could be handling add-ins. Each engine would be able to get its own threads and RAM from Windows, and the load would be distributed across all the cores available. The necessity to coordinate activity between multiple processes adds an overhead, of course, but FSUIPC and WideFS already do this, and whether it's across the network or across processes in a single box doesn't really matter. I suspect Peter Dowson might make an excellent distributed member of your dev team

As a software architect (albeit one with no particular speciality in gaming or simulation) it seems to me that going for a grid-computing approach is probably going to be easier than comprehensively re-writing the software to be inherently multi-threaded. But that would definitely involve running lots of copies of the core engine, so the licensing model would have to change.

Anyway - far be it from me to tell you how to do your job I'm just really keen to see what you've been able to do so far and to take Prepar3d for a spin.

For the record, my own sim - still very much in the building phase, but planned out in terms of software and hardware for the initial build - consists of 2 main desktop-type PCs (but in a special housing I built for them to minimise the space they occupy), 1 tablet PC that will be mounted directly into the MIP, and an old ultra-mobile PC (an ASUS R2H) that will just run custom software I'm building for a radio stack and some other stuff. My project will be heavily touch-screen based with lots of UI written by me. So I'm probably a bit different than most around here. The precise division of functions is not finalised yet but I will almost certainly need to run 3 copies of FSX (I have 4 licenses). I may add another desktop PC to the mix. So to run Prepar3d in all the same places I'd arguably need 3 or 4 clients. However, I guess if I were to adopt a Prepar3d solution I would look to rearchitect around a 2-client system and take advantage of your performance increases.

Hope that helps! I think you can probably get a lot of feedback from people here if you start a poll - just ask Matt or one of the admins. It's very on-topic so I guess there'd be no objections that it's a disguised advert. Perhaps Lockheed would fancy advertising here anyway

As always, John, thanks for your participation and all the information you're providing. I'll be signing up in November!

[JNicol]

Hi Neil,

You have again raised some good points for discussion. I have to say that this is P3D, so even though we are starting from the same code, we are going to be getting the graphics processing off the CPU and where it belongs into the GPU. It won't happen for a while as we upgrade the core, but it is on the roadmap. Performance might be slightly increased with what we have done now, but dont expect 100 fps just yet! In any case, I think a poll sounds like a great idea. I will ask Matt. So the question would be, For a typical Prepar3D development set-up, how many client licenses would be required - 1 - 2 - 3 - 4 - 5 or more. I will also look at getting a poll up on the P3D website.

Regards,
John