FSDS V3.5 and FSX - Xtomdl vs makemdl - must read

Chris · Post by **Chris** » 15 Sep 2007, 09:11

Dear All,

I'm writing this to provide some important to my opinion observations after some quick testing which I believe will rock everyone thinking to start moving to FSX..after a period off course

Last night I've downloaded the upgrade for v3.00 users and I decided to give it a shot so as to see what is happening.

Well...

I have exported a version of my model under developement (an F-16) using the makemdl (fs9 sdk wise) and the xtomdl (fsx sdk wise) so as to see the differences in quality as also the frame rate impact in FSX. So I'm presenting below the 2 screenshots I took for comparison. The model is untextured so in FSX is purely black..

1st observation

After Using Makemdl (fs9) the fsds V3.0 shows 3915 poly's and the ACM shows 5917 polygones - Using xtomdl (fsx) the FSDS V3.5 shows 4617 poly's... I don't know what kind of optimization is hidden behind but the result is positive.

2nd Observation

The quality of the exported model is superb. You are able now to prepare a project using 0.001 scale and the welding taking place within xtomdl after does not destroy the model small parts (i.e looking at pitot tubes, antennas etc). In fact it was the first time for me that the details of the model within fsx were absolutely the same as within FSDS 3d view! No strange shadows, jerking poly's etc

and the observaton No 3 (it's all about fps) right?

well here is the first screen shot taken after exporting using xtomdl with 30 aircraft parked in view. Look at the fps locked at 25 I got 25.

and here is the same model exported with makemdl for fs9 - fps is 6...

If that is not enough let'me add one more more important observation...What you see in these screnshots is a SINGLE LOD model

For those interested, My specs are vista ultimate on an quad core 6600 (2.4), 2 giga ram, ATI XT2900 512 mb , fsx with SP1 ofcourse running at 1600X1200 in a T965 eizo and with all sliders (on both ATI and within FSX moved to full right except autogen and any other traffic apart from planes which are set to zero.

In fact this single LOD model exported via xtomdl caused no impact at all which to me is impressive. So apart from any other problems prevent people to start moving to FSX which is not my subject, I would suggest that if you are going to produce anything for suitable for FSX make sure you use the xtomdl ...

Cheers

Chris

MIKE JG · Post by **MIKE JG** » 15 Sep 2007, 12:45

Interesting results there Chris, very encouraging.

RipPipPip · Post by **RipPipPip** » 16 Sep 2007, 15:36

I feel I'm convinced

BTW: the back-test would be very interesting as well - are there any profits when putting FSDS3.5 generated AI into FS2004, comparing to those plain, old FSDS3.0 ?

Chris · Post by **Chris** » 16 Sep 2007, 18:21

Rysiek,

I'm Just guessing that apart from any specialised keyframe animation which might exist in FSDS V3.5, or the enhanced boolean operation as Abacus claims (I'have'nt test it) I'm afraid that there will be no benefit as you will still have to go via makemdl so as to produce the mdl for fs9 and that is the limiter.

On the other hand as you already know FS9 does not accept mdls exported via xtomdl

so I suppose that we also cannot earn something following this route either.

aerogator · Post by **aerogator** » 17 Sep 2007, 15:44

I'm a little confused Chris. Is it possible to specify which compiling tool you use: either for FS9 or FSX or do I have to configure the FSDS for one or the other. I have downloaded the update but I'm not going to install it until I know for sure. I've read the "new features" list on the web page and it seems to be OK for FS9, but I'd like some feedback from you guys as well.

Thanks

Chris · Post by **Chris** » 17 Sep 2007, 16:31

Jake,

It's very easy

. I will give you the instructions but I' give you also a trick that I'm using which I believe is more comfortable and practical in order not to switching between FS9 and FSX.

So, After installation, when you first run the program, it will ask you to configure the program according to where do you want to export your models and sceneries FS9 or FSX. The program is doing that by asking you to browse the relevant SDK folders depending on your decision.

In other words if you choose to export to FS9 it will ask you to browse the folder of FS9 and the folder on which makemdl (from FS9 SDK) exist - also for the sceneries the Bglcomp and that's it.

If you choose that you want to export to FSX the proccess is similar but with the exception that this time you can choose between makemdl or xtomdl. This is happening because as you already know you can produce an aircraft using makemdl for exporting to FS9 and then to run it within FSX (with all known limitations - not recommended).

Note that after initial setup, whenever you want, you just go to the preferences and change through the compile options button the target of export by switching to any direction you want. That means ofcourse that you will have again to browse the paths described above.

Personally I don't like this so I'm following a different approach.

So here is what I have done in order not to switch preferences all the time.

1. I haven't touched my FSDS V3.00 installation (I kept it as is in the folder of my original installation - for example FSDS V3.00.
2. I have installed the FSDS V3.5 into a new folder for example FSDS V3.50.
3. During initial FSDS V3.50 run I have choosed by default to export to FSX by using FSX SDK (that means use of xtomdl).

So, depending on which project I want to work and where do I want to export (FS9 or FSX) I'm opening the relevant version of FSDS by it's desktop shortcut. I'haven't seen any improvement of using the FSDS V3.5 for exporting to FS9 so there wasn't any reason for not using the FSDS V3.00 when my project was strictly FS9 oriented. On the other hand if there is something that I want to export it to FSX I'm just opening the FSDS V.3 .fsc file from FSDS V3.5 and exporting straight using the correct compiler which in this case is the xtomdl.

Also note that like before with fsds v2.XX and v3.00 there is no interference between them.

I hope that helps

Chris

BTW Sorry for my bad english, I'm trying to improve them but is going to take some 100 years more

ronniegj · Post by **ronniegj** » 17 Sep 2007, 17:29

Chris - doesn't the improvements in V3.5 -- esp the snap to at .001, and the improved boolean operation -- make it desirable to do your modeling in v3.5 only, instead of v3.0?

Ron

aerogator · Post by **aerogator** » 17 Sep 2007, 17:37

Chris - doesn't the improvements in V3.5 -- esp the snap to at .001, and the improved boolean operation -- make it desirable to do your modeling in v3.5 only, instead of v3.0?

That would be my question as well.

And BTW, your English is much better than any other language I try to speak

!!

Chris · Post by **Chris** » 17 Sep 2007, 18:50

Ron and Jake,

I agree that ofcourse it's desirable but at the very end this is a right, when the model exported, is, the model you have designed. What I mean is that snap to at 0.001 is a benefit when you are able to see this detail at the end result, and that result can only be seen in FSX and only if you export the model via xtomdl.

The snap 0.001 existed to FSDS V3.00 already but makemdl compiler was not powerful enough to produce such detail in FS9. So what is the point of entering a design proccess with such detail when your aim is not FSX?

So in other words, from what I'm getting so far from testing, it's not a matter of FSDS really. It's a matter of compiler. That is what I'm saying. You may spent many manhours designing a super detailed project and at the end to discover that unfortunately you have loosed many of the details in FS9 because you are limited to the makemdl use for exporting the model. So what is point behind?

I really don't know.

As for the boolean operation I' have'nt tested enough but to be honest I haven't seen any difference apart from a feeling that seems more stable as operation. From this point of view you are probably right.

But again is a matter of design methods and project. I would say that I'm trying to avoid any boolean operation as much as I can and I prefer the method of delete and re-poly the surfaces where needed manually.

Of course the above are describing my personall approach to this issue, and there is a strong possibility that I'm far away from what is right. However what I'm saying, I'm trying to say it based on personal experience and examples where possible.

So...take a look at this example. It is the area of the teardrop shaped fairing houses the RHAWS on the RH side of the FWD fuselage and the radome including the pitot tube of my F-16 project. Snap is set to at 0.001.

Maximize the pictures to see the detail.

here is how is looks within FSDS V3 and the FSDS V3.5 (it's absolutely the same)

Here is how the model exported via makemdl from FSDS V3.5 to FS9 and is absolutely the same if you do it from FSDS V3.

and here is how the model looks when exported again from FSDS V3.5 via the xtomdl to FSX.

I think that is very clear the difference. In the FS9 (makemdl) its clear that several poly's disapeared from the static ports as also look at the intake of the pitot tube. It has been destroyed compared to the fsx version which has a perfect cycle shape.

I hope that example to be helpful explaining what I mean.

ronniegj · Post by **ronniegj** » 17 Sep 2007, 19:06

Yes, clearly there is a difference, and there is plenty of food for thought, and need for testing to develop some guidelines to follow when making choices. I appreciate your efforts in sharing this information.

Ron

Chris · Post by **Chris** » 17 Sep 2007, 19:26

Like you mentioned, food for thought...

So at this point let me add a strange (to me) observation before I go for sleep... (it's 2230 hours here in Athens, Greece)

When you are exporting your model from FSDS V3.5 to FSX you are getting a message showing the number of vertices of the model created. I repeat - the number of vertices(points) - not the polygones.

This was looking strange to me but then I remembered some articles I have read lately, I think were coming from ACES team) in the net about the importance of the vertices in games lately designed and in the future.

Well the output as I took it was that we are in front of a new challenge as it looks that from now on the number of vertices will be much more important compared to the number of polygones in any design. That means that if we follow that route, in a way, we should start considering the number of vertices instead of polys if we want to gain performance....

How these articles might be related to this message created from FSDS V3.5 I don't know but I feel that something is going on and probably will change many things...

aerogator · Post by **aerogator** » 17 Sep 2007, 20:30

Thanks, Chris, for your input.