Transform and Lighting (T&L) revisited

[Guest JonP01]

I am just wondering if anyone else has had the same experience that I have regarding the T&L setting in both CFS2 and FS2002. I have found on two different computers, both using Geforce 4 cards, that enabling T&L actually reduces the frame rate rather than increases the frame rate. I find this quite interesting and I can't help wondering whether those T&L settings in both these sims actually directly control the onboard GPU T&L engine, or they are something else to do with the proprietary graphics rendering engine itself.I set up a few different scenarios of an aircraft sitting on the runway at some "detailed" airport and noted my observations. All the graphics sliders were maxed out, frame rate was uncapped and vsync disabled. That way I was getting a true reading of the frame rate produced. I also had 5/8ths cloud, moderate precipitation and light wind. All I did was toggle the T&L on and off about 6 times and noted the frame rate variation between the two settings. Both machines used the exact same components except the CPU and mainboard. I used Win98 lite with 256Mb RAM and a Geforce 4 Ti 4400 and used the Cessna Caravan in VC view. My point isn't at all to comment on the actual quantum of the figures themselves, as both these machines are quite slow and used slow SDRAM as well as chipsets over 12 months old. The interesting thing is that in both cases the performance was better with T&L off. I didn't test what happens in actual flight, as I couldn't work out a fair way of doing this. I guess replays are possible, but a lot of the number crunching performed by the sim in real time doesn't need to happen with replays, so I don't know how valid those figures would be. My observations were consistent regardless of resolution, colour depth or any other GPU-specific settings, or even GPU drivers.Here are the figures from my YSSY example: Machine # 1: Pentium III 933:T&L on: 16.9 fpsT&L off: 19.2 fpsMAchine # 2: Althon XP 1800+T&L on: 24.6T&L off: 29.3The only two theories I can come up with are that either:(i) T&L in FS2002 isn't really true GPU T&L at all, but it is something elseor(ii) the T&L calculations are performed better by both CPUs in this particular situation than on the Geforce 4.I realise this is not at all scientific and is probably of little or only academic interest to most people, as many people (including myself) don't really bother about what the fps is so long as the sim is smooth. Nevertheless, I find my observation interesting from a technical point of view given the emphasis many people put on how T&L improves performance. I should also add that these two sims are the only games I have that do not seem to benefit from T&L. I would have thought that FS2002, being so processor bound, would give much better performance if the T&L burden was passed onto the GPU rather than CPU. That is why I can't help wondering whether the T&L options in FS2002 and CFS2 are something different to GPU T&L and relate to something else.I am interested to see what other people's experiences are.

[Guest Stamatis]

I am a complete layman when it comes to such technical/hardware issues, so I may be talking nonsense here, but should we be surprised that the frame rates went up when the feature was disabled?I mean, take anti-aliasing as an example. If you disbale it, frame rates will go up, quite drastically so in some situations.What I am trying to say is that when you disable a feature that takes up some processing time, isn't it natural that frame rates will increase?Finally a question: Did you observe any quality degardation in the display when you disabled T&L ?Stamatis

[Guest JonP01]

>I am a complete layman when it comes to such >technical/hardware issues, so I may be talking nonsense >here, but should we be surprised that the frame rates went >up when the feature was disabled? >Well, theoretically, enabling T&L tells your grahics card (GPU) to perform this step of the rendering pipeline rather than the CPU. So at least in theory that frees up the CPU for other tasks. So I had originally felt those extra CPU cycles would be of benefit to FS2002, as FS2002 is supposed to be extremely processor bound. This is on the premise, however, that a Geforce 4 is more efficent at T&L than either of these CPUs. AA is a little different though. When you enable this, you place extra load and bandwidth demands on your GPU, but you don't provide any "relief" for want of a better term to any other components. It is purely a hardware function within the Geforce 4 GPU engine. So I'd fully expect a drop in performance with AA under all circumstances. Mind you, I'm being theoretical here. As your post intimates, being theorectical seems to be dangerous :-lol>Finally a question: Did you observe any quality degardation >in the display when you disabled T&L ? None at all. It looked exactly the same regardless of whether I had it on or off. I also tried different "number of lights" settings as well, but didn't notice any differences in this case either.

[Guest Stamatis]

"Well, theoretically, enabling T&L tells your grahics card (GPU) to perform this step of the rendering pipeline rather than the CPU."Oh OK, I thought that by unmarking the box it meant that there was not going to be any T&L processing, no matter the processor.Stamatis