Sorry for the rant as we all know rendertimes for TG2 are slower than sludge.
Unfortunately, this is seriously holding back my image creations with TG2. Of course it is far superior to the other apps I use. But I have abandoned many image projects due to them taking so long to render! Its pretty sad when 800x600 images are taking days to render at decent quality settings (settings to where you can't really see the grain). And this is on an AMD AM2 4600, with 2GB RAM.
There has to be something in the works to resolve this. Right? This program is amazing, but its limited as of now. Or atleast until 20GHz PC's come out.
Now, "IF" we are stuck with this, then one option would really help: image tiling. I'm not talking about the crop feature either. With image tiling we could render small sections of big and detailed images, and once done TG2 could just keep track of which tile goes where and assemble it into an image. It would still take weeks or months to complete and image. But the PC won't be clogged up and you wouldn't have to leave it on constantly for renders.
As we have previously discussed in similar threads there is significant optimization yet to be done. Our focus thus far has primarily been on feature implementation and stability. Optimization work will come later in a logical progression of development. This is simply the nature of working with a pre-release product. You can certainly expect the final version to be much faster, especially on multi-core/multi-CPU systems as the current rendering engine is single-thread.
Only time will tell just how much optimization is possible but I think that the combination of quality and render time TG2 offers will be very good when it is finalized later this year, particularly for those with computers purchased within the last 1-2 years. Older systems will of course suffer from poor performance, but this is to be expected with any modern graphics application on old hardware.
- Oshyan
looks like some one is having a bad day!!
Quote from: dhavalmistry on April 09, 2007, 07:47:16 PM
looks like some one is having a bad day!!
Heheh! Nahh not really even though I am stuck at work... :o
Just get bummed when I see a great image at 400x300 resolution (that I want to make in a poster-quality 3200x2400 image) take nearly a week at 1024x768 and still not be finished.
And thanks Oshyan for the info. This is just another 'hot-poker-in-the-side" for the development team to let them know we are their biggest fans...atleast they released a pre-release to play with.
I'm curious what hardware you're working with. A 1024x768 image really shouldn't be taking a week to render. It's possible you're just doing extremely complex scenes, or working on something that taxes TG2 more than usual (reflections for example), but my guess would be you just need better tuning of your detail settings. Although things are still not as fast as you would probably like, I think we can probably help reduce your render times without sacrificing the quality of the end result.
- Oshyan
that's the one thing that will stop be buying tg2 right now is the extreme render time.. maybe if there is significant optimizing later I may think again. also its not a final release.
Quote from: Oshyan on April 09, 2007, 08:38:28 PM
I'm curious what hardware you're working with. A 1024x768 image really shouldn't be taking a week to render. It's possible you're just doing extremely complex scenes, or working on something that taxes TG2 more than usual (reflections for example), but my guess would be you just need better tuning of your detail settings. Although things are still not as fast as you would probably like, I think we can probably help reduce your render times without sacrificing the quality of the end result.
- Oshyan
oh yes....I have send them many files and they have fine tuned for me....the support is great....also if you dont want to bother them and let them work on the next update (so that they can release it sooner ;) ), you can consult about your problems here. We are not as good as the support team but we ARE here for u.....I am sure you know that by now ;D
Really? You must not have the trouble I have with $$. The price will likely sky rocket by that time. I would get with it now while the getting is good. Then again, I'm not you.
Quote from: king_tiger_666 on April 09, 2007, 08:41:31 PM
that's the one thing that will stop be buying tg2 right now is the extreme render time.. maybe if there is significant optimizing later I may think again. also its not a final release.
Quote from: calico on April 09, 2007, 11:11:19 PM
Really? You must not have the trouble I have with $$. The price will likely sky rocket by that time. I would get with it now while the getting is good. Then again, I'm not you.
I don't know about that. the standard deep is 275$nz deep+ani is 414$nz
damn that american dollar >:(
Will the optimisation include memory management and a render catch which are some of the ways you could go about it I understand that version 2 of the Brazil Render System will be getting them, and they are present in other renders as well, just throwing ideas out there.
Regards to you.
Cyber-Angel
As with king tiger, I would REALLY suggest buying TG now.
Man thought it was only $200..lol
As Oshyan has said in the past, the final release price will skyrocket, dunno, maybe $1000+?
So I'd really suggest buying now :P
You could...buy...15..and resell them..and make a nice profit :P Think of it as a nice long term investment, lol :P
Quote from: Oshyan on April 09, 2007, 08:38:28 PM
I'm curious what hardware you're working with. A 1024x768 image really shouldn't be taking a week to render. It's possible you're just doing extremely complex scenes, or working on something that taxes TG2 more than usual (reflections for example), but my guess would be you just need better tuning of your detail settings. Although things are still not as fast as you would probably like, I think we can probably help reduce your render times without sacrificing the quality of the end result.
- Oshyan
I am working with an AM2 4600 with 2GB of RAM. I don't have any specific TGD files to post, but most of my scenes are highly complex (just because they can be). I do use alot of reflections on rocks and such and tons of detail in the rocks. Of course the skies are super detailed as well, with high quality clouds, etc.
I've seen a few images that were completed at larger resolutions, and the clouds were the most amazing 'rendered' clouds I've ever seen. I want clouds like that too. ;D
Moodflow, is it possible the number of samples on your clouds/atmosphere are above 256? I usually use 16 to 32 on the atmosphere and 128 to 300 in the clouds. Of course, the clouds seem subjective. If we want all the grain in the clouds gone, then all of the quality settings are adjusted upward to some degree. That degree is where some of us have gotten into similar troubles. And, in case you don't know (you probably do), the clouds' setting for Ray Tracing doesn't usually need to be on and will definitely add time to the render.
Quote from: calico on April 10, 2007, 11:13:34 AM
Moodflow, is it possible the number of samples on your clouds/atmosphere are above 256? I usually use 16 to 32 on the atmosphere and 128 to 300 in the clouds. Of course, the clouds seem subjective. If we want all the grain in the clouds gone, then all of the quality settings are adjusted upward to some degree. That degree is where some of us have gotten into similar troubles. And, in case you don't know (you probably do), the clouds' setting for Ray Tracing doesn't usually need to be on and will definitely add time to the render.
Yea, the samples are at minimum 256, otherwise there is noticeable grain. Some of the tests I've done have had samples as high as 2000+. I am thinking I will stick with 256 samples across the board, with a .75 render quality on some of my next tests.
I literally have a folder of unfinished images due to their taking too long to render at a decent resolution.
Now here is the crazy talk: I like to work with atleast a 2:1 step down, where the native image size is 2X as large as the final (for detail compression). Higher ratios are even better, technically. So if I want an image of 1600x1200, the minimum native resolution 'should' be 3200x2400. I was able to pull this off on Terragen 0.9 and Bryce. I'd love to see a high quality TG2 image of this resolution - I bet its amazing! And for decent posters, the final would need to be around 3200x2400, meaning the native resolution would have to be 6400x4800 (for best results). This would be a job for the BlueGene/L supercomputer.
I think we just need some simple renderfarm options (or does this exist yet?). I'd go find some lower end PC's and set this up if it were implemented. Find some old 1GHz jallopies and string them together with NetBEUI. Watch my electric bill skyrocket in the name of kickass images!
Generally speaking the old "downsample trick" (working at 2x resolution) is unnecessary with TG2 now due to the improved detail and antialiasing options. In some cases it still helps but this is mostly due to needed improvements in the renderer, not a fundamental need for downsampling to achieve max quality. Ideally a renderer *should* output a properly detailed image without the need for downsampling and it should do so in less time than it would take to render the 2x resolution equivalent.
Of course you "can" create extremely detailed scenes and with the expanded capabilities of TG2 the possibility to do so is much greater. But you must also accept the increased render time as a result. You always pay a price for pushing the limits of what is possible.
- Oshyan
My thoughts on TG2 render times are well known ( just do a search here)....so, I won't add much to what has already been said here.
However when we look at this issue from the artists viewpoint, an interesting question comes to mind....
What do you as the artist consider to be an acceptable render time to produce a print quality 3200X2400 size image?
I fully understand that image complexity, objects, amount of water and reflections can dramatically affect times....so, the answer would not be precise....but, would you be happy with 70hrs, 100hrs, or 500 hours for a typical image of this size? What would be acceptable to you? As a hobby "renderist", what would you be willing to put up with in the way of render times?
If todays state of the art "consumer" computer still requires 150hours to produce a 3200X2400 render...is that acceptable to you? Will it affect your decision to purchase this software?
Secondly, has Planetside in their development of TG2, looked at this issue from this viewpoint?....in other words, have they considered render time implications on potential product sales...or, was the focus on producing a software package that gives the desired end result, without factoring in how long it would take to render that result?
A lot of questions....and I know there won't be a standard answer to any of them....but it would be interesting to hear what folks think about some of this....
What we are talking about here basically is economies of scale, that is the given length of time it takes to produce a high quality product that is market ready and and compete with the rest of its market place.
Given the magnitude of TG2 and the relatively small size of the development team in comparison to say that at Autodesk or Newtek say, I would say that what we have with the TG2 is a substantial achievement, but yes it is at the moment not without its share of problems such as slow rendering, but the rendering could be worse as any one who comes like my self form a Bryce background will tell you, even Vue in its early years was slow, but you expect troubles like this early in the life of new software how ever they do need to be resolved if your product is going to compete in the market place.
How Planetside decides to proceed with the development of Terragen depends on their business model and also weather or not more programmers are added to the development team at some latter point in time.
Production facilities are going to insist that TG2 render times come down by many orders of magnitude if they are to use TG2 as part of their pipelines, reduced CPU loading and better all round stability and memory management and render catching abilities to make TG2 more adaptable to the ever shrinking deadlines faced in production today.
Regards to you.
Cyber-Angel
Quote from: Oshyan on April 10, 2007, 05:22:08 PM
Generally speaking the old "downsample trick" (working at 2x resolution) is unnecessary with TG2 now due to the improved detail and antialiasing options. In some cases it still helps but this is mostly due to needed improvements in the renderer, not a fundamental need for downsampling to achieve max quality. Ideally a renderer *should* output a properly detailed image without the need for downsampling and it should do so in less time than it would take to render the 2x resolution equivalent.
Of course you "can" create extremely detailed scenes and with the expanded capabilities of TG2 the possibility to do so is much greater. But you must also accept the increased render time as a result. You always pay a price for pushing the limits of what is possible.
- Oshyan
Oshyan, I agree with what you say above, but this does not tackle the problem of outputting renders on paper. If you want to output a 3200 render off of a good printer it'll only come out at approx 10" wide.... not much use if you are trying to do posters (A1 = 33" wide) cause you need an image of at least 300dpi. Same also applies if Planetside wants to see some of their softwares images in "Print" (by that I mean books and magazines etc) For press print you would again need a minimum 300 dpi. If you want to step into the world of high quality "Fine Art Prints" then you are talking 4000 dpi upwards!!!! :o
Just a small example.... A friend in work liked one of my renders and wanted it framed at just under A1 size for her living room wall (this was a TG 0.9X image) In TG 0.9 the largest I could render was 6000 x 4500 on a twin 3ghz xeon workstation with 1 gig ram. At 300 dpi that is only 20", not the 33" required. Luckily enough I have a copy of Genuine Fractals PrintPro (most excellent software for upsizing bitmaps without loosing quality) so I was able to do the job. I can't imagine the spec of the PC I'd need to get a 6000 x 4500 render out of TGTP ;D
Don't get me wrong, I love TGTP and fully realise that it is still in beta (I beta test another commercial bit of software, so I am well aware of how the process works) and that things will get a lot better re. render times before its final release, but if one is serious about the print side of things (and wants to use TG2 commercially), then speed / size of render will have to be reduced / increased by huge ammounts, otherwise you are going to have to spend £4000 upwards on a PC to run a bit of £200 - 400 software!!!!
but the $200-$400 software would equal maybe $2000-$4000 compared to the other competition on the market ;)
Quote from: DiscoBall on April 11, 2007, 07:19:16 AM
but the $200-$400 software would equal maybe $2000-$4000 compared to the other competition on the market ;)
I agree, as far as value for money is concerned, TG is about as good as it gets, but if it wants to be taken seriously as a commercial tool, as opposed to just a hobbyst bit of software then the area of print production would have to be addressed, regardless of price. Screen viewing is all well and good, but at the moment, and for the foreseeable future, I can't sell anything I produce with TG because I can't render it big enough. I'm not having a go at Planetside etc by saying this, just playing devils advocate, but these are real concerns if they want their software to fly off the shelves when it is a finished product.
Mr. Miley is right and hopefully Planetside are all nodding their heads. :)
I never really gave it a thought, but I wonder if the Planetside Developers, Oshyan or Testers have ever created a scene with TGTP and printed a large 36x24 poster from a 1280 x960 bitmap output from TG? I would think, if print output was considered during the development of TG2 this would have been attempted. If so I would love to see the results.
There has to be a boxed version with a physical paper manual (Spiral bound so you can lie flat on you desk for reference please) not just the down load version of the software. Also when are we going to see Terragen have its own print advertising to increase awareness of the Terragen brand, I mean look at the rendering market (Not Landscape) you see full page adds for Maxwell Render and Vray both of which are made by small companies relatively speaking.
Whats the aim with Terragen make enough money to keep the lights on as it where or aim to be the world leader in Terrain creation software and give the mighty Vue product line a run for its money.
Terragen has in the past on some graphics forums been accused of not been a professional tool nor its images true digital art such things should not be seen as insults but rather the catalyst to make Terragen not just a good product, but a truly great product, if you want image correction you think Photoshop, If you want high end Animation and CGI for motion pictures you think Maya or Maybe XSI, let Terragen be the software you think of when you need digital Terrain not just for digital terrain creation but for terrain replacement to, give it the tools to match its terrains and lighting to a live action plate shoot on location etc.
A Strong product identification and image are essential today, but if you go further you could with enough work turn Terragen into a household name, a brand name product if you will and go beyond the meteoric and stand on the shoulders of giants and when you finish using it at the end of the working day you can say "That was a job well done".
Regards to you.
Cyber-Angel
Quote from: Buzzzzz on April 11, 2007, 10:11:06 AM
I never really gave it a thought, but I wonder if the Planetside Developers, Oshyan or Testers have ever created a scene with TGTP and printed a large 36x24 poster from a 1280 x960 bitmap output from TG? I would think, if print output was considered during the development of TG2 this would have been attempted. If so I would love to see the results.
Don't think I would ;D The last time I tried something like this it looked horrible. That was the point I was making in my 1st post. A 1280 x 960 render at 72 dpi would only be 17" x 13.33" anyway and 72 dpi is nowhere near enough to get a decent printout. You need at least 300dpi for that
Quote from: mr-miley on April 11, 2007, 11:04:07 AM
Quote from: Buzzzzz on April 11, 2007, 10:11:06 AM
I never really gave it a thought, but I wonder if the Planetside Developers, Oshyan or Testers have ever created a scene with TGTP and printed a large 36x24 poster from a 1280 x960 bitmap output from TG? I would think, if print output was considered during the development of TG2 this would have been attempted. If so I would love to see the results.
Don't think I would ;D The last time I tried something like this it looked horrible. That was the point I was making in my 1st post. A 1280 x 960 render at 72 dpi would only be 17" x 13.33" anyway and 72 dpi is nowhere near enough to get a decent printout. You need at least 300dpi for that
well I wouldn't either ;) I print and have sold my Images rendered with TG. The main reason for my post is that someone here keeps telling us that rendering large for quality isn't necessary with this version of TG. And I'm really having a hard time digesting it.
Maybe the key will be (I'm only guessing) that someone at Planetside has plans to include a variety of DPIs in the final product. Or maybe it isn't a priority. I would think the version coming out this year would include something like this.
Otherwise, I wonder too about its future usability. Not that I don't love the software. But, it must have a target audience in someone's mind other than simple (though fun) hobby use. That would be a neat discussion here.
Just to add in reference to the render times. I don't really care if a render takes 8 days if it's something I think is worth the Wear and Tear on my Processor, Energy Consumption and just the long wait. My main concern is render size for quality prints. And I'm not referring to (if you stand back 6 feet and squint it looks good) ;) I want small quality detail at close up viewing. Sorry for jumping off topic of render times but that doesn't concern me as much as it may others. :)
Quote from: calico on April 11, 2007, 11:21:38 AM
Maybe the key will be (I'm only guessing) that someone at Planetside has plans to include a variety of DPIs in the final product. Or maybe it isn't a priority. I would think the version coming out this year would include something like this.
Otherwise, I wonder too about its future usability. Not that I don't love the software. But, it must have a target audience in someone's mind other than simple (though fun) hobby use. That would be a neat discussion here.
yes...at this point you cant really use TG2 for anything other than pure digital art. I am talking about TG2 to be used at film level....the only purpose of this software is visualization or maybe background use on some films/animations
Just to address a couple of questions - rendering at really high resolutions (poster size) generally causes crashes. A few of the testers have tried this and Planetside is well aware of the problem. The crashing issue with high resolutions will be fixed eventually, but even with the standard optimizations, rendering at high resolutions will always take longer than rendering at lower ones.
As for this:
Quote from: Buzzzzz on April 11, 2007, 11:11:56 AM
I print and have sold my Images rendered with TG. The main reason for my post is that someone here keeps telling us that rendering large for quality isn't necessary with this version of TG. And I'm really having a hard time digesting it.
In TG 0.9, if you rendered at 4000*3000, you often had to downsample to 2000*1500 or lower to get good looking results. With TG2, the downsampling process is no longer necessary.
I personally hope that TG2 will be used in more general purpose 3D environments. When the 16 shader limit is removed, animation controls are improved, and the renderer is optimized, I could see TG2 being used exclusively for certain types of CG shots such as air and possibly even space-based scenes.
As for this:
Quote from: Buzzzzz on April 11, 2007, 11:11:56 AM
I print and have sold my Images rendered with TG. The main reason for my post is that someone here keeps telling us that rendering large for quality isn't necessary with this version of TG. And I'm really having a hard time digesting it.
In TG 0.9, if you rendered at 4000*3000, you often had to downsample to 2000*1500 or lower to get good looking results. With TG2, the downsampling process is no longer necessary.
This is what Oshyan keeps saying, but this still needs to be Proven to me. Do you print your TG Images?
Me, too. I believe it already has been used this way, correct? I think it was Star Trek. I would like to see TG2 become mainstream. I'm sure you do, too...and it's without question the Planetside staff would like this.
It would just be nice to have clarification about DPI control, even if (and it probably is) at a later time this year.
It's a tough business and I hope it only gets better (rather than worse) for the Planetside programmers. This thing is a jewel.
Quote from: old_blaggard on April 11, 2007, 11:51:30 AM
I personally hope that TG2 will be used in more general purpose 3D environments. When the 16 shader limit is removed, animation controls are improved, and the renderer is optimized, I could see TG2 being used exclusively for certain types of CG shots such as air and possibly even space-based scenes.
Buzzzz - I haven't printed them yet, but just viewing them on the monitor, TG2 images are as detailed without downsampling as TG 0.9 images were with it. It could be that for your purposes you *do* need to downsample your TG2 image. However, a similar TG 0.9 image would need to be downsampled even more to achieve the same level of detail.
Quote from: old_blaggard on April 11, 2007, 03:16:14 PM
Buzzzz - I haven't printed them yet, but just viewing them on the monitor, TG2 images are as detailed without downsampling as TG 0.9 images were with it. It could be that for your purposes you *do* need to downsample your TG2 image. However, a similar TG 0.9 image would need to be downsampled even more to achieve the same level of detail.
I agree with you that TG2 is more detailed than TG.9 rendering the same sized images. I just don't think that one can render a small (say 1288x960) with TG2 and print a large poster from it.
Even with the improved detail of TG2 one would probably need to render @ 3600x2700 to get a good 24x18 print. I say probably because I don't know this for fact. Whereas with TG.9 I had to render 7200x5400 for that size image.
I guess I will have to do a test with a tree population, 3D clouds, etc., rendered @ 3600x2700 to determine the render time.
And I do understand completely that this version of Terragen is in alpha and no where complete. However there seems to be a really long long way to go for this Product to be complete in 2007 ? We are already a quarter of the way through the year and have received only "One" what I consider minimal update.
Buzzzz -
I imagine you know all this, but there is a non-linear relationship between image size and rendering times, controlled, I suspect, by the amount of system RAM.
I tried rendering a fairly complex scene (2048x2048)with a large number of objects at a moderate GI detail level. It hadn't completed 20% of the initial rendering pass when I aborted it after 30 hours or so.
I tiled the same scene into quarters, 1024x1024. Each quarter completed in an average of about 8 hours, running on the same machine. (!!!!)
I suspect that the larger image, coupled with the large number of objects, overwhelmed the machines RAM and it had to do an inordinate amount of swapping to the disk, where the smaller images didn't do this to such an extent and so the rendering process was MUCH faster. The exact parameters would depend on your machine, of course, but it's worth looking into. There is likely some level of tiling that will give you much better performance than trying to render the image as a whole.
Quote from: Harvey Birdman on April 11, 2007, 04:24:04 PM
Buzzzz -
I imagine you know all this, but there is a non-linear relationship between image size and rendering times, controlled, I suspect, by the amount of system RAM.
I tried rendering a fairly complex scene (2048x2048)with a large number of objects at a moderate GI detail level. It hadn't completed 20% of the initial rendering pass when I aborted it after 30 hours or so.
I tiled the same scene into quarters, 1024x1024. Each quarter completed in an average of about 8 hours, running on the same machine. (!!!!)
I suspect that the larger image, coupled with the large number of objects, overwhelmed the machines RAM and it had to do an inordinate amount of swapping to the disk, where the smaller images didn't do this to such an extent and so the rendering process was MUCH faster. The exact parameters would depend on your machine, of course, but it's worth looking into. There is likely some level of tiling that will give you much better performance than trying to render the image as a whole.
Thanks for your input HB, and in fact my last render Sedona http://buzzzzzart.com/Gallery/Buzzzzzs-TG2/Near_Sedona_copy (http://buzzzzzart.com/Gallery/Buzzzzzs-TG2/Near_Sedona_copy) was rendered in 6 crops because TG was crashing trying to render the entire scene at once. Many of my renders have been done this way.
Yeah, I figured this wouldn't be news to you, but thought I'd toss it out there anyway.
It seems like there are a lot of misconceptions here as well as some apparent misunderstandings of things I've posted previously. I'll try to address each issue in the order it was posted. This will be a long message as there is a lot to reply to. Skip down to where I mention your name if you're only interested in replies to your questions. ;)
First, to mr-miley: I think you will find that the vast majority of prints above portfolio size (roughly 8x11) are *printed* at 100-200dpi, usually about 150. 200-300dpi is often requested for the source files and in some cases provides better results depending on how the printer RIP (http://en.wikipedia.org/wiki/Raster_image_processor) (raster image processor) handles the bitmap processing. But usually this is just specified so that there is a buffer and the RIP can downsample appropriately. Ultimately a 1:1 render to print resolution is usually most ideal and efficient, but you do need to know the details of the printer's capabilities.
Furthermore a high quality printer with a sophisticated RIP can do as much as 2x upsampling with very good results, primarily because it's done at print time and in a way that meshes best with the capabilities of the printer. With that in mind if you aim for 150 or at most 200DPI you should be fine, as long as your print target is poster-sized (above 8.5.x11).
I'm not sure where you get the 4000dpi requirement for "fine art" prints (no generally available printers that work at sizes above 4x6 are capable of *true* 4000dpi printing period). Art prints are generally not printed above 600dpi and the difference between 300 and 600 is literally indistinguishable at a normal gallery or home viewing distance. The vast majority of "art" prints are made at 300dpi or less. For example the well-known CHROMIRA and Giclée print processes are targeted at 300dpi natively by most print houses. These are processes used for prints in the finest art galleries the world over. You may find these articles of interest http://www.allpconline.com/giclee_dpi.htm http://desktoppub.about.com/od/resolution/a/lpichart.htm
There is also a significant difference between the fundamental process used for home inkjet printing and most production processes such as those used for commercial-grade posters (so "dpi" capabilities of a home inkjet aren't going to equate to the same level of quality in a commercial print). Furthermore there is a difference between both of those and true "fine art" prints (which will cost you a good deal more than an average poster print at $15-30). Source file resolution above 600dpi is seldom necessary or advantageous in any of them however.
It's possible you're simply confusing scanning resolution (for scanning slide or traditional film for example) with print resolution. Slide scanning is routinely done at 4800dpi and above.
The problems you had rendering in TG 0.9 at sizes above 6000x4500 were probably related to memory and may have been solveable by reducing the render buffer size to something like 5mb. 1GB of memory is not very much to be rendering at such resolutions (and is surprisingly low for a dual 3Ghz Xeon). The image buffer alone, not counting antialiasing buffers and other memory needed for rendering, is 81MB at 6000x4500. That gets multiplied by your antialiasing level by I believe up to 4x (for Ultra) and is combined with memory needed for terrain rendering, render buffer, etc.
From my own testing on a machine with a similar amount of memory I found the maximum possible render resolution was about 11,500x11,500. This was consistent across multiple scenes and generally required use of a 5mb render buffer. I tested on another machine with 3GB of RAM and found the same results so this seems to be the practical limit with TG 0.9 on a PC, regardless of additional memory. On a Mac use of the Render To Disk feature made higher resolutions possible. We plan to include a similar feature in TG2 for the final release.
In any case you can rest assured that we are well aware of the issues with render time and know what level of improvement is necessary to be competitive in the market. Optimization is clearly needed and is a large mark on our "to do" list before final release.
Terragen 2 is a combined value proposal of quality, capabilities and ultimately speed. If you examine only one aspect - speed in this case - then it may not be competitive, even in the final release. But ask yourself if you can achieve the same level of quality and realism in any other application and in lesser time. If the answer is yes then perhaps TG is either not appropriate for your needs, or is simply non-competitive. The latter is something we at Planetside have to struggle with on our own, and we have full intentions to bring performance up to the best level it can be by the time of the final release later this year. Your own choices in applications for your work are up to you, of course. But it only makes sense to reserve judgment on the product's commercial applicability until the final release with all features and optimization in place (although several major effects studios are already using it in production even in its current state).
Buzzzzz: Certainly rendering for print output is being considered in TG2's development. I haven't done any specific print tests at this point, but I'm very familiar with the resolution that is fundamentally necessary to produce quality prints. I have printed photo-quality output from a variety of home-oriented inkjets as well as local print shop processes, online "on demand" print sources (CafePress, Zazzle), as well as "fine art" print sources. I have no illusions that the higher quality of TG2 output as compared to TG 0.9 changes the fundamental need for high resolutions in print source files and I think it's a bit strange to have concluded that from my earlier messages.
What I was actually saying saying is only that the downsampling approach used to get maximum quality with TG 0.9 should no longer be necessary with TG2. I consider downsampling to be a separate consideration from baseline resolution necessary for a given print size, so I was not at all implying that you could use say 72dpi output from TG2 to get equivalent results of a 150dpi (*after* dowsampling) TG 0.9 render. Rather I am saying that an equivalent 150dpi render in TG2 could be made natively at the target resolution, while you would need to render at twice that resolution and then downsample to get the same overall level of quality with TG 0.9.
The reason for this is that TG 0.9 lacked appropriate levels of detail, antialiasing, etc. TG2 can theoretically use any level of detail, AA, etc. that you want, and in general using an appropriate balance of detail settings will get you a high quality image in less time than rendering at 2x resolution and downsampling. You can count on a 4x increase in render time for every doubling of render resolution in each dimension (the total pixel area is 4x more). Increasing AA, detail, and other settings in appropriate measures can give you the necessary detail with a potentially much lower increase in render time.
Also just as a small note, don't read too much into the specific features, fixes, etc. that you see in the updates. Development is not a strictly linear process and there are many things still cooking in the ovens that will come out as more significant changes in the future, closer to release. Feature implementation is not likely to be a linear process so you shouldn't necessarily expect to see a steady growth in that regard. The end result is what matters.
Cyber-Angel: We do plan a boxed version of TG2 but many high-end, expensive products are sold completely digitally now and the digital distribution approach is more and more widely accepted. I don't think boxed distribution is any kind of pre-requisite for a successful high end graphics product.
As far as marketing, do you really think it would be appropriate to spend significant amounts of money on high profile marketing efforts at this point with the product in its current state? Of course we will market the product properly; after all, it's in our best interest. In fact the Technology Preview is part of our marketing plan and we feel it has done a great job of accomplishing our goals. But it makes little sense to have a huge marketing push at this time and I think if you look at the history of any significant new product (Maxwell and Vray are both good examples), they absolutely did not start out with such high levels of marketing when they were as early in development and sales as we are with TG2.
Both Maxwell and Vray are now more established, with lots of sales, high profile clients, etc. They can afford to pay for that level of marketing and their level of income demands it in order to ensure necessary levels of market expansion. For us it takes a much lower level of marketing to impact our registration levels to a significant degree, and this fits nicely with the smaller overall company and income levels we are working with. If we had multi-million dollar profits yearly we might want to do big print ads, but for now we're starting small and building the brand appropriately, not jumping the gun straight away with expensive advertising before we even have a final shipping product.
I think if you look at where we are in development and in the market then you'll agree this makes sense. If TG2 is released commercially and you still see a lack of marketing *at that time* it would be a much more appropriate time to call attention to it.
calico: Having native DPI control is really unnecessary for any rendering application. It is often included as a convenience, to help people to know what resolution is necessary for a given size print, but the actual DPI specification in rendering has no bearing whatsoever on the resulting render quality. It is only the specific render resolution that matters. You could render an image at 6000x4500 and 72dpi, which equates to an 83x62 inch print, and then simply adjust the DPI later in your image editor (as most people must do anyway) to get a 300dpi print at 20x15 inches. This is very common. Likewise you could do the reverse, if you wish.
The dpi specification in an image file is largely meaningless until it is printed and even then due to print scaling and other factors it often doesn't translate. It is useful mostly for reference and ease of setting appropriate render resolutions for an application like TG2. As such we may include it as a nicety, but it's certainly not required for good quality print output.
- Oshyan
Thanks, Oshyan. This makes sense and I apologize for my clear lack of graphical science education. This has been very helpful. Whether it was meant to be negative by anyone else isn't what I'm saying when I say I hopefully won't get dragged into any more of these negative fubars. It was my ignorance that led me to ask the question I asked and not because I believe that Planetside is stupid or assinine. Hope this clears it up.
Onto better things...
I didn't mean to imply any negative connotations, just trying to answer questions as best as possible. Glad things are cleared up. ;D
- Oshyan
Thanks Oshyan for the time taken to clear things up. For some reason I was under the impression it was being said that we didn't need to up size at all. If I understand now double should be fine?
You know me, Mr. Negative. ;D
Hmm, I'm confused about your question Buzzzz. Essentially you should just render to exactly the resolution that will give you your desired print size at the target DPI. So if you want to print 8x11 at 300dpi, you should render at 2400x3300 in TG2. No down or upsampling necessary. If you wanted to do a poster of 20x30 inches at 150dpi you should render at 3000x4500. This is all assuming your quality settings are correct to get desired quality, of course.
- Oshyan
QuoteI just don't think that one can render a small (say 1288x960) with TG2 and print a large poster from it.
Nor would I, No renderer, whether its TG2, Bryce or what ever, is capable of doing that. What pixel output would you expect from a digital camera, to give you quality
and print a large poster? 10+M at a minimum I would suggest. The above render is only about 1.24M, can't make silk purse out of a sows ear. Not that I suggest that Terragen is a sows ear I might add, quite the contary.
.