Just did my first water render test.....OUCH.....
http://www.renderosity.com/mod/gallery/index.php?image_id=1363782 (http://www.renderosity.com/mod/gallery/index.php?image_id=1363782)
This was 800x600 with a P4 2.2GHz machine...quality setting at 1...render time 69hrs 35 minutes!
The simple atmo took only a few minutes....then.....it rendered the subsurface terrain, then layered the water over the already rendered subsurface terrain....can't see why it has to waste time rendering the subsurface texture if there is no water transparency yet.....
its probluably just in preperation for the future, and 69 hours of rendering times....wow just wow My max so far has been 11.8 of course surface renders take longer then orbital renders.
well i suggest everyone to NOT render water. The render times are obviously not optimized (yes water will take long even in final but not like this) and then the water doesnt really have any new functions excluding the 3D waves.
Nice render Gradient! Just thinking that 60 hours is rather long for a 800x600 render, however the one I posted on my site was rendered @ 1280x960 and took 36 hours mostly because of the large quantity of cloud samples. 3.02 p4 with Hyper Threading using only 50% cpu. I have had TG 9.43 take 8 days to render water scenes so TG2 might not be as bad as some think. Gradient? I'm wondering if you have ray traced shadows checked in quality atmo settings and render settings. If so I found that to really add to render times and I don't know if it's always needed.
@Jay; Just checked my tgd...I didn't have ray tracing checked...so, methinks its just the darn water!!!
I'm glad I didn't have any clouds in it or the thing would still be cookin'...LOL!
It's the longest render I've ever done...had a few 0.9XX ones go around 40 or 50 hrs. I was just hoping that we wouldn't have a power failure...then I would have "lost" it....in more ways than one!
I think I'll stick to dry land till things get sorted out....
Quote from: gradient on January 14, 2007, 01:16:38 AM
@Jay; Just checked my tgd...I didn't have ray tracing checked...so, methinks its just the darn water!!!
I'm glad I didn't have any clouds in it or the thing would still be cookin'...LOL!
It's the longest render I've ever done...had a few 0.9XX ones go around 40 or 50 hrs. I was just hoping that we wouldn't have a power failure...then I would have "lost" it....in more ways than one!
I think I'll stick to dry land till things get sorted out....
I'm thinking of rendering the land in 9.43, render a matching sky in TG2 and then blend the two together in photoshop. At least until I figure this fricking thing out! LOL
There must be something wrong. Of course water takes a little longer, but that's way too much. You wrote "quality setting 1". Do you mean the detail setting in the render/quality tab? As far as I know there was mentioned in some earlier thread that this would increase the render time very much, but wouldn't give that plus of quality. I always leave it at 0.5.
I rendered the following picture with lots of samples in the clouds and used the "Full render" preset setting. The water is a little rough but that's a matter of the roughness setting I think. It took about one and a half hour to render.
(http://img153.imageshack.us/img153/4015/cloudfrontocean2medzz0.jpg)
Hannes
sounds like some atmosphere or cloud samplings where very hi too....
A simple test. It took 21 minutes to render.
(http://img219.imageshack.us/img219/7064/watertestjt8.jpg)
Hi,
Quote from: buzzzzz on January 14, 2007, 12:27:12 AM
3.02 p4 with Hyper Threading using only 50% cpu.
I'd just like to point out that a P4 with Hyper Threading is not quite the same as true dual core machine, or one with multiple independent processor chips. When a P4 HT says it's using 50% it's really using much nearer 100%. An HT chip has one full processor and a "virtual" processor. In a somewhat simplified description, this virtual processor works by putting instructions into the gaps between the instructions the full processor is working on. Chips like the P4 often have gaps in their instruction queue or pipeline. The maximum theoretical performance increase you can get with HT is maybe 20%, realistically probably somewhere around 10%, compared to theoretically 100% for a chip with two full processors like a dual core one.
Windows sees the HT chip as two separate processors, even though the bulk of the processing is always being done by the full processor and the virtual processor doesn't contribute nearly so much. Due to this a machine with an HT chip will say it's using 50% CPU when really that means it's using 100% CPU. When TG2 is multithreaded you might get a 10% increase in rendering speed, maybe, whereas a true multiprocessor machine could potentially get a 100% speedup ( though it might realistically be more like 50-60% ).
I need to write a FAQ for this :). I just thought I would mention this to let you know your machine is running much closer to 100% CPU use than 50%. My PC is a 3.0 GHz P4 HT machine too.
Regards,
Jo
@Hannes; My Render quality tab was set at 1 and antialiasing set at 3
@King Tiger 666; There were NO clouds...atmo was set at quality setting of 16 samples....
As I said I used a few shaders to get the berg texture...it rendered all that first including the entire full subsurface parts...then layered over the water. The actual rendering part of water overlay didn't take that long....the most time was taken to render the subsurface parts....But, I don't know if the two are linked in computation?
@Hannes...try a multi-shadered surface with water shader at render quality 1 setting to see what you get for render time....also, what processor are you using?
@Jo......you didn't touch on water render times.....
69 Hours. Phew, I guess you turned up some quality sliders? This image rendered in 20mins 5s.
Did this one quickly in about 5 minutes. I do wonder how you got the iceberg lit the way it is. Did you put in an additional lightsource?
ps. I did do autolevels in PS to get the colours a little less hazy.
@3Dguy; Quality sliders and shaders were set as indicated.....also, don't forget...from what I read here elsewhere, you have a high end machine too!
Regards lighting....yes you are correct...an additional lightsource was put in from the front. I have found this to be the best way to accentuate areas that would be in the shadows...similar to a "fill flash" in photo terminology.
BTW...many thanks for your pillar tgd and explanation!
69 hours is ridiculously long. My computer could do top-quality renders involving water at hi-def quality in 10 to 20 minutes with the old Terragen. Now, with T2TP, it takes over an hour to render images with the 'Full Render' preset, which does not produce very large or particularly-high quality images (and it sometimes renders an image in which everything looks suspiciously purple or everything is white and nothing is visible - clearly errors!). Clearly a lot of bug fixing and optimisation needs to be done...
@Dark Fire; some difficult atmo settings used in combination with water often resulted in long render times with version 0.9XX as well....as Buzzzzz above indicated. You wouldn't get anything of "high-def quality" in 10 to 20 minutes with version 0.9XX.....
However, I do agree with you that some optimization clearly needs to be done (and Planetside had made this very clear at the outset)....I just hope that there can be adequate optimization accomplished, otherwise many of us will not be able to participate.....
I see you used quality set at one.. you don't need that. 0.6-0.7 is enough.
Quote from: 3DGuy on January 14, 2007, 07:39:43 PM
I see you used quality set at one.. you don't need that. 0.6-0.7 is enough.
In my opinion the difference beetween 0.8 and 1 it's still visible.
I attach some very similar render one with 0.8 Quality AA 3 and GI 3 3 and the other one with
Quality 1 AA 4 and GI 4 4 and slight little bit more of atmo quality.
btw agree that in some cases 1 quality is damnly slow to render: on the example i put, quality 1 render need 4x time if compared to 0.8
Regards Max
I agree - the difference is clearly visible.
yes, I take it the first is the one with the lower quality its defnialy clear.
regards
Will
Hi Superza,
Do you still have the .tgd files for that scene? If so, could you post them?
Many thanks!
Quote from: moodflow on January 15, 2007, 10:04:26 AM
Hi Superza,
Do you still have the .tgd files for that scene? If so, could you post them?
Many thanks!
I have TGD still in render so i'm not sure what of the two images is this .tgd i sent, so try to render it and see which of the two images goes out :D
I have a scene I've been trying to render for awhile now but after it's crashed 3 times about 7-8 hours into rendering, I've given up. It keeps crashing when it gets to the water too.
Quote from: gradient on January 14, 2007, 07:13:23 PM
@Dark Fire; some difficult atmo settings used in combination with water often resulted in long render times with version 0.9XX as well....as Buzzzzz above indicated. You wouldn't get anything of "high-def quality" in 10 to 20 minutes with version 0.9XX.....
However, I do agree with you that some optimization clearly needs to be done (and Planetside had made this very clear at the outset)....I just hope that there can be adequate optimization accomplished, otherwise many of us will not be able to participate.....
I am sure that your quality settings or AA; maybe both where set too high, if you set the Quality to 4 and the AA to 3, you can even throw in some GI and render at 2500 pix and you will seehow fast you can render the same thing at higher printable quality. ;D
if you wanna make a comparison you should render the same image.
changing the detail but also the AA and GI will not give you the correct information.
You should only change one at a time
Quote from: gradient on January 14, 2007, 03:56:48 PM
@Hannes; My Render quality tab was set at 1 and antialiasing set at 3
@King Tiger 666; There were NO clouds...atmo was set at quality setting of 16 samples....
As I said I used a few shaders to get the berg texture...it rendered all that first including the entire full subsurface parts...then layered over the water. The actual rendering part of water overlay didn't take that long....the most time was taken to render the subsurface parts....But, I don't know if the two are linked in computation?
@Hannes...try a multi-shadered surface with water shader at render quality 1 setting to see what you get for render time....also, what processor are you using?
@Jo......you didn't touch on water render times.....
@edlo;
As I indicated....my quality was set at 1 and AA was set at 3
@fmtoffolo; yes, of course.....but I am not in the mood to try a different setting and wait for let's say... 56 hrs to go by....then another at 42 hrs....etc.
Quote from: MeltingIce on January 15, 2007, 10:56:49 AM
I have a scene I've been trying to render for awhile now but after it's crashed 3 times about 7-8 hours into rendering, I've given up. It keeps crashing when it gets to the water too.
When you say 'crashed' do you mean Terragen stops responding or closes or something like that? I have never had the new Terragen crash, but I have wasted hours rendering images that come out with a purple tint or that just come out as a huge rectangle of one colour (blue and white are common colours with this error).
Quote from: gradient on January 15, 2007, 02:49:07 PM
Quote from: gradient on January 14, 2007, 03:56:48 PM
@Hannes; My Render quality tab was set at 1 and antialiasing set at 3
@King Tiger 666; There were NO clouds...atmo was set at quality setting of 16 samples....
As I said I used a few shaders to get the berg texture...it rendered all that first including the entire full subsurface parts...then layered over the water. The actual rendering part of water overlay didn't take that long....the most time was taken to render the subsurface parts....But, I don't know if the two are linked in computation?
@Hannes...try a multi-shadered surface with water shader at render quality 1 setting to see what you get for render time....also, what processor are you using?
@Jo......you didn't touch on water render times.....
@edlo;
As I indicated....my quality was set at 1 and AA was set at 3
@fmtoffolo; yes, of course.....but I am not in the mood to try a different setting and wait for let's say... 56 hrs to go by....then another at 42 hrs....etc.
Yes Pal sorry about the misunderstanding I meant to say 0.4 Quality and rendering a 2000+ pixel image that is of course if you have the registered version.
@edlo....of course setting quality to 0.4 will reduce render times....but if you look at Superza's posts in this thread...you will see that there is already a substantial difference in Q between 0.8 and 1.0. Setting it at 0.4?....well it would look like crap.
Quote from: gradient on January 15, 2007, 03:11:55 PM
@edlo....of course setting quality to 0.4 will reduce render times....but if you look at Superza's posts in this thread...you will see that there is already a substantial difference in Q between 0.8 and 1.0. Setting it at 0.4?....well it would look like crap.
Not if you increase the pixel count to 2000+; this is rendered at 0.4 take a look http://www.deviantart.com/deviation/46256703/?&q=by%3Aedlo&qh=sort%3Atime+-in%3Ascraps it surely doesn't look like crap ;)
Ok, I gave it a shot. 1 piece was rendered at quality 0.7 and one at 1.
(http://temp.theglasseye.nl/water_test.jpg)
That's 15 mins versus 1h22m
yea look pretty much the same
regards,
Wil
@3DGUY In your case 0.7 and 1 is quite similar also if the lighting is clearly better on the second one;
More similar than in my tests, on which as Edlo said the different settings are unfortunately more than one.
4 sure that increases also GI and AA improves render using different way.
Actually i found enough quality only using no less than 0.85 Quality, AA 4 and GI 3 or 4;
Atmo about 48 samples and clouds at least 0.8-0.9.
I think it's also a question of personal taste.
Best Regards
Max
You like the second one better? Guess what.. That's the one with 0.7 detail ;)
Detail setting is the only thing I changed for that.
The only diffrence I saw was a slight varience in reflection.
onguards,
Will
Quote from: edlo on January 15, 2007, 03:25:11 PM
Quote from: gradient on January 15, 2007, 03:11:55 PM
@edlo....of course setting quality to 0.4 will reduce render times....but if you look at Superza's posts in this thread...you will see that there is already a substantial difference in Q between 0.8 and 1.0. Setting it at 0.4?....well it would look like crap.
Not if you increase the pixel count to 2000+; this is rendered at 0.4 take a look http://www.deviantart.com/deviation/46256703/?&q=by%3Aedlo&qh=sort%3Atime+-in%3Ascraps it surely doesn't look like crap ;)
@edlo;
In order to increase the pixel count, it will cost me considerable $$$$...something I currently can not afford, nor would I be prepared to do given the current state of development of TG2.
Quote from: 3DGuy on January 15, 2007, 06:04:10 PM
You like the second one better? Guess what.. That's the one with 0.7 detail ;)
Detail setting is the only thing I changed for that.
OMG anyway in that close visual is quite impossible to appreciate the better depth that the 1 details give.
In my opinion the light of the second looks more "photographic" :)
I'm quite sure that in complex scene the detail on far part is more incisive.
Best Regards
Max!
Quote from: gradient on January 15, 2007, 07:40:59 PM
Quote from: edlo on January 15, 2007, 03:25:11 PM
Quote from: gradient on January 15, 2007, 03:11:55 PM
@edlo....of course setting quality to 0.4 will reduce render times....but if you look at Superza's posts in this thread...you will see that there is already a substantial difference in Q between 0.8 and 1.0. Setting it at 0.4?....well it would look like crap.
Not if you increase the pixel count to 2000+; this is rendered at 0.4 take a look http://www.deviantart.com/deviation/46256703/?&q=by%3Aedlo&qh=sort%3Atime+-in%3Ascraps it surely doesn't look like crap ;)
@edlo;
In order to increase the pixel count, it will cost me considerable $$$$...something I currently can not afford, nor would I be prepared to do given the current state of development of TG2.
Ah my bad then... I assumed you had the registered version because of the quality of your work, it would be a very nice investment for you to have the registered version though.
Quote from: edlo on January 16, 2007, 10:04:48 AM
Ah my bad then... I assumed you had the registered version because of the quality of your work, it would be a very nice investment for you to have the registered version though.
T2TP is too expensive for most people if it is not going to be used commercially.
This came up in another thread, but I whole heartedly believe that as a hobby, a license of TG2 is relatively cheap compared to other hobbies. Compare it to photography (*very* expensive), video gaming (console + many games), collecting *anything*, model building and painting, etc.
I could go on with many hobbies that are much, much more expensive, and chances are you would easily spend lot's of money on other things. But many people believe that because it's software on a computer, that anything other than free is too much. I don't understand...
Quote from: njen on January 16, 2007, 10:59:50 AM
This came up in another thread, but I whole heartedly believe that as a hobby, a license of TG2 is relatively cheap compared to other hobbies. Compare it to photography (*very* expensive), video gaming (console + many games), collecting *anything*, model building and painting, etc.
I could go on with many hobbies that are much, much more expensive, and chances are you would easily spend lot's of money on other things. But many people believe that because it's software on a computer, that anything other than free is too much. I don't understand...
Photography can be cheap if you are good at spotting bargains and you are very patient. Video gaming isn't even that expensive when you consider that the PS3 and other consoles are now designed to be the hub of your media experience. You can collect Tube Maps for free. With TG 0.9 you could render near hi-def quality images and animations without paying. Programming is generally a free hobby, and you can even have a website for free these days.
Anyway, many hobbies are free and many people give out free software. I do not think T2TP should be any cheaper than it is now, but I cannot justify spending a huge sum of money on it because, if I am not going to use it commercially, I would require an army of computers and a working command line to get enough enjoyment out of it and, unfortunately, I do not have a load of computers and T2TP does not yet have a working command line.
Once TG2 is released it should have a working command line and it should have been optimised quite a bit, so it may be worth purchasing then.
Sorry folks... I didn't want, nor did I think this would turn into a thread concerning the price vs. value of TG2. That is something each one of you must decide on your own.....
Now, let's get back to water render times.....hmmmmm, no comment from Jo yet.....
heh yea I think you guys scared him off ::) But I know that using the water shader with a non water object (like terrian or clouds) the render times are good....yea that was off topic too.
regards,
Will
A further thought on render times....it is becoming apparent that in order to reasonably be able to use TG2's features....3d clouds, water and imported objects all require enormous processing power.
I wonder what Planetsides estimate of potential optimization is....it may affect registered version purchasing decisions.....In other words if we can only reasonably expect a 25% improvement in render time....it may not be enough to convince folks to purchase the software if their CPU systems are not state of the art.
Jo?...Java?...anyone?
Quote from: gradient on January 17, 2007, 08:46:45 PM
...a 25% improvement in render time...
Optimisations are really badly needed - T2TP takes too long to render reasonably good images.
Quote from: gradient on January 17, 2007, 08:46:45 PM
A further thought on render times....it is becoming apparent that in order to reasonably be able to use TG2's features....3d clouds, water and imported objects all require enormous processing power.
I wonder what Planetsides estimate of potential optimization is....it may affect registered version purchasing decisions.....In other words if we can only reasonably expect a 25% improvement in render time....it may not be enough to convince folks to purchase the software if their CPU systems are not state of the art.
Jo?...Java?...anyone?
Btw i think that Tgd looks like to be at least with great scalability.
25% could be a great improve. It will be also the multicore support, so for example using a dual core processor we could expect another 60% boost performance or even more using a quad core processor, If this is not enough, i read that in final version network rendering will be supported.
Suppose to have a 100 hours render with the actual version;
with 25% of speed optimization we speed up to 75 hrs,
the dual core owner can expect 45-50 hours and the quad core owner, 22-25 hours.
if this is not enough with network rendering and another quad core time should be around 11-13 hours!
So render time becomes only matter of hardware power and unfortunately money to bought it!
Best Regards Max
@Max....the 25% is just a number I made up....It is not based on any information!
Really, what I am saying is that to produce a print quality(3200X2400) of the same image that took me 70 hrs (for 800X600)...would take 16 times as long...in other words, 1120 hours (or about 47 days!)!!!!
Even IF there was a 25% optimization success...it would still take 840 hours!
Based on the render time benchmark database...even IF a CPU was 3x as fast as my machine....that same render would still take 210 hours! Keep in mind, my render had NO clouds and NO objects, few shaders....it was technically very simple!
Unless SIGNIFICANT optimization is achieved in future builds of TG2, render times will be unacceptably long for "casual" users....and, I think that may affect registered program purchase decisions.
Quote from: gradient on January 18, 2007, 03:04:13 PM
Unless SIGNIFICANT optimization is achieved in future builds of TG2, render times will be unacceptably long for "casual" users....and, I think that may affect registered program purchase decisions.
That's where my software comes in. I was a casual user of the old Terragen but I wanted to produce high quality animations, so I wrote software that lets an entire network help out. I have started upgrading the software so that it will work with T2TP - an extreme beta is available on my website. However, the beta is useless at the moment because the T2TP command line is broken. :(
guyz i think u are forgetting that TG2 is meant as Professional tool (thus you are expected to own workstation). Its not the amateur TG1 anymore. I honestly dont know how long does take renders with vue/houdini, but i dont think there will be some big difference.
I don't know about hudini but VUE6 Infintite is faster but it has been optimised andthere are no overhangs or plantary renders. Also to reiterate what so many others have siad: This is just a tech preview wait till Terragen 2 is released to judge the final product.
Regards,
Will
Quote from: gradient on January 18, 2007, 03:04:13 PM
@Max....the 25% is just a number I made up....It is not based on any information!
Really, what I am saying is that to produce a print quality(3200X2400) of the same image that took me 70 hrs (for 800X600)...would take 16 times as long...in other words, 1120 hours (or about 47 days!)!!!!
Even IF there was a 25% optimization success...it would still take 840 hours!
Based on the render time benchmark database...even IF a CPU was 3x as fast as my machine....that same render would still take 210 hours! Keep in mind, my render had NO clouds and NO objects, few shaders....it was technically very simple!
Unless SIGNIFICANT optimization is achieved in future builds of TG2, render times will be unacceptably long for "casual" users....and, I think that may affect registered program purchase decisions.
I dont' know Grad, May be there is some strange issue in your render; the fact that takes ages to render the underwater surfaces looks strange, my last in ReRO' with water, was not render the underwater surfaces: i was humbling about some setting similar to the 0.9 flag "Back face Culling" in your case.
I was not saying TDG is fast, my last took 57 hours with quality 1 @1600*1000, and i was render two new renders about 30 and 90 hours repsectively and they are to half render, to that point is damnly slow(especially with two lightsource)!!!!
But for a professional use TGD will have a great scalability.
Take in mind that with a multitrheaded version of tgd, a quad core processor within few months will costs few hundred dollars and could be 6-8 time faster than your actual cpu, and with another similar computer and network rendering could be a theorical 12-16 time faster.
It could be not so bad if the 1000 hours could be 70!
Best Regards MAX
@Superza, Will, buchvecny, and Dark Fire... I guess what you guys are all saying is confirming my suspicion....that is that in order to produce "print quality" images from TG2 it will require not only the cost outlay for the reg'd version....but also significant hardware $ outlay in order to bring render times within a reasonable limit.
Perhaps buchvecny's words are most telling...."TG2 is meant as Professional tool (thus you are expected to own workstation). Its not the amateur TG1 anymore"
Unfortunately if that is the case,.....As MOST of us don't have render farms, we won't be able to play the TG game anymore.....
It is somewhat sad to think that the "hobby" user....the one that used TG 0.9xx, helped promote the product and, to some degree helped develop TG2 to it's current state...is now being shut out of the game....
Quote from: gradient on January 18, 2007, 03:04:13 PM
Based on the render time benchmark database...even IF a CPU was 3x as fast as my machine....that same render would still take 210 hours! Keep in mind, my render had NO clouds and NO objects, few shaders....it was technically very simple!
Your render was technically simple *as far as you know*, but from it sounds like in your description earlier in the thread it was actually the land under the water that took much of the time in rendering. The culling of features that will be rendered over will be improved in the future, so that will speed this sort of thing up a lot, but the real question is whether it's truly the water that took so long, or something else in your scene. You have described the detail settings, but they are by far not the only thing to really impact render time. It's clear you have used very extreme displacement for example, which will take a while to render in itself (even without water). If you used an additional Compute Normal/Compute Terrain for some purpose that could help explain such long render times.
It would really help to see the .tgd file to determine whether there is some issue here, a bug, incorrect settings, etc., and I think it's only fair not to assume it's a fundamental TG2 problem without a full review of the settings used. You may send to support@planetside.co.uk if you would like us to take a look and see if there are any notable issues.
As far as hardware upgrades go, I'm afraid you'll just have to accept the reality of this as a necessity for working with any modern application - TG2, Vue 6, etc. Programs are developed to take advantage of the hardware of their day, not yesterday's and only occasionally tomorrow's. The latest games that are released require good hardware for good performance and the same is true for Terragen 2 or any application, particularly a 3D rendering system.
It's unfortunate that you have to work with an older computer (a 2.2Ghz P4 would have been new perhaps 5 years ago, which is quite a while in the computer industry), but the simple fact is that the advanced capabilities of programs like TG2 are not possible to accomplish in reasonable time using older hardware. TG2 is capable of so much more than TG 0.9 - displacement alone is a huge difference - and these capabilities must naturally come with an increase in rendering time. Doing more in the same time would be a rare miracle of coding and I think you can count the number of times this has happened in commercial applications on two hands.
Fortunately I think it's fairly cheap to upgrade to something decent, which would give you a great speed boost. Since TG2 is not multithreaded at present you could get a very cheap, fast single core Athlon 64 (2.4Ghz for perhaps $80, equivalent of a P4 at 3.2+Ghz) - an entire system upgrade would be less than $500, and if you kept your hard drive, case, optical drive, etc. it could be much less. Or you could go dual core and still keep it very cost-effective - $700 or less for a fully new dual core Athlon 64 X2. Yes these are not really small amounts, and I wouldn't recommend you only upgrade for TG2, but if you have a need for better performance overall the price can be justifiable. Many people spend much more than that for computers that hardly do more than email. :D In any case I think it's pretty clear though that any modern program is going to be making greater and greater demands on your aging system and it would be good to upgrade for your overall computing experience, if possible. I hope it's something you can do in the near future.
As far as optimization goes, well it's very difficult to give any specific numbers. I can tell you that there are many areas of optimization, from improved culling (not rendering parts that will be covered up) - which will effect all aspects of scene render time - to improvements in specific areas of the renderer such as reflections and volumetrics. The latter two will of course only affect scenes that have those features - water and 3D clouds, specifically. So the improvements may depend in part on what aspects of TG2 you focus on, or that impact your current render time the most.
Another area of optimization is multithreading, which will probably create the greatest improvements, at least on multi-core/multi-CPU systems. Fortunately multi-core is the way of the future and dual core systems have already surpassed single core for new purchases (even for laptops), and will quickly become the dominant CPU in modern systems (systems newer than 3 years). There are also possibilities in optimizing for SSE/2/3 and other specific CPU features, although I don't expect these to necessarily show large gains. I won't place any kind of official number on any of these optimization paths as there really is no way to know, but I am confident that overall TG2 render time will improve a great deal even in single threaded rendering, and I'm sure my dual core Athlon will really crank out renders once TG2 is multithreaded. Anyone running a dual core or better CPU should see an amazing difference in the final version with mulithreaded rendering.
In regards to professional use, although it's true that most "professionals" have access to more powerful computer hardware, it's also true that they are often much more demanding of low render times than the average hobbyist. A professional would never put up with 69 hours of rendering time for a single frame. Even the 20-30 hours Gradient and others endured under TG 0.9 would usually be unacceptable for professional use. As we are indeed partly catering to the professional market optimization is definitely a great concern for us. In a way you can perhaps thank the professional users for they probably set a higher standard than anyone, and we do need to live up to it if we hope to compete. I think in general you will only benefit from the work done in catering to this market - it will mean faster render times, more power and flexibility, broader interoperability, and better ultimate image quality.
Finally, let me make clear that nobody is being "shut out of the game", certainly not intentionally. We are trying to make a modern, powerful, flexible 3D rendering system with broad appeal. The simple fact is that this is a demanding task for a computer. Any application that even comes close to what TG2 can do is going to be taxing on your system, that's just the way it works. We are doing our best to make TG2 accessible and enjoyable to all, and that includes optimization as well as UI improvements and potential simplification.
It is unfortunate to hear that any of our users feel we may be abandoning them as that is certainly not our intention. I hope you can understand the difficulty here - we want to support those who were used to TG 0.9, both in user interface and performance, but these things present a problem for the improvements that are necessary for TG2 to compete and excel. It simply wasn't possible to make "TG 0.9 with more" or we would have. To get the kind of power needed we had to basically start over, so many things will be different. We are balancing it all as best we can and I do hope you'll stick with it, at least on the sidelines, to see where it takes Planetside and the community. I have a feeling you may ultimately enjoy where we end up. I certainly hope so. :)
- Oshyan
Thank you for your response Oshyan.
I guess time will tell.....regards optimization successes.
As I indicated to you in our dialogue over at Renderosity, I do have a newer machine.........unfortunately it is still a 1 1/2 year old P4 2.8Ghz HT. I have not yet installed TG2TP on it yet....I intend to do so to see what differences in speed I can expect....but from looking at the benchmark database, I am not too hopeful that it will help me too much.
Quote from: gradient on January 18, 2007, 09:22:26 PM
As MOST of us don't have render farms, we won't be able to play the TG game anymore.....
You don't need a render farm. My software can run off a USB device so you can render wherever you go. Most workplaces have computers these days, so you could render while at work.
I am considering creating an internet version of my software, which should solve problems people are now having with limited processing power, but I am currently concerned about security - if you leave your files in a place where anyone can access them and help with rendering, they may be stolen and/or abused. Even if I were to somehow come up with secure methods of transfer and storage of your data, I would have to keep the internet-based code (most likely PHP) private, thus limiting jobs to being hosted on servers I own.
I was just saying that beween the two Terragen and VUE VUE is faster but thats becuase it has been optimized lets not knock Terragen 2 before it is released. But then again I have a Dell XPS because Im an avid gamer so my box is pretty nice.
Regards,
Will
Quote from: Dark Fire on January 19, 2007, 01:42:19 PM
Quote from: gradient on January 18, 2007, 09:22:26 PM
As MOST of us don't have render farms, we won't be able to play the TG game anymore.....
You don't need a render farm. My software can run off a USB device so you can render wherever you go. Most workplaces have computers these days, so you could render while at work.
Right, try that at most companies and you get fired on the spot.
Nahh, you just render in the background :)
Quote from: njen on January 19, 2007, 01:53:33 PM
Nahh, you just render in the background :)
Do you speak from experience? ;)
Regards,
Will
Quote from: njen on January 19, 2007, 01:53:33 PM
Nahh, you just render in the background :)
Uhuh, misuse of company property. There's plenty people fired for doing similar stuff with SETI@home or similar programs.
Hes just playing around 3DGuy don't worrie and if he aint well lets just say I hope his boss does no read this fourm.
Regards,
Will
Quote from: jo on January 14, 2007, 07:25:03 AM
Hi,
Quote from: buzzzzz on January 14, 2007, 12:27:12 AM
3.02 p4 with Hyper Threading using only 50% cpu.
I'd just like to point out that a P4 with Hyper Threading is not quite the same as true dual core machine, or one with multiple independent processor chips. When a P4 HT says it's using 50% it's really using much nearer 100%. An HT chip has one full processor and a "virtual" processor. In a somewhat simplified description, this virtual processor works by putting instructions into the gaps between the instructions the full processor is working on. Chips like the P4 often have gaps in their instruction queue or pipeline. The maximum theoretical performance increase you can get with HT is maybe 20%, realistically probably somewhere around 10%, compared to theoretically 100% for a chip with two full processors like a dual core one.
Windows sees the HT chip as two separate processors, even though the bulk of the processing is always being done by the full processor and the virtual processor doesn't contribute nearly so much. Due to this a machine with an HT chip will say it's using 50% CPU when really that means it's using 100% CPU. When TG2 is multithreaded you might get a 10% increase in rendering speed, maybe, whereas a true multiprocessor machine could potentially get a 100% speedup ( though it might realistically be more like 50-60% ).
I need to write a FAQ for this :). I just thought I would mention this to let you know your machine is running much closer to 100% CPU use than 50%. My PC is a 3.0 GHz P4 HT machine too.
Regards,
Jo
Ok Thanks for the info jo. :) Just thought I would mention that I tested running One instance of TGTP with Hyper Threading Enabled and Task Manager said that tgd.exe was using 50 % cpu as I noted earlier. So I rebooted and disabled HT and then tgd.exe was using 100% cpu as expected. I rebooted again renabled HT, started a First instance of TGTP which was back to using 50% cpu. So I thought What the Hell? I'm only using 50% so I started a Second instance of TGTP and cpu usage went up to 100% as expected right? 2 x 50% = 100% right? Didn't run that way for long but seemed to be ok. I main thing I like about Hyperthreading is that I can be rendering and doing pretty much anything else I want without any noticeable lag in performance. I thought about disabling HT again and and trying to run 2 instances of TGTP but was afraid to without a Fire Extinguisher near by. :o
Thanks Again jo!
HT only simulates 2 cores. Windows sees 2 cores of which 1 core is apparently using 100% of it's power. Using the simple math that windows uses, it 'thinks' 1 core @ 100% another core at 0% that makes 50%. But with HT it's not that simple. It's actually really close to 100% actual usage. HT doesn't mean you actually have 2 cores like the core duo 2 chips.
So in short, with HT the taskmanager is simply reporting the wrong figures.
Quote from: 3DGuy on January 20, 2007, 10:16:39 AM
HT only simulates 2 cores. Windows sees 2 cores of which 1 core is apparently using 100% of it's power. Using the simple math that windows uses, it 'thinks' 1 core @ 100% another core at 0% that makes 50%. But with HT it's not that simple. It's actually really close to 100% actual usage. HT doesn't mean you actually have 2 cores like the core duo 2 chips.
So in short, with HT the taskmanager is simply reporting the wrong figures.
You just said the same thing as jo only in different words and I understand that. I also know I don't have two cores because I built my machine.
In addition I performed a render test using the same tgd with and without HT and the render times were nearly the same. So, I guess what I'm trying to convey here is that with HT enabled I can render and run other processes without any ill effects on performance or render times. Which for me is a good thing.
Quote from: 3DGuy on January 19, 2007, 01:52:23 PM
Quote from: Dark Fire on January 19, 2007, 01:42:19 PM
Quote from: gradient on January 18, 2007, 09:22:26 PM
As MOST of us don't have render farms, we won't be able to play the TG game anymore.....
You don't need a render farm. My software can run off a USB device so you can render wherever you go. Most workplaces have computers these days, so you could render while at work.
Right, try that at most companies and you get fired on the spot.
I have never heard of that happening. I think whether it is misuse or not depends on what you do with the resulting render. If something is found with SETI, it could potentially damage some companies, so I can understand the fact that some companies would not like it.
I could search for registry entries T2TP makes and make it into a totally portable program, thus making its previous usage of it undetectable. If you are desparate I could also make a button of key combo that would kill T2TP and any associated software.
Also, realistically, a company should not care too much about you using waste that does not pose a security threat. For example, if a company was throwing out a load of blank paper (save the trees!) it would not be called misuse of company property if you offered to take the paper and actually use it. The usage of spare processing time can be thought of in the same way - the company cannot recover yesterday's wasted processing time so, if it is not going to be used by the company, it makes sense for you to use it providing you do not pose a security threat (and using T2TP is not a security threat).
I've been part a group that did something similar to SETI@home, which used distributed processing. That process also runs in the background at the lowest priority. There were several reports (from the Anandtech crew I think it was) of sysops that got fired just because they installed that software on every server.
If you're using spare cpu time, you're actually consuming more power. A cpu that's idle uses less energy than one that's fully occupied, in that respect you're stealing money. More power=more electricity=money. Fact is, most companies don't want you installing anything on their computers. It could even be in your contract. But a decently configured workstation won't even let you install anything, that is if the sysop knows what he's doing.
If your boss lets you, by all means, use workplace computers as much as you want, but if you do it without telling anyone, you could be in serious trouble (anything from a slap on the wrist to losing your job). It doesn't matter if a program poses a secutity threat or not. You just don't install 'random' software on company computers.
Quote from: 3DGuy on January 20, 2007, 06:22:33 PM
But a decently configured workstation won't even let you install anything, that is if the sysop knows what he's doing.
I've never met a good sysop. :D
Quote from: 3DGuy on January 20, 2007, 06:22:33 PM
It doesn't matter if a program poses a secutity threat or not. You just don't install 'random' software on company computers.
If I ran a company, I would allow distributed computing stuff to run on the company computers. However, I may never end up running a company...
Quote from: 3DGuy on January 20, 2007, 06:22:33 PM
There were several reports (from the Anandtech crew I think it was) of sysops that got fired just because they installed that software on every server.
If you're using spare cpu time, you're actually consuming more power. A cpu that's idle uses less energy than one that's fully occupied, in that respect you're stealing money. More power=more electricity=money. Fact is, most companies don't want you installing anything on their computers. It could even be in your contract. But a decently configured workstation won't even let you install anything, that is if the sysop knows what he's doing.
Firstly, I doubt anyone here would be stupid enough to start installing T2TP on every server in their workplace. Secondly,you are not always using more power when using the CPU more. Some CPUs (and maybe all, but I'm not an expert on CPU technology) power up different parts of themselves in steps. However, to exploit just the spare powered parts of the processor would require a truly amazing program.
Quote from: 3DGuy on January 20, 2007, 06:22:33 PM
...(anything from a slap on the wrist to losing your job).
It's amazing how closely related those two things are...
All processors use more power when used more. Why do you think a processor gets hotter when using 100% usage? That's because more power is required which in turn produces more heat. It's not different parts of the proc that get activated. What you're referring to is speedstepping which is used in laptops for instance. When it's idling or on battery power it switches to a lower frequency... to save power. According to your assumtions it wouldn't matter if you ran a background render on a laptop or not and still get the same batterylife out if it, that's simply not true. If that doesn't show you more power/energy is required than I don't know what will.
Quote from: 3DGuy on January 21, 2007, 11:55:35 AM
All processors use more power when used more. Why do you think a processor gets hotter when using 100% usage? That's because more power is required which in turn produces more heat.
My CPU gets hotter the longer I leave my computer on - it is not affected much by rendering. Maybe I just have good cooling.
Quote from: 3DGuy on January 21, 2007, 11:55:35 AM
According to your assumtions it wouldn't matter if you ran a background render on a laptop or not and still get the same batterylife out if it, that's simply not true. If that doesn't show you more power/energy is required than I don't know what will.
According to my assumptions if you were using a program that, for example, used the CPU at 99%, if I ran Terragen with a low priority so that it used up that final 1% the battery life would be unaffected. Obviously running a render would have a huge effect on tha battery life compared to a laptop running nothing and, besides, I rarely get
any battery life out of a laptop, so I can't test that. I was probably talking about a rare CPU type that I read about in PCPlus...
It gets hotter because your case warms up. In other words, you need some airflow in there. Measure the temp when windows is idling, then let tg run a render and check the temp after 5 mins, you'll see an increase of a few degrees.
Anyhow, this has less and less to do with the title of the thread ;p
Quote from: 3DGuy on January 21, 2007, 01:11:33 PM
...you need some airflow in there.
I have plenty of that (there are very few parts of the case without holes). I feel I didn't explain myself properly last time - my CPU gets hotter the longer I leave my computer on until it reaches about 40
oC. After that it stays at about the same temperature no matter what I do. The same thing happens with my GPU as well. As I said, maybe I just have good cooling.
Sorry, but I just don't buy that. Unless you have a kick ass watercooling system, there's no way that your cpu doesn't get hotter under full load, that's simply impossible with air cooled systems, no matter how advanced it is.
Unless you have some background program running already using 100% that is.
I think it's got something to do with the fact that I am running an old processor with a much newer case and fans.
Somehow we have progressed from water render times to water cooling... :D
Having an older processor has nothing to do with it. You're either reading it wrong or using the wrong software. I've seen and done enough tests and measurements to know there's no processor that doesn't get hotter under full load vs. idle.
Hot Water Springs! Great Idea for a Render! Thanks you two! ;)
Ok...bringing this thread back on topic.....
Let's see some HQ water renders from you guys...I want you to feel the pain...LOL!
Quote from: gradient on January 21, 2007, 04:37:11 PM
Ok...bringing this thread back on topic.....
OK...back to water... :(
Quote from: gradient on January 21, 2007, 04:37:11 PM
Let's see some HQ water renders from you guys...I want you to feel the pain...LOL!
Quote from: buzzzzz on January 21, 2007, 04:16:30 PM
Hot Water Springs! Great Idea for a Render! Thanks you two! ;)
I think some steaming pools of hot water in Iceland would take painfully long enough to render...
Quote from: 3DGuy on January 21, 2007, 11:55:35 AM
...It's not different parts of the proc that get activated.
Sorry for bringing this up again, but these days it is different parts of the processor that are powered up and down (http://news.bbc.co.uk/1/hi/technology/6354225.stm): 'It consumes 62 watts, and the cores can power on and off independently, making it more energy efficient.'
Anyway, back to water. I was right about the steaming pools - they took ages to render. Interestingly, the water in this picture (http://ccgi.emptosoft.plus.com/main_site/e107_plugins/autogallery/autogallery.php?show=Volcano_in_T2TP....png) didn't take too long to render, despite the fiddling with settings I did to make it as reflective as possible...
It's not parts of the processor.. it's the seperate cores that can be turned on/off. Each core is a CPU. Albeit that that all the cores are placed on one chip. Dual Core or Quad Core just means it's 2 or 4 processors on a chip ;)
Quote from: 3DGuy on February 13, 2007, 08:10:01 AM
It's not parts of the processor.. it's the seperate cores that can be turned on/off. Each core is a CPU. Albeit that that all the cores are placed on one chip. Dual Core or Quad Core just means it's 2 or 4 processors on a chip ;)
I think we'll eventually get to the stage where we call multiple cores on a single chip 'a CPU' - especially if the processor in that article becomes common (80 cores!). A multiple-core CPU is generally treated as a single CPU for licencing purposes (as far as I am aware)...
Anyway...Stop dragging this thread off-topic!
Back to water...Has anybody else got any good water renders?
Quote from: Dark Fire on February 13, 2007, 10:31:06 AM
Quote from: 3DGuy on February 13, 2007, 08:10:01 AM
It's not parts of the processor.. it's the seperate cores that can be turned on/off. Each core is a CPU. Albeit that that all the cores are placed on one chip. Dual Core or Quad Core just means it's 2 or 4 processors on a chip ;)
I think we'll eventually get to the stage where we call multiple cores on a single chip 'a CPU' - especially if the processor in that article becomes common (80 cores!). A multiple-core CPU is generally treated as a single CPU for licencing purposes (as far as I am aware)...
Anyway...Stop dragging this thread off-topic!
Back to water...Has anybody else got any good water renders?
Can we talk about "Water Cooled" Quad-Core CPU's? *Pulling Dark Fires leg* LOL
Hell I can't even get water to render. Sorry
Quote from: buzzzzz on February 13, 2007, 12:05:06 PM
Can we talk about "Water Cooled" Quad-Core CPU's? *Pulling Dark Fires leg* LOL
Hehe...Very clever...That could actually be on-topic as the topic title is quite ambiguous. "Water Render times" could quite easily expand to 'Render times with a water-cooled CPU'.
I hope that 80-core processor doesn't go the way of the Cell processor. I have trouble seeing how that could be efficient to make. Think of how much this multiplies the chances for something to go wrong as well. I understand the idea behind the chip was to make a very power efficient processor, but I would like to see how much power they can pack into it.
Quote from: buzzzzz on February 13, 2007, 12:05:06 PM
Quote from: Dark Fire on February 13, 2007, 10:31:06 AM
Quote from: 3DGuy on February 13, 2007, 08:10:01 AM
It's not parts of the processor.. it's the seperate cores that can be turned on/off. Each core is a CPU. Albeit that that all the cores are placed on one chip. Dual Core or Quad Core just means it's 2 or 4 processors on a chip ;)
I think we'll eventually get to the stage where we call multiple cores on a single chip 'a CPU' - especially if the processor in that article becomes common (80 cores!). A multiple-core CPU is generally treated as a single CPU for licencing purposes (as far as I am aware)...
Anyway...Stop dragging this thread off-topic!
Back to water...Has anybody else got any good water renders?
Can we talk about "Water Cooled" Quad-Core CPU's? *Pulling Dark Fires leg* LOL
Hell I can't even get water to render. Sorry
So Jay...you can't get ANY water to render?
No John, not since that first scene I posted with water. No mater what I do as soon as it gets to the water, Poof TG just leaves town. Hoping that the update will fix this from me and others??? I'm running 3 Ghz P4 and 3Gig of Ram so I don't know? Everything else is fine.
maybe terragen needs to be reinstalled?
Quote from: king_tiger_666 on February 14, 2007, 11:06:11 PM
maybe terragen needs to be reinstalled?
That's a thought, thanks.
Yeah, that's weird Jay....Did my one and only water render on a crappy 2.2GHz with 512ram. Took ages, but it did work.
Maybe a TG re-install might be the answer.....