I don't know if it's just me or not, but my renders seem to take a long time. My average render for a simple scene is about 2-5 hrs, depending on render settings. Here's an example of one of my render's.
It's done completely in Terragen2, here are my render settings:
resolution: 800*600
detail: .5
Anti-aliasing: 3
GI relative detail: 2
GI sample quality: 2
Here's my computer's hardware (it's old):
cpu: Athlon chip over-clocked to 2.11 GHz
memory: 768 MB RAM, but really old (either DDR1 or before, I can't remember)
Video card: Don't know if this matters, but NVIDIA GeForce4 Ti 4200 with AGP8X
it really depends on the quality settings, especially when you got several cloud layers i think.
right now i'm rendering at same resolution and .5 quality and GI 1/1, 50% rendered after 4 hours, 3 cloud layers at quality 1, that's roughly 400 samples for the closest-to-cam-layer.
my specs are
amd64 sempron 3000 @ 1.8Ghz (no o.c.)
2 gig ddr 400 samsung ram
I'm guessing its your hardware, your render settings sem fairly standard. Follow this link http://tg2bench.kk3d.de/index.htm (http://tg2bench.kk3d.de/index.htm) for a benchmark test, that will let you see if its your system or not.
Happy Rendering
Richard
GI = 2 = maxed out render time.
Usually GI > 1 means long renders..but high quality...
yeah run that benchmark and it should give you a rough estimate of how long your renders should take .. the list has gotten quite extensive too ;D ;D ;D
It's interesting to see how people with the same systems can have such a big difference in times. For instance I have an Intel Core 2 Duo E6600 running at 2.4ghz with 2GB of RAM. So I took a look at the table to get an idea of where my system would fit in. Sure enough there I see 4th from top a guy (Christian Petrik) with the same processor, same amount of ram, same type of ram and same chipset as me - time to render 0:03:34. So to compare I download the test file and open in terragen, close all other applications, open task manager and close all non essential processes and render the image. Time to render 0:04:58!! How can two systems differ so much? I thought there would a couple of seconds difference between 2 systems of the same spec. Anyway I went into my BIOS and overclocked my system to 3Ghz so now my system has the same chipset and amount/type of memory but is running 700mhz faster. I render again and this time get a time of 0:04:24. So even overclocked that time would put me down with the Athlon 64 systems. I guess there must be some setting in my BIOS or windows that's really clogging my system up, although I'm not quite sure what could be making such a big impact, especially with all the processes shut down etc.
So just to summarise, that chart doesn't necessarily give an accurate picture of what speeds you can expect to get as shown above.
I agree. There needs to be a better benchmark testing against what matters to the core program, which I thought someone said has to do with the CPU core and maybe the amount of RAM.
Heh, look at the list, there's a guy with 6 Gigs of RAM and Quad core :P
Somehow, he/she has a Mac..interesting lol
Alright, here are the results of my render: 5:42
It's slightly slower than faster than this:
Intel Pentium D 840, 2 cores, 3200 MHz, 2 GB Dual DDR2, Nvidia nForce4 SLI, Windows XP Pro SP2, 0:05:47
but slower than this:
AMD Athlon XP 3000+ (Barton), 1 core, 2167 MHz, 1024 MB DDR, Nvidia nForce2 Ultra 400, Windows XP SP2 , 0:05:36
Not bad for such an old system! So, it's not my computer that's slow (at least, not entirely I think). I think it may be just the scene.
Oh, I ran CPU-Z and here are the specifications:
CPU:
(http://forums.planetside.co.uk/index.php?action=dlattach;topic=1079.0;attach=2650;image)
Motherboard:
(http://forums.planetside.co.uk/index.php?action=dlattach;topic=1079.0;attach=2654;image)
Memory:
(http://forums.planetside.co.uk/index.php?action=dlattach;topic=1079.0;attach=2652;image)
Cache:
(http://forums.planetside.co.uk/index.php?action=dlattach;topic=1079.0;attach=2648;image)
it's your ram, only sd, the other guy got ddr
Hmm that scene with those settings should render in a few minutes. Did you turn on raytrace shadows? If so, that's probably the culprit, turn it off. I don't see a reason for raytraced shadows in this image.
Quote from: DiscoBall on March 18, 2007, 12:22:22 AM
Heh, look at the list, there's a guy with 6 Gigs of RAM and Quad core :P
Somehow, he/she has a Mac..interesting lol
That's me :D
It's not a Mac, it's a Dell. Got it in November as I'm about to graduate and intend on doing visual effects work.
Edit: I didn't notice the mac one previously... mine is the other 6gb ram quad.... the 1866 MHz one
To check the render times of the new release(1.8.76.0) vs. the previous version(1.8.64.0) I used the bench test( http://tg2bench.kk3d.de/index.htm ) file in both versions. To my shock the new version was 63% slower than the previous version.
My computer is as follows: AMD 64 4200+, 2.2 GHZ, dual core; 2 GB DDR-400 ram; XP+SP2.
The bench time for the previous version was 4:33, similar to the times for this computer configuration in the bench test table. But when I ran the bench file with the new version( TG2 deep+animation 1.8.76.0) the time was 7:25. That's 63% increase in time. I ran the tests three times and got the same results.
Both versions were licensed TG2 deep-animation.
This can't be possible. Does anyone have an explanation? Has anyone run this simple test?
i will try this test and see if the newer version changes my render time.... :( if it ends up being longer than it previously was
@digidon.....I hope that is not the case with the new version. I have not yet downloaded it.
Can others here confirm this?
I have both versions installed seperately and I tried the benchmark on my system and there was no appreacianle differance, might have been a couple of sec faster. I do think we might need a differant benchmark however. I'm not sure how these things work but I would assume that the calculations for haze, reflection and populations for example use differant approaches and therefore differant systems might score differantly ie one king of architecture would be quicker at working out reflections that another of equal "power". Idealy a benchmark should have examples of all types of processing chalenges in it.
Richard
For what it is. all things being equal the benchmark does give a measure of how fast your pc is at doing a render. Though I would agree, a benchmark that has water "reflectivity"? /3d clouds "with ray tracing"?, maybe a population of grass/ 1 tree or more would test more areas of terragen.
Confirmed on my machine. Rendertime increase of about 67% :-\ From 232s to 389s
Thanx 3Dguy!
It is certainly not the direction I had hoped things would go.
I will not waste my time with the update then.
i just ran the benchmark on the new version and previously had a render time of 6mins 06seconds now my render times 10mins 47seconds:( so much of optimisation?
i have P4 2.8ghz northwood so it looks like it affects both intel and amd
Yeah seems bad. It maybe AMD specific. I don't know what CPU cyphyr is running, but he says he sees no difference. So if that's an Intel that might be an indication.
I'm going to try one of my own scene's tomorrow, one that requires a bit more time and see if it's exponentional or just a fluke in this particular scene.
I'm really sorry to say this....but if what you guys are experiencing is truly an indication of update render times....I think that is really poor work from Planetside.....especially after all the threads concerning excessively long render times in the original version.
I didn't notice anything concerning render time optimization in the docs...so, I didn't really expect any...but to see it go 60-70% higher is, to put it very mildly....a dissapointment.
I'll reserve final judgement until more folks have compared....
I tried another landscape comparison between the two TG2's (1.8.64.0 vs 1.8.76.0) and saw a 28% slower render time by .76.0. Of course there will be diferent degrees of difference in relation to the number and type of parameters chosen. But, more files should be used in the comparison of the two versions. And please mention which processor, or better still, complete computer configuration.
I've kept a record of all my render times for tg2 renders so i may go back and revisit some... though if it is a indication then my clouds and Atlantis renders will be taking over 106hrs to about 115hrs respectively:(
Hmm, I think I have the older build of TG2 (got it ~ the day it came out and haven't updated yet). Yes, I know the RAM is causing it to be slow, but to change that I would have to get a new motherboard (it's ancient!), new video card that supports PCI, etc.
From what I can tell by the other render times, it's on the slightly slower end (42 faster, 29 slower).
Also, the render was a fairly decent time, so I'll try adjusting some settings on my scene and see if it improves the render time.
Although RAM will have some impact on render speed, generally speaking faster RAM isn't going to make a very appreciable difference in render times. Perhaps a few seconds in a 3-5 minute render, but not much more. Amounts of memory below 1GB would have a greater effect, as of course would the CPU.
Setting up benchmark scenes is a tricky thing. Ideally you want as many people as possible to run it to get the widest basis for comparison, but many people are impatient, so a quick benchmark is desirable for that reason. On the other hand the faster the benchmark the greater the margin for error, especially when time is being measured in seconds. To illustrate this, think about the difference between a 50 second render and a 100 second render - the 2nd is twice as long. Now let's imagine the impact of starting another program, even a simple document editor or email program, while the render is happening. Let's say it slows down the render for about 10 seconds as the application loads. That's going to have an impact of as much as 10% on overall render time! But now imagine the imapact of that 10 second slowdown on a 1000 second render - it's only a maximum of 1%. So the longer your render, the more such small and usually inevitable CPU spikes are minimized in terms of their effect on the overall render time. In this particular case the render times are fairly short - I would have aimed for something more like 20-30 minutes - so accuracy may be suffering as a result of that. There are also often processes that are difficult or impossible to truly get rid of before rendering - these would include antivirus scanners and other security software. Depending on the configuration of your security systems this may have a significant impact on render time. Finally remember that the results here are reported by individuals and subject to human error and possible embellishment. It's useful to have a wide variety of results from similar machines to get an average, in case someone is misreporting for some reason. Never assume an unusual result is simply someone who knows how to optimize their system; it's more likely an error.
As far as the increased render times, although some changes to the rendering engine were made in this update, you should not be seeing significantly increased render times in most cases. As noted in the change log there were some accuracy improvements in cloud rendering that may result in slightly slower render times, but the numbers being reported here are unexpected. We will certainly look into these potential issues. It may be useful if people continue to report their results here in this thread so we can get a better idea of whether this is affecting everyone the same. We've already seen at least one person saying they can't duplicate the issue.
- Oshyan
Quote from: Allegro on March 18, 2007, 06:00:46 PM
Quote from: DiscoBall on March 18, 2007, 12:22:22 AM
Heh, look at the list, there's a guy with 6 Gigs of RAM and Quad core :P
Somehow, he/she has a Mac..interesting lol
That's me :D
It's not a Mac, it's a Dell. Got it in November as I'm about to graduate and intend on doing visual effects work.
Edit: I didn't notice the mac one previously... mine is the other 6gb ram quad.... the 1866 MHz one
Heh lol, I wonder how you fit so much? :P You only have 4 slots...right? :P
4 slots..did you get like 2 2 GIG sticks of ram and 2 1 GIG sticks of ram...fits out to 6 gbs of ram :P
Nice...but do the extra 4 do any extra good? lol
OK I gave the benchmark a re test and I guess I must have remembered wrong, big differance first render on TG2 1.8.64.0 came in at 5.31 min and the seccond on the new update, TG2 1.8.76.0 came in at 8.45min. Thats nearly 60% slower (I think, pardon my math) I think I stick with the first release for now. Also comparing the two images there is a definate differance in the clouds, The new release (lower pic) has more defination/contrast in the clouds.
Richard
the never dells support up to 8gb so its not hard to see how one can have 6gb installed and it probably helps if they are into multimedia design... or the like...
Here's mine:-
amd athlon 64 3800+ @2400
via K8T800 Pro chipset
2Gb PC3200 DDR RAM
Quadro 4 900xgl w/ driver 9.1.3.6
Win XP Pro (fully updated)
Classic Theme
With each version I ran the benchmark with as much shut down from the Task Manager as I could.
TG2 1.8.64.0 - fastest time 4:04, slowest time 4:11
TG2 1.8.76.0 - fastest time 6:30, slowest time 7:02
Edit: - Deep + Anim
I just tested it with one of my own scenes. Difference was less dramatic, but still the new version was 35% slower (24 minutes vs 32.5 minutes)
Thanks for the continued reports everyone. We have identified an issue whereby the acceleration cache for the clouds will have a minimal or at least significantly lessened effect due to the implemented fixes for the "blockiness" issue. The effect of this will vary from scene to scene and it will not always be so dramatic as in this benchmark scene.
We will work on a better balance for the acceleration cache and blockiness fixes but we may not be able to release an interim update to resolve it at this time. It will certainly be addressed and better optimized in a future update though.
Unfortunately issues like this will be a part of any ongoing public test such as the Technology Preview release. We do apologize for any inconvenience or frustration this causes.
- Oshyan
I'm totally cool with this and I think its a great way for a company to test a product like terragen. Because of its nodal system there are SO many differant ways of rigging things up it would be next to imposible to do this level of testing in-house. I'll try a couple of renders without clouds in them and see how they compare.
Cheers
Richard
I re rendered my scene, taking out a cloud layer, reducing atmosphere quality, etc. and amazingly render time improved from 3-5 hours to 5-10 minutes! However, I didn't test each setting one at a time so I'm not sure exactly what was taking all that time.
Made another test, this time disabling the "Cumulus Layer 1" (enable, enable primary and enable seccondary all un-ticked) and the render times were significantly improoved. Build 1.8.64.0 rendered the scene out in 2.24min and Build 1.8.76.0 rendered out in 3.10min. For referance my system is a Rock Extreme Ti 3.4ghz with 1Gig ram
Richard
Quote from: king_tiger_666 on March 18, 2007, 09:19:56 PM
i just ran the benchmark on the new version and previously had a render time of 6mins 06seconds now my render times 10mins 47seconds:( so much of optimisation?
i have P4 2.8ghz northwood so it looks like it affects both intel and amd
I've the same exact situation on my own - same processor, same render times (6:38 on 1.8.64.0 and 10:55 on 1.8.72.0)
Oshyan;
I must say I am flabbergasted that Planetside would release the update without testing the render time impact of the changes you describe.
From the posts so far, it appears that render times have increased at least 30%....possibly up to 70% depending upon the scene. I don't think that constitutes "slightly slower render times" or a "minimal or at least significantly lessened effect".
I'm not impressed!
@cyphyr...your numbers also confirm increased render times for the new build.
@caramel...you need to compare build 1.8.64 to build 1.8.76...otherwise your numbers are meaningless.
Quote from: DiscoBall on March 19, 2007, 06:08:06 AM
Quote from: Allegro on March 18, 2007, 06:00:46 PM
Quote from: DiscoBall on March 18, 2007, 12:22:22 AM
Heh, look at the list, there's a guy with 6 Gigs of RAM and Quad core :P
Somehow, he/she has a Mac..interesting lol
That's me :D
It's not a Mac, it's a Dell. Got it in November as I'm about to graduate and intend on doing visual effects work.
Edit: I didn't notice the mac one previously... mine is the other 6gb ram quad.... the 1866 MHz one
Heh lol, I wonder how you fit so much? :P You only have 4 slots...right? :P
4 slots..did you get like 2 2 GIG sticks of ram and 2 1 GIG sticks of ram...fits out to 6 gbs of ram :P
Nice...but do the extra 4 do any extra good? lol
Well, I've got a 64 bit machine... so my motherboard will allow me to have 64gb of ram. 8 Slots, that I can stick 8gb in each slot. Right now I've got 6 of the dimms filled with 1gb each. With a 64 bit machine, having 4 gb of ram is pretty much the same as having 2gb on a 32 bit machine... but after that it's the icing. I can work with Photoshop files that are more than a gig in size. I can have Maya, Photoshop, After Effects, Premiere, & Itunes running all at the same time, with no slowdown. I rendered out a skybox for a scene last week. I had Terragen rendering three frames at once, and was still able to use my 4th processor to do uvmapping in Maya.
It's an advantage for sure, and it will become more of an advantage as companies gear their programs for 64bit machines.
Since posting the render time problem with 1.8.76.0 in reply#12 of this thread I have noticed another difference between 1.8.76.0 and 1.8.64.0. In .64 if you open the program and render a scene, then render it again and again( without changing any settings) there will only be slight differences between the render times. But if you do the same with .76 there are larger differences between the render times. The first render after opening the program will be the shortest time. After that they increase at amounts significantly greater than .64.
This may be just another manifestation of the accelerated cache and blockiness fixes mentioned by Oshyan in Reply #31 of this thread.
I recently made a 3000 image movie using .64. The increase in .76's render times could seriously impact such a project( depending on the degree of atmosphere). However, I think I will use .76 for single images since there are improvements in many places.
DigiDon
Is there a change log anywhere so we can actually tell what changes they made?
The changes of .76 over .64 can be found by going to "Announcements" at the top of the forum then clicking on "New version: ........"
I think if anyone has a problem with a whether it's faster or slower at this point isn't "playing fair". This is a PREVIEW or isn't it? And to expect things to be built solidly and then later to be built faster is normal. That's how I understand programming.
Ever read any of John Carmack's logs (if you can fish through the genius intricacies of his mind)? I've read on more occasion where speed just wasn't an issue until the very end. Same with the 3DRealm's logs about what they're programming. Making such an issue about render times might be helpful, but I'm more interested in features - like transparency, opacity, shaders, functions. I'm 100% confident that Jo and Matt will make this baby hum once all of these things are working.
So, Oshyan, I for one am willing to wait and be patient. Disappointment will come and go until it is finally what I expect it to be - excellent.
Quote from: Oshyan on March 19, 2007, 02:29:41 PM
Thanks for the continued reports everyone. We have identified an issue whereby the acceleration cache for the clouds will have a minimal or at least significantly lessened effect due to the implemented fixes for the "blockiness" issue. The effect of this will vary from scene to scene and it will not always be so dramatic as in this benchmark scene.
We will work on a better balance for the acceleration cache and blockiness fixes but we may not be able to release an interim update to resolve it at this time. It will certainly be addressed and better optimized in a future update though.
Unfortunately issues like this will be a part of any ongoing public test such as the Technology Preview release. We do apologize for any inconvenience or frustration this causes.
- Oshyan
Gradient, some speed tests were done, but they didn't generally indicate such a dramatic performance loss. I will agree that not testing on the currently available benchmark scene by Karsten was an oversight and I will be sure to include it in future tests of alpha builds for comparison.
That being said please do keep in mind that this is a pre-release Technology Preview. Our current focus is on feature implementation and bug fixing. The initial Technology Preview release had many well-known problems, some of them fairly serious (frequent crashes, etc.) so fixing those issues was of greatest importance for us on this update release. Speed improvements are always desirable and will be implemented as time allows, but hopefully you'll agree that actually being able to work in the application without worrying about crashes is a bit more important than rendering as fast as possible.
You can still look forward to very significant optimization in the future. Perhaps it will not be enough for you, but only time will tell. In the meantime if the ups and downs of these update releases - the 2 steps forward and sometimes 1 step back - bothers you then I would simply suggest taking a break from TG for a while and waiting until the final release. You have to accept a certain amount of frustration and "wasted" time in dealing with any pre-release and its inevitable shortcomings. If you're not prepared to do that, there is little use in following the every development of the program and being upset when it doesn't go in the direction you want it to.
I do hope you'll enjoy and be able to comfortably use the final release but I don't think the Technology Release period will be of use to you given what you've said so far. We certainly didn't intend it to be used by everyone; if we thought it was ready for that we would have released it in final, or at least as a beta! ;D
- Oshyan
Quote from: Allegro on March 19, 2007, 08:45:16 PM
Quote from: DiscoBall on March 19, 2007, 06:08:06 AM
Quote from: Allegro on March 18, 2007, 06:00:46 PM
Quote from: DiscoBall on March 18, 2007, 12:22:22 AM
Heh, look at the list, there's a guy with 6 Gigs of RAM and Quad core :P
Somehow, he/she has a Mac..interesting lol
That's me :D
It's not a Mac, it's a Dell. Got it in November as I'm about to graduate and intend on doing visual effects work.
Edit: I didn't notice the mac one previously... mine is the other 6gb ram quad.... the 1866 MHz one
Heh lol, I wonder how you fit so much? :P You only have 4 slots...right? :P
4 slots..did you get like 2 2 GIG sticks of ram and 2 1 GIG sticks of ram...fits out to 6 gbs of ram :P
Nice...but do the extra 4 do any extra good? lol
Well, I've got a 64 bit machine... so my motherboard will allow me to have 64gb of ram. 8 Slots, that I can stick 8gb in each slot. Right now I've got 6 of the dimms filled with 1gb each. With a 64 bit machine, having 4 gb of ram is pretty much the same as having 2gb on a 32 bit machine... but after that it's the icing. I can work with Photoshop files that are more than a gig in size. I can have Maya, Photoshop, After Effects, Premiere, & Itunes running all at the same time, with no slowdown. I rendered out a skybox for a scene last week. I had Terragen rendering three frames at once, and was still able to use my 4th processor to do uvmapping in Maya.
It's an advantage for sure, and it will become more of an advantage as companies gear their programs for 64bit machines.
Lol restricting Terrgaen to 1 core each, eh? :P Wish I could do that...man I seriously need to upgrade my CPU :P lol
Btw, did you have THREE Terragen TGD's open at once, and if not, how'd you manage to render 3 at a time with 1 TG2TP program? :S
And nicely said Oshyan, the final release should suit everyone :) Soon..enough :P
probably three instances of terragen running at once
@Oshyan;
I had indicated in another thread earlier on...that I had already taken a break from TG2. In fact it has been almost 2 months since my last render.
I do however take the time to follow progress on this board....when I saw that an update was available, you will note that one of my first questions was whether or not render times changed. Although I wasn't expecting any optimization yet...I certainly didn't expect things to go the other way...at least by the amounts folks here have indicated.
In my testing of the original T2TP version, the crashes and bugs were far less of a problem to me than the render speed.
You said;
"there is little use in following the every development of the program and being upset when it doesn't go in the direction you want it to."
Well, it seems that your statement should not be focussed at me alone...as this thread shows, there are many more sophisticated testers who seem to be "surprised" at the increased render times. If you lose the backing of those folks...that would not be good....
As for the final version being "for me", yes I understand that if I want to use the program it will require hardware upgrades on my part. That is something I will need to wrestle with.
In the meantime, I still plan to follow progress here....I hope you do not mind.
Gradient, I certainly don't mind you following the developments. In fact we are glad to have people so interested in the progress of the product. I only hope people - many besides yourself included, of course - will keep in mind the in-development nature of the product and that some "backward steps" are going to happen. What matters of course is the final result. Only time will tell us how well it turns out. ;D
- Oshyan