I notice that TG is getting very sluggish after a few hours of working on one file, and doing testrenders. Refreshing of everything takes longer, also loading a new file. But also explorer gets sluggish (with TG still open). Could it be that TG 'retains memory' or so?
Usually a reboot helps to speed everything up again, but would there be a faster way?
Quote from: Dune on November 05, 2016, 11:37:50 AM
I notice that TG is getting very sluggish after a few hours of working on one file, and doing testrenders. Refreshing of everything takes longer, also loading a new file. But also explorer gets sluggish (with TG still open). Could it be that TG 'retains memory' or so?
Usually a reboot helps to speed everything up again, but would there be a faster way?
Been finding that doing torture tests on my new one as well. It had been happening on the old dell but I wrote it off to the computer not the program...Restarting the computer helped more than restarting TG4
Yes, though I just restarted TG and it was fast again.
Quote from: Dune on November 05, 2016, 11:53:00 AM
Yes, though I just restarted TG and it was fast again.
That suffices on the Corsair but the dell needs a complete restart. Must be the age...and as an aside the dell has win 10 but the corsair only win 7
I've have experienced this behavior as well. On a fairly complex scene over several iterations the length of similar test renders grew substantially longer. Also, the viewport became increasingly sluggish. After a reboot the program was noticeably quicker. My current practice involves a reboot about every half dozen iterations. Unfortunately between reloading the scene and subsequent population updates this has become a rather lengthy work-around. :(
Good to hear I'm not the only one :) Perhaps a 'flush memory' button would be good (if that could work)...
Quote from: Dune on November 06, 2016, 02:45:17 AM
Good to hear I'm not the only one :) Perhaps a 'flush memory' button would be good (if that could work)...
Hmmm, maybe that has impacted the way my new box runs...I'll get it back Tuesday and find out then if there's any difference.
Thought it was just me having issues................been trying to figure out what was wrong with my pooter. A reboot rather than a restart is the only thing that helps mine. The longer I'm in TG the slower it gets. ???
I guess in my non-techie jargon this is what I was trying to ask about the other day Ulco. I had noticed a slow down and saw 'population cache' somewhere.
I know a web browser used to load up and store items that you had to clean stuff out of now and then.....
I wondered if TG handled objects the same way or each new render it just keeps adding more and more to TG internal memory, if there is such a thing.
I don't need a detailed answer, I know someone will figure this out. :)
Quote from: Dune on November 06, 2016, 02:45:17 AM
Good to hear I'm not the only one :) Perhaps a 'flush memory' button would be good (if that could work)...
Did u guys tried the Purge Undo under the project menu? Might help.
Does Windows Task Manager show high memory usage in these situations?
Are you all using the new clouds (Easy Cloud or Cloud Layer V3), or do you notice this slowdown even when you have not been using them?
Are you all using the RTP, or do you notice the slowdown even if you haven't been using RTP?
Matt
with my ancient video card I rarely use RTP but it happens in any piece I work on for any length of time. both on this computer which I reinstalled TG4 and the new one that's gone back to the shop for tweaking. I'll keep Task Manager open next session and look at that.
Just so you know Bobby, the RTP actually taxes the graphics card *less* than the regular preview. ;) It does, however, use more memory, so if memory is also an issue on your old machine, it would be best minimized too.
- Oshyan
On the render I'm doing right now............not using RTP. Just using 8 cores. Memory usage is at 29% with a very simple scene................no populations, two objects (6 MB, 161 MB), one "easy" cloud layer. If you need any more info let me know.
What is your total amount of memory, so we know what "29%" actually means? Also, is it behaving "sluggish" right now, as described earlier in this thread?
- Oshyan
Sorry, thought I had put that in there, 64GB. ;)
Woah, so nearly 19GB. Well, part of that may be due to the recent update changes to the subdiv cache size, and part is almost certainly due to Easy Cloud and the cloud v3 scattering model it uses, as well as part coming from Embree. Basically, TG4 uses more memory than TG3, to do things better and faster. :D But we're looking at how optimize.
Our greater concern for this thread is this feeling of "sluggishness", and determining where it may be coming from. In the case of this most recent render I'm still unclear whether you are simply reporting memory use (helpful but not quite on the original topic of the thread), or if you are also experiencing sluggishness and reporting memory as a possible link. If you do see sluggishness now, it would not seem due to memory use since you have so much free...
- Oshyan
Like I said, I'm not using RTP. But if I change the seed on my "easy" cloud, it takes almost 20 seconds before the preview updates. And while it is waiting to update you basically can't do anything else, no mouse clicks, no nothing.
OK, that's good info, thank you. Can I ask what your cloud Quality setting is? And number of Voxels?
- Oshyan
Hey Matt, using TG 40010 with V2 clouds the 'sluggish' behavior was very noticeable so I did check the memory usage and was surprised that it wasn't ramping up, and still had a fair margin of available RAM. I didn't take notes, but I still have the scene in progress and will track the numbers. As my scenes are pretty loaded with high poly models I had assumed that perhaps I had hit the 'ceiling' with my current way of building TG environments. In particular: the viewport even with many objects and pops 'hidden' became progressively slower, the crop renders not only took longer, but additionally the app after render became slower to finish and release the image for saving. I was going to try the scene in the 3.4 vers to see if it was my scene construction or a change in the program, as well as the new vers. 4004. I'll still do that as it appears that others have something similar happening. The V3 clouds are another conversation. I did have a go at some V3 clouds in the current scene and the render time hit is pretty severe (even for me :-[). The RTP seemed to do pretty well, and I didn't really notice any progressive slow-down.
Quality: 1.00053
Voxels: 49.85 million
It is the file Luc uploaded in this thread:
http://www.planetside.co.uk/forums/index.php/topic,22370.0.html (http://www.planetside.co.uk/forums/index.php/topic,22370.0.html)
Yossam, I believe the Luc Bianco file has 500 million voxels (499.85). That's a pretty robust setting.
I stand corrected...............old, tired, need new glasses eyeballs. Sorry............ :P
Mm, yeah, 500m voxels is going to be taxing, especially on memory use. Lots of nice detail though. :D
Zaxxon, very interesting that you're seeing this even with v2 clouds and that memory use does not seem to be an issue at first glance. Please do check it in 3.4 as that would definitely be an important comparison. In addition, are your v2 cloud layers localized and using voxel acceleration? Were you using RTP prior to or during the experience of sluggishness?
- Oshyan
Ah, this is going to be interesting... I will keep your questions in mind, Matt, and check things out while working. I didn't have much memory load by objects, just a few small pops, did have a V3 cloud in there, and used RTP ocassionally (not all the time). Perhaps hitting RTP on and off also adds to memory, if it keeps every state without auto-flush?
Just for the sake of it, I tried 500 million voxels and Terragen crashed. After that I put it back to 50 million voxels. It was worth trying at least.
AP, how much memory do you have? Did you watch memory use while attempting this?
- Oshyan
Quote from: Oshyan on November 07, 2016, 01:22:44 PM
AP, how much memory do you have? Did you watch memory use while attempting this?
- Oshyan
Only 4 gigabytes, however I can try it again and look at the memory reading.
The memory is at 1.5 gigabytes before opening Terragen and when Terragen is running it jumps to 3.63 gigabytes. That would probably explain the crash.
Yeah, 4GB is the minimum for good reason. I'm afraid you don't have a lot to work with when it comes to using Terragen 4.
- Oshyan
Perhaps in some cases, yes. I just adapt to what I know I can do and what my computer is capable of.
Quote from: AP on November 07, 2016, 10:08:26 PM
Perhaps in some cases, yes. I just adapt to what I know I can do and what my computer is capable of.
And some fairly impressive output if I do say. Probably a faster machine than this old dell though.
I have a Intel Core i5 1.7 GHz. I am wanting to upgrade some day, all in due time.
Quote from: AP on November 07, 2016, 10:50:59 PM
I have a Intel Core i5 1.7 GHz. I am wanting to upgrade some day, all in due time.
this dell is a duo core at 3.0Ghz but task master never reports it reaching that speed.
Working hours on one file without using RTP or V3 clouds shows no sluggishness, so it's either. I'll see if I can work with one of those used and see what happens.....
Here are two separate renders, that using the same settings have a pretty large render time difference. The top one was an area crop render and took 7 hours 18 minutes, the bottom one was cut out of the completely rendered image which took 5 hours 7 minutes in total. 900X1600, detail 1, AA 8. The top image was rendered after a number of scene changes and RTP viewings, the bottom was rendered after a reboot and reload. Additionally the top image has only two populations and no hero objects, whereas the bottom image is fully populated with numerous hero objects. So that's a obviously significant variance. The base scene takes 24 gigs to load; with the top image at render end reporting 63.1 gigs used (out of 64 gigs installed), and the bottom image topped out at 58.4 gigs used. For now my process is to reboot and reload every so often, but especially to render out full res images. I still have to retest the V2 performance, but despite the pain I opted for the V3 clouds (they're just prettier!).
Interesting results Zaxxon (and gorgeous clouds!). I would suspect that with that much memory used for the top crop, it may well have been using your disk-based swap space for additional memory, which would slow down rendering dramatically. It could be due to the RTP not releasing memory after the previous use, though that's pure speculation on my part. But given the clear difference in reported memory use you show there, it does seem as if there's some kind of memory-related issue going on.
Thanks for those details. Hope to see the full and final image soon, it looks very promising!
- Oshyan
Quote from: zaxxon on November 08, 2016, 04:45:45 PM
I still have to retest the V2 performance, but despite the pain I opted for the V3 clouds (they're just prettier!).
I have to agree, and I just flashed how you keep the same shapes...they both use the same Density shader, D'oh....
Wonderful sky, Doug! Looking forward to a big one too.
It indeed looks like TG doesn't flush memory before rendering, or something in that sense. The second one should have taken longer with all the veggies.
Quote from: zaxxon on November 08, 2016, 04:45:45 PM
Here are two separate renders, that using the same settings have a pretty large render time difference. The top one was an area crop render and took 7 hours 18 minutes, the bottom one was cut out of the completely rendered image which took 5 hours 7 minutes in total. 900X1600, detail 1, AA 8. The top image was rendered after a number of scene changes and RTP viewings, the bottom was rendered after a reboot and reload. Additionally the top image has only two populations and no hero objects, whereas the bottom image is fully populated with numerous hero objects. So that's a obviously significant variance. The base scene takes 24 gigs to load; with the top image at render end reporting 63.1 gigs used (out of 64 gigs installed), and the bottom image topped out at 58.4 gigs used. For now my process is to reboot and reload every so often, but especially to render out full res images. I still have to retest the V2 performance, but despite the pain I opted for the V3 clouds (they're just prettier!).
Are the clouds default settings Easy clouds or Cloud layer v3 clouds?
AP: the clouds are 'Easy Clouds' and no, these are not the default settings. When I post the final image I'll put the settings up in the forum.
Dune: Exactly. I'm certainly not a programmer, so I'll not hazard a quess here. And as Oshyan pointed out when the ram is near the limit performance will suffer. Looks like it's time to add more ram at any rate.
Is this being investigated as a possible memory leak?
It is being investigated as a potential bug. A memory leak is just one possible element.
- Oshyan
Just seen this, I too have experienced this sluggishness. When it happens I simply save and reboot TG.
A couple of new observations after some more work on the scene referenced above. My latest test render of the entire image took 4 hours 56 minutes, and ended with 46.6 gigs in use. This was rendered without a reboot/reload. The scene had numerous object tweaks and small to medium crop renders and spans about 6 saved revisions, all this with the Easy Cloud voxel count reduced to 10 million (500 million setting used in final render), and RTP not being used (as I had already finalized the cloud shapes and placement). The key factor seems to be the RTP use. As I recall, while using the V2 clouds in prior iterations of this scene, I did use the RTP to preview those clouds. Indeed, the RTP is an amazing pre-viz for clouds and atmosphere. Also, in the current work, I didn't notice any progressively 'sluggish' viewport performance.
Excellent, thank you for the continued and in-depth feedback zaxxon! Super helpful.
- Oshyan