Old Stuff

[rcallicotte]

@efflux - I understand about software.  Seems like much of it (not centering on any one app) is just so loosely thrown together and lied about it through advertising.  This, in fact, was one of the things that attracted me to Terragen - the general support.

[efflux]

TG2 is a cool app. I've absolutely no real criticism of it at all. Planetside are doing a great job. The long render times are just how it is with landscape apps.

I've got very frustrated lately with computers. It's because I've been using Linux and free software which is in many cases streets ahead then for example, I go on the Modo forum where they are bashing Blender's UI. Blender's UI is great because they ignored conventions and made it better than other apps. They should take a closer look at Modo's UI first. It is a good app but overhyped. As for many of the other so called "industry standard" apps, I couldn't even bear to use them. I couldn't even install XSI on Linux it was such a licensing nightmare and I didn't even get any response from them about the issue. These are companies who want you to part with £1000s. They are in a cuckoo world. The proof is going to be in the pudding. You'll see this especially with Blender in the near future. I'm experimenting with some Linux renderers here. For example a new GPL one called Luxrender because the Blender render is a bit too simple for some stuff. This is the kind of quality that is coming out of Blender - rendered with Luxrender:

http://www.luxrender.net/gallery/main.php?g2_itemId=1166

There are various new builds coming from Blender with some new stuff including volumetrics. 2.5 is going to be a whole rebuild of the app.

I've also looked into buying a Mac and it just doesn't make sense anymore. My next system will be a PC running Linux.

[efflux]

The thing that is clearly obvious with open source software is that it's an attempt to make a tool as good as possible not to sell more than your rival or add more useless features to extort money for another upgrade. They do things that proprietary companies never do. If Adobe did some of the things they do with Gimp there would be uproar and lost sales before people realized the change might have been an improvement.

Luxology screwed up by trying to compete with ZBrush except all their lost time in attempting to do this meant that the important thing - animation, has been delayed. ZBrush had the last laugh by releasing a new demo I think the day after the new Modo or something like that. It was a mockery. A total wiping of the floor with all competition.

[Matt]

Thanks for reporting those results.

As to why different scenes would make a difference, well that's because they will use (or reduce efficiency in) different parts of the renderer by different amounts. Clouds, for example, use the Cloud Shader which has its own caches and calculations to make, and it also calls on the lighting systems of the renderer in different proportions, with different parameters, and with different levels of coherence between adjacent calls which will stress various caches differently. That's just scratching the surface. Different scenes take different amounts of time to render not only because of how "much" is in them but because of what's in them and how efficiently TG can handle different scenarios.

Since we don't know the reasons why multi-threading performance seems to be worse under Wine, it's helpful to narrow down which rendering systems (or at which scenarios) lead to less efficiency - efficiency relative to different thread settings and efficiency between OSs.

Matt

[Matt]

I also asked about disabling shadows, because disabling those would avoid using the ray tracer in the final pass (unless you have reflective surfaces) among other things, and it would reveal a lot of if we knew whether that made any difference to the relative render times between OSs or between thread settings.

Matt

[Matt]

Wine:

0    0.46
1    1.53
2    0.56

That tells me there's something wrong with threading even when there's only 1 render thread. The time for 0 (which renders all buckets in the same thread as the GUI and render schedule) should be very close to the time for 1 (which renders buckets in a separate thread and only occasionally has to synchronise with the main thread). It gives me a good starting point.

Matt

[efflux]

OK thanks for the reply Matt.

I might run more tests. TG2 is the only further app Id like to get running under Wine with acceptable results.

As a matter of interest I just installed Modo on Linux Wine. it's not really usable because of a few UI issues but rendering speed is very good. In fact so good it would be worth running it just for rendering. This is the first rendering app I've installed where rendering has been equivalent To Windows.

[efflux]

Apologies for this but I must have used the wrong version of TG2 for the Linux test. Unless something else is different. I have just run a test on what is definitely the very latest 1.10.23.1. which I did have at the time of the last test. New results are as follows:

0  0.28
1  0.24
2  0.34

These are consistent though. I ran them a few times.

[efflux]

Disabling shadows causes Linux to catch up marginally against OSX to roughly one third longer time instead of twice the time or over. That is by comparing the new results from Linux. I must have started the wrong TG2 before running those first tests. There isn't any other explanation. I have tested more complex scenes but results seem fairly consistent. Clouds don't seem to make too much difference. I haven't done any new tests on Windows but I know the previous test times anyone. It would obviously be a fairer test.

[efflux]

Just some more points. The drop down menu in the node graph is now functioning correctly in Wine. In previous versions of Wine you had to use the arrow keys to navigate them. No icons are visable anywhere in the main UI but this is of little relevance once you know where stuff is. The only issue is that I now get random crashes with Wine 1.1.7. These things are problems that will sort. It's always UI. Every so often UI issues get fixed then sometimes a few issues revert back to a problem but it gets better with every Wine release. I don't think it will be long before it's 100% perfect except for the render times.

[Matt]

OK, those results look much better. In your shadow vs. non-shadow tests how many threads were rendering? Also, how are you controlling the threads? The most reliable way to force it to use a specific number of threads is to set both minimum threads and maximum threads to that number.

Matt

[Matt]

You Linux machine is P4 HT, so that only has one core? (HT can't really be treated as another first-class citizen when it comes to complex rendering). If that's true it makes sense that the render became less efficient when you tried to use 2 threads instead of 1. It's not really fair to test this against a dual-core Mac.

Matt

[efflux]

I know the Windows machine was doing 15 seconds (at all thread settings) on the default when you open TG2 so that's what I'm comparing with. It's simply that the Windows machine (same PIV specs) isn't even hooked up to a monitor or keyboard. I never use it now.

24 seconds under Wine as compared to 15 under Windows is still not too good but certainly better.

Yes my Windows machines are a single CPU with HT.

An interesting test will be when I get possibly an 8 core or whatever is available at the time.

[efflux]

Probably a quad core will be best because although 8 cores may be out by the time I build a new system, they will most likely be too expensive.

[Matt]

24 seconds under Wine as compared to 15 under Windows is still not too good but certainly better.

What about a longer render? With such a short render time it could be dominated by things that become less significant in more difficult renders. Possibly.

Matt