Recent Posts

Pages: [1] 2 3 ... 10
And too how we can see to a texture very details, when to the camera are far ? And for see of near the texture to the object ?

To see the details, you gonna need more resolution/quality when you render, while more far is the camera from the ground/object.
this is of course a very limited option, as Terragen canīt import shaders natively as "textures", if that is what you are referring.
Terragen Discussion / Re: Rock Won't Displace
« Last post by Oshyan on Today at 10:39:09 PM »
If you use the Mesh Displacer input you get correct shadows and don't have to use Force Displacement.

- Oshyan
Terragen Animation / Re: Canyon dog fight
« Last post by Oshyan on Today at 10:36:44 PM »
Super cool!

- Oshyan
Terragen Discussion / Re: How to export Terragen 4 to Maya?
« Last post by Oshyan on Today at 10:27:48 PM »
You can export terrain geometry with the Microexporter, detailed in the documentation here:
I recommend using only a single render thread, which reduces geometry overlap. But you'll still need to do some geometry cleanup in another application.

There is no built-in texture export. You can render a top-down orthographic image as a workaround.

For HDR/IBL/environment map, just make a spherical render (requires the Professional version) and save as EXR. You have the option of 16 or 32 bit per channel high dynamic range data.

- Oshyan
Terragen Support / Re: Spherical camera not rendering
« Last post by Oshyan on Today at 10:25:01 PM »
Can you send us the TGD to support AT

- Oshyan
I do not understand.

- Oshyan
Image Sharing / Re: Retreating ice, etc.
« Last post by Oshyan on Today at 10:23:45 PM »
I like this latest one quite a lot Ulco. Lots of dynamics in the character poses, good variety and realism in surface shading, and a nice scene composition with the distant glacier. It tells a lot of story.

- Oshyan
Image Sharing / Re: Good morning world...........
« Last post by Oshyan on Today at 10:20:53 PM »
Nobody in their right mind drinks a cocktail out of a coffee cup, so I'm going with sunrise here. :D

- Oshyan
Terragen Discussion / Re: The ultimate Terragen machine(s)
« Last post by Oshyan on Today at 10:14:02 PM »
Hi Niel,

First, make sure you have adjusted the graphics card driver settings of any of your machines running Quadros, as we discussed in the support thread previously. If you haven't already done that, then it's not surprising you're seeing highly variable render times. Assuming those settings are correct already and your render times are the best they can be, it would be interesting to know the specific hardware in question. But perhaps not that relevant to your question, if you're looking to upgrade.

When looking at hardware, for Terragen it is generally fairly simple to get the best performance, and then it just comes down to how much money you want to spend for each extra bit of performance. Terragen is currently a CPU-only renderer. The GPU is only used for limited OpenGL display purposes in the 3D preview and node network, basically. So for best performance you just want to get as many CPU resources as possible for your money, and then get as much RAM as you can afford, especially if you're rendering more complex scenes. I would recommend 64GB minimum for render boxes at this point.

With the Zen CPU family from AMD (including Ryzen, Threadripper, and Epyc) now being very similar to Intel's "Core" series (including Xeon's) on a clock-for-clock basis, you can now basically just use clock speed times number of cores to arrive at a roughly equivalent level of performance for a given CPU or system (in the case of multiple CPUs). The one caveat to be aware of is that the more render threads you are using, the more efficiency is lost to "overhead" (inter-thread communications, work distribution, thread scheduling, etc.). The amount of overhead varies for each renderer, and can be difficult to quantify, but hopefully it suffices to say that if you have the choice between more threads at a lower clock speed, or fewer threads at a higher clock speed and they result in the same basic number when multiplied together (e.g. 4 cores at 4Ghz = 16 vs. 8 cores at 2Ghz = 16), then you want to go with fewer threads at a higher clock speed.

So obviously there are multiple factors you have to balance - hardware cost, power usage, heat, space (do you have tons of room to put lots of computers, or do you need them all small and rack mounted, for example), etc. I'll give a couple example scenarios which will hopefully be helpful and get you a good basic idea, and then we can talk more specifics if you need to.

First, I think it's worth talking about new vs. used hardware here. Due to the lack of competition from AMD until recently, Intel was in a period of pretty slow improvement for a number of years. Performance would improve maybe 5-10% annually with each new CPU release. Since a lot of companies replace hardware on a time basis for reliability reasons, rather than strictly on a performance basis, it meant that there was a lot of older used hardware that was still very high performance, perhaps 80-90% of today's CPUs but at 50% or less of the cost. You can still find a lot of used dual Xeon machines on eBay and other sites that are a really good deal for hardware that's 3 or fewer years old and still performs really well. To outfit a render farm on the cheap it can be a great way to do it, especially if you buy a really reliable workstation model/brand and/or you have in-house hardware expertise and can e.g. swap out an SSD or RAM should the need arise.

Anyway, if you want to buy new, here are the options as I see them:

Starting at the best price/performance ratio, which assumes space is not an issue (because we're talking about single CPU machines). I'd recommend a Threadripper in that case. Ryzen 7 1800x is 8 cores at 3.6Ghz, which is 28.8 "Ghz equivalent" (this is the term I use for these simplistic calculations, and keep in mind they are not precise measures of expected performance, just ballparks). The 1800x costs about $500 right now. The top of the line Threadripper 1950X is 16 cores at 3.4Ghz, a slight drop in per-core clock speed, but twice as many cores, 54.4 Ghz equivalent, so obviously a lot more powerful than the 1800x. It's about $1000, so twice the price of the Ryzen CPU for not quite twice the Ghz equivalent, however when you factor in the cost of the rest of the components (making total system cost less than 2x) and inherent limits of space, it's still a better deal to get 1.9x the performance for less than 2x the price.

So a single Threadripper 1950x machine with an SSD and basic graphics card (because TG doesn't use it for rendering) would be a great render box, and will be hard to beat for price/performance ratio. In comparison, the Intel i9-7920X is $1200, and 12 cores at 2.9Ghz = 34.8, much worse than Threadripper (not to mention the motherboards are more expensive, I believe). The 7900X is a slightly better deal at around $1000, 10 cores at 3.3Ghz = 33. So slightly slower, but cheaper and remember we talked about efficiency, so 10 threads at a higher clock is probably going to be better, especially given it's only 1.8 "equivalent Ghz" slower (and $200 cheaper). But both are a good deal slower than Threadripper, so no contest really.

The upcoming Core i9-7980XE - launching this month -  is going to be 18 cores at 2.6Ghz = 46.8, but it's $2000 (msrp). In leaked benchmarks it appears to be outperforming Threadripper on Cinebench, but Terragen is a bit less efficient with high thread counts than C4D, so you're still probably better off with Threadripper, even putting aside the doubled cost of the 7980. That being said, one complexity with the i9 CPUs and - to a lesser extend Ryzen and Threadripper - is that they can overclock themselves, sometimes on all cores at once, depending on thermal conditions. So if you cool your CPU and overall system really well, these Ghz values might be well exceeded, which could account for some of the i9's performance lead over Threadripper (aside from architectural differences, of course). The 7980XE can theoretically do 3.4Ghz on all cores in the right conditions, which brings it up to 18x3.4 = 61.2 equivalent Ghz. Still if you are rendering for many hours on each system you can't necessarily count on that overclock being maintained, so I always feel it's best to consider base clocks, where Threadripper wins handily.

Now if you need more performance with the same space footprint (i.e. a single computer case), you'd go dual CPU. This is going to get you more performance still, but at a notable cost increase. It's not the most performance or the most money you can spend (that honor goes to quad CPU or octa CPU machines), but due to threading efficiency it's probably the most you'd want to spend on each machine, unless space is at a real premium.

Unfortunately for AMD, their Epyc line is pretty low clocked, with lots of threads, so not the best for Terragen performance. A single socket Epyc 7551 is 32 cores at 2.0Ghz, or 64Ghz equivalent, but since you're using twice the threads of Threadripper, actual Terragen rendering performance might not be much greater. And that's also a $2100 CPU. If you step up to dual CPU Epyc, for the same performance level in the 7501, it's $3400 *per CPU*. The Epyc line tops out at 32 cores at 2.2Ghz = 70.4 per CPU in the 7601, but that's even worse at $4200/CPU. Still, Intel's Xeon line doesn't do any better at full retail price. If you buy used, as I mentioned above, you can get some slightly older but still very high performance systems for a bargain, which would easily compete with Epyc for price/performance. But if you're buying new, I'd suggest either Threadripper (a real bargain for the performance), or a dual Epyc 7501, which is going to give you 64 cores total, at 2.0Ghz = 128Ghz equivalent, which is crazy. Honestly with that number of threads (including hyperthreading it's 128 on a dual CPU setup like that!), you'd probably want to run 2 parallel instances of Terragen for best performance.

With both Epyc and Xeon you can go up to quad CPU, but then you're looking at $10-20k or more per machine, and it just doesn't seem worth it. :D

So bottom line: Threadripper is the sweet spot, best price/performance for new hardware. Get a dual Epyc if you need more and are buying new, but you'll save a lot of money buying a used dual Xeon in that case. For dual CPU I'd definitely recommend the used Xeon route. I have an HP Z820 workstation with dual 2.9Ghz 8 core CPUs in it, plus 128GB of RAM and dual SSDs, and it works great. It's in the top 10 of benchmark results for the old Terragen 3 benchmark and it was about $1800 on eBay. It even included about 6 months of remaining warranty support, with on-site tech visits.

Let me know if you have any questions.

- Oshyan
Terragen Discussion / Re: Rock Won't Displace
« Last post by fleetwood on Today at 10:12:53 PM »

Did you also transform the powerfractal to make sure that both rock objects receive the same info?
Say the left rock is at the origin 0,0,0 and the right rock is at 2,0,0.
Then for the right rock you'd need to transform the powerfractals output with +2 meters for an exact match of input.

Sorry, I neglected to transform the fractals in those earlier tests, but have done it here.

Both examples here do show the incorrect shadows made by a displaced rock object. The shadows of the rock object appear to be pre-calculated based on the undisplaced object.

Example 1 is the 20 face rock object with one subdivision on the left and the 80 face rock object with no subdivisions on the right.

Example 2 is  a 256 face rock object with 4 subdivisions on the left compared to a Terragen displaceable sphere on the right.
        Note that the shadow of the rock object is just a simple round shape as compared to the more complex shadow of the sphere object.

I would repeat: anyone wanting to make a hero rock, it is probably better use of time to work with one of the Terragen displaceable objects like the cube or sphere or an imported high poly object.

Pages: [1] 2 3 ... 10