Daz render speeds?
neo_xenophanes
Posts: 6
First off, sorry if this has been answered before, but I'm wondering which computer components are the most important in reguards to speeding up the render process. Is it the CPU? The GPU? The RAM? I'm looking to get more serious about rendering and I'd like to make some imformed decisions when I upgrade my machine so I get the most out of the upgrade that I can. Thank you in advance!
Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
Everyone will say that depends on your render engine (3dlight or iray) but that's only applicable if youriray scene fits in the GPU's memory.
Mostly I would say CPU as 3delight only uses CPU for rendering. And in Iray, if the scene doesn't fit in the card's ram, then it will fall back on CPU. (If it does fit on the card, then GPU can be much much faster than CPU)
Ram is also pretty important. The more Ram you have, the bigger the scenes you can create, but also if you computer has to resort to dumping the ram to temp files because it ran out, that can slow down rendering a lot.
As Kaotkbliss says, it depends on your rendering needs. However, since you say you are wanting to get more serious about rendering and since you appear to be thinking about a serious computer upgrade (or new computer?) I would suggest that you first answer the question about 3Delight or Iray. Iray can give technically better results as the lighting system is more akin to reality (if you are willing to spend some time learning how to use it, you can even get spectactular photoreal images). 3Delight is quicker and you can cheat a bit with lighting.
If you choose 3Delight as your main render engine then normal RAM and CPU speed are important.
If you choose Iray as you main rendering platform then CPU is much less important since the bulk of the processing will be done by your GPU. In this case you must use an Invidia card. If you can afford it go for the new Nvidia10 series. For Iray renders the number of CUDA cores is important. This is correlates to the size of you GPU RAM as Kaotkbliss said. The more CUDA cores the bigger scene you can render without dumping to normal RAM and the faster the render will take place. Having a few gigabytes of normal RAM is not a bad idea anyway.
Remember that specific render settings and lighting setups can greatly affect render times (for both 3Delight and Iray). Some setups give good results for relatively low resource cost, whilst other CAN give great results but WILL use a lot of resourses and hence be slow rendering.
With rendering it is always a trade off.
Thank you so much for the feedback guys!
I am looking to use more Iray rather that 3Delight, so it looks like graphics card will be my next move. I'm only upgrading parts rather than the whole computer. My current computer is just a frankenstien of parts, as things break or get outdated, I just upgrade. I upgraded to a GTX 960 at the start of this year and I upgraded from 6RAM to 24RAM 2 months ago. MY core is quite old, but appaerently it's not too bad, it's a quad core i7 @ 2.67Ghz.
As of right now in my basic test render scenes, it took about 2 hours to render them. I'd say they weren't too overly complicated, mostly just a female model posing in a studio setting with some backdrop curtains and a chair and a few poses and outfits. I guess 2 hours isn't bad, but if I get more complex, I'm worried because I've heard some stories about renders taking as long as 10 hours or more?
This is why I was thinking about maybe making a jump to a Titan X model GPU (or two Titans) unless somebody has a better suggestion?
Again, thanks for the info already provided!
I've had an iRay render going for a week now and it's only at 23% converged LOL
That's mostly to do with LOTS of emmissive and semi-transparant items with lots of detail behind them
What kind of specs on that machine kaotkbliss? It's renders that take that long that intimidate me and are making me seriously consider dropping another 2-3 grand on upgrades. :x
Size, content and composition of the image probably play the biggest role in how long it takes...massive resolution, with a lot of detail/SSS/reflections are going to take a long time, no matter what the machine specs are.
this is only like the 2nd render or so that's taken more than 4 hours to get to 95% convergence. Most take under 2 hours.
I've got an AMD athlon x3 with 16GB ram, so it's an older processor and I'm running in CPU only mode
I've got 15 different objects in my scene that are set to something like .003 opacity with emmissions, 6 of those also have the water shader applied so they also have reflections. Then I've got 20+ characters in the scene.
I have a thread here about looking for help speeding up iray renders and in that thread is the image I'm currently rendering when it was at about 2 or 3% converged.
Thanks again for all the help Kaotk!
I went in to bestbuy today to talk shop with one of the people there. I'm pretty sure he gave me the wrong information though. When I told him my intention was to buy two GTX 1080 cards or Two Titan X (Pascals) and use them in SLI, he said that doing so would not increase render speeds and that it would only be useful for high end gaming. I'm pretty sure this is not accurate info.
If anybody can clear this up for me it would be a great help. 2 cards in SLI will render noticeably faster than one card alone? Yes or no?
Perhaps they meant, you should switch SLI off for rendering and switch it back on again for gaming.
Which would be correct.
So you're saying that using 2 cards won't help render speeds? Hmmm. That's troubling
You can have multiple cards, and they will speed up renders, but they don't have to be in SLI.
SLI is a specific type of inter-GPU communication which can interfere with some 3D programs, it's generally recommended to switch it off just in case.
Yes multi GPU's will decrease render time as long as they meet the requirements for iray. You just can't run them in SLI mode. Iray will read both cards.
Don't you mean increase speed or decrease time?
Two of the same card will not be 2x as fast as one card, but they will be noticeably faster.
Yes sorry. I fixed it.
Oh okay I understand now, I'd use both card, just not in SLI mode. Thanks for filling me in.
ALso,
I tried a generation GTX1060 card and ran it concurrently with my GTX 980TI and I was not getting any joy with the new 1060 card. I just found out that DAZ has not finished testing the support for GTX10xx cards yet and won't support any 1080 cards - yet... FYI
Check out this post:
http://www.daz3d.com/forums/discussion/115686/dual-cuda-card-dilemma#latest
It's not just Daz...nothing that uses Iray is supporting 10x0 cards...unless there was an update to something the past week or so. Nvidia has been slow to release the updates on its end, so nothing downstream will have support until Nvidia releases the final version (and then that will need to go through whatever testing the application developer does). Studio does have support in the private beta pipeline...at this point it is anyone's guess as to when it will make it public or release.
I installed my second 980TI and am more than pleased. I thought one 980 was fast but I have effectively halved the time of all my renders today. I went from 3 day moderate sized renders on the CPU to gigantic renders in 15 minutes. I stayed with CUDA values inside the 'MAXWELL' generation of cards and the 5632 CUDA performed twice as fast as expected. I am diggin' the speed. If I wanted I can now do SLI with games since they're identical but I would need an SLI bridge long enough for the inside card to reach the card sitting 1.5 feet away on the outside. I'll worry about SLI if I ever play a game with SLI capability. Since I shouldn't use SLI with Iray renders I probably won't even bother hooking it up.
I've done three renders already today of varying sizes and complexity and so far two identical cards have proven to be twice as fast, almost exactly halving the time of all my renders.
It can be close...but most of what I've seen is closer to 1.5 to almost 2x. It does, to some degree, depend on other factors, too, as to how close you'll get to 2x (or 3x, etc). But in any case it will be faster.
Really? That's strange, mine have been coming out to something like a whole 1.9 - 2.1 times faster. I have yet to see anything dip below 1.9x faster. And I take it all the way to 100% convergence. I was reading in the Nvidia software developer forums about what are the best languages to code for the CUDA platform and I happened on a post where one of the developers was making suggestions to make sure to uncheck the CPU box in the render setting advanced tab besides checking all the CUDA GPUs. I suggested this to a friend who was running both CPU and her GPU in her Iray renders. When she unticked the CPU box in interactive/photoreal sections and left the GPU ticked/enabled she saw a marked increase in speed after that. The post mentioned that CPU + GPU creates bottlenecks in processing time. I have yet to even try CPU + both of my 980s enabled but I don't intend to test that out. I am enjoying faster renders to even want to go back...
This render with an HDR background but lots of polygons on the ground and reflections- took 21 minutes to render to 100% with one 980 TI card. With two running it took 10 min 10 secs to get to 100%. As you can see it got to 90 % in about a minute and a half but true to form 91-96% took 3 minutes and 99% was stuck on the decimal fractions for 6 minutes till it finished at 100%
Not saying it can't but it's more typical to be on the lower side not the upper...but a 1.5 to 2x speed up is worth it.