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
You need the right drivers...I think you need to install the Mac CUDA drivers.
Thanks... I was able to install the drivers from the Nvidea site, and sure enough, the GPU is now listed as an option. Thanks!
My first test showed a *dramatic* reduction in render time. However, my system crashed in an ugly way after the render was complete and I was trying to switch to another application. Hopefully an isolated incident... or something that will be fixed soon.
At 2GB of VRAM, you're on the lower end of the ability to use Iray. If your scene exceeds the memory of your card, then ugly things can happen. Textures are your main direction to look. Try to minimize the memory use of textures in your VRAM.
Kendall
I'm trying to follow this discussion and just plain can't. I got lost the minute someone threw out numbers. You throw out numbers and I'm going right with the conversation through the window.
I have no idea why my render is so slow but it's been stuck at 54% for 3 days. I'm not even able to look up the settings to tell you guys properly what's going on. But I do have one of the settings changed from memory to speed. I have a NVIDIA card. I told it optimization to speed things up. It gets to 54% in just a few hours and then... hangs.
I have the scene set to use only lights within the render because the dome would bring too much light to this. I have 2 planes that are emitting light as the sky. I have a huge house with every window lit. I have 3 spot lights on the central characters, three women. One is behind them because I'm a fan of rim lighting.
The odd thing here is when I was rendering it slightly smaller it got a bit further. I stopped it because I had to tweak a costume. Now that I'm trying to render full size, I get this issue. And I'm not sure what to do to resolve it.
When I have long render-time issues in a scene, what I do is process of elimination. Maybe because I am a scientist. I turn everything off in the scene but ONE thing, say a character. Bald. Naked. (Turn off clothes, hair). And ONE light. Render. How long does it take? Usually it's quick, couple minutes. OK, so it is not the character. Now I add back in the lights, one at a time. Render. Time it. All lights on, OK? Hair next. Render. Then clothes, render. OK. this character is rendering fine. Next we add back one prop. Render. Then another prop. Render. Then character #2, naked. Then hair. Clothes. And so on.
Usually, you have some thing, often ONE thing, that is crushing your render speeds. For me it was 2-way lighting on a tubular fluorescent light. That one thing alone was shooting render times from 45 minutes to 2.5+ hours. I found this out by process of elimination.
Finally, try HDRI if you are not using it, instead of mesh lights. My back-of-the-envelope experiments indicate that HDRI is a good order of magnitude faster than even a single mesh light, for the exact same scene. Sun and sky setting is also faster. If you can get a good HDRI light setup, you might see faster render times. And, IMO, the renders tend to look a little better, especially for human skin. I dunno why, but this is my admittedly anecdotal experience.
LOL, if you were a mathematician or an engineer you would turn off half the items, render, then decide whether that half or the other is the problem, and half again.
Heheh. I suppose you could do it like that too.
The reality is we do have some good reasons to guess certain things are causing more trouble for render times than others. For instance, I've learned over time, mostly from threads here but also from experiments, that things with high subD displacement maps will kill render times but things with bump maps won't... and if possible (it isn't always, but sometimes it is), you can speed things up by converting your disp map to a bump map and raising the bump level a lot (2+) and taking out the displacement. Doesn't always work but it sometimes does and it can dramatically speed up the render.
I think the Iray addition to Daz is awesome but it sounds like I am having similar problems with render speed as well. Besides the speed I love it Iray is an awesome add on to the Daz studios already awesomeness. I have a good deal of the NVDIA supported products in my pc due to the fact of performance with some of the other software I run and have ran in the past. such as mudbox which I switched to zbrush after getting a pc just to run 3d applications like mudbox which will not run without NVDIA support. I imagine that the Iray will improve over time cant wait to see what Daz implements next. the pic attached was my first with Iray and took around twenty minuets to render
As I just posted in another thread, the huge advantage of Iray is that, once you figure out how to do the lighting (it took me a good month), you can get good results out of the box. I don't have to do as much fiddling with lighting angles to get realistic looking renders as I did with the old 3DeLight system.
But the render times... my goodness. 40-50 minutes to do what 3Delight can do in 3 or 4 is just sad. It has dramatically slowed down the production time on my webcomic. Fortunately I have a 26-page buffer, but still... I'm going to run out of buffer eventually.
Daz iRay Render Speeds is the title of the thread, well, I have numbers!, depends so much of the kind of light and the setup of iRay materials, and you don't need to reach 100% for achieve a decent render, the images speak for themselves:
I'm sure it was already mentioned, but GPU-based render speeds are influenced by the video card, specifically the number of CUDA cores, memory bandwidth, and memory interface width.
A single GTX 780 Ti will render the same scene faster than a GTX 980. I know this because I have both, and compared render times on both.
The 780 Ti has 2880 cores, 336 GB/sec memory bandwidth, and a 384-bit memory interface width, and a higher texture fill rate but slower base and boost clocks than the 980.
As well, and this has been mentioned in another thread, adding the CPU into the mix does not always speed up renders, but may actually increase render times.
Rendering the same scene (single clothed and haired figure using the default Iray HDRI image), I got the fastest time and fewest iterations running two 780 Tis with SLI disabled. I don't have my notes with me, but IIRC it did 250 iterations in 37 seconds. Adding the CPU (i7-4770K) took 10 seconds off that time. I'll double-check when I get home and post the results. I did 12 tests I think, with all the various combos I could think of (i.e. CPU+1 GPU with SLI enabled, both GPUs with and without SLI, the display-driving GPU alone, etc).
Granted a more complex scene will take longer, but as a general rule, the setup that gives the shortest time for a simple scene *should* also give the shortest time for a more complex scene.
I'll also note that rendering with CPU + both GPUs gives my 1200w battery backup unit fits. Time for a new one, I guess.
Fully true ! Yes, as said a good member previously in this thread, It is possible to stop and save a render unfinished 100%, lets say in a range of 15-50% state of progress without noticeable loss of quality, but this is true for backgrounds or solid props only , NEVER for people skins!
As an example my V6 1024x1024 was up to 95% render time after 18mn. but reached 100% 24mn. later only! And take care, be sure that all values in skins render quality are processed during those last 5% of final touch ! Volume, luminance, subtle skint tones of beauty come at this moment only. However the last step (99%) looks quite unfinishable long and very hard to bear, lol.
I also own Reality sine Version 2 and the only thing you get here are errors after errors. Since Version 4 you can get stable renders and if you're lucky enough, there are no unwanted freckles, and, if you decide to make a big effort and the final picture has dimensions like 20 MP or above, you can count on up to three day for a desired result. Thats not the way anyone who work within this world of creativity would like or need!
So I'm very happy about this solution because IRAY gives full access to the whole bandwith of any tools that I was dreamed about. Any other at this level would costs thousands of bucks!
(this is a scene I was made with 4.6 and converted with IRAY shaders)
Greetings
Image removed for nudity. Please see this thread for info: http://www.daz3d.com/forums/discussion/3279/acceptable-ways-of-handling-nudity#latest
The big difference between Iray and all the rest is PA/vendor support at purchase. It is easy to do and since a huge volume of people have it/use it it is more than worth the time to make products support it. There is no "buy in" for the content creators either so we all have it. Lastly since Iray is the Studio default render engine promos done using it are actually examples of what the user should expect to be able to do with the product.
I’d like to run this test using my GTX660ti (2MB), but can’t get GPU to show up as a render function (even in the most ridiculously small scenes using only a sphere with a shader applied), I think due to a possibly outdated Cuda driver (I’ve got 5.0, I think).
In order to try and make this work, I am wondering:
(1) What is the minimum version driver I need to make use of my Cudas?
(2) Will that specific driver work on Mac OS 10.8.3?
(3) Do I have to install a “CUDA toolkit” with that, or just a (the) new driver to make things work?
(4) If I mistakenly install a Cuda driver not supported by my system, would my display go dark (requiring me reboot using an install CD in order to revert and reinstall the last driver that didn’t make my display go blank)?
Answers to any or all of these questions would be most appreciated!
Installed a more recent version of Cuda (6.5.18) and it's working. Awesome speed difference. The update didn't create a blue screen of death or any problems with my monitor. Thanks for the help
6.5.5.1
http://www.nvidia.com/object/macosx-cuda-6.5.51-driver.html
anything newer will not work in OS 10.8, you might consider upgrading which is free but back up everything with Time Machine or Carbon Copy Cloner before you set out. Yosemite 10.10.4 is giving some users a lot of grief to get working.
my PC cofiguration Athlon x2 240 2.8ghz (two cores), 4GB RAM and Geforce GT 630 2GB i know it's slow for 3d but if change something I must change everythingfrom motherboard and can't rightnow.
So here is a question. When i start render I must wait about 20-30 minutes before iratiation starts and can't see how it's going just empty screen. So please tell me how to reduce this time that I can make quick tests?
i tryed change converged ratio from 95% to 0% and also render quality but it's not reduce and i think even increase render time.
The render on standart quality took about two hours sure if resolution not extreme but how to set daz to start irratiation faster?
What's killing you is the dual core processor. Getting a new video card is possible...and a 4 GB GT 740 is pretty cheap and may allow the entire scene to fit in the card's memory...which with a 2 GB card is not likely, so you are running in CPU mode.
4 GB of RAM isn't helping either, but I'm guessing it's DDR2 which is outrageously priced right now.
iRay demands a lot of powerful hardware, or at least most recent hardware to perform the way it must be intended for. doing this with a videocard with limited ram and just 192 CUDA cores is painful. converged ratio must be in 95%, render quality must be at 1 value and ON. for quicker rendering, if any, try do HDR rendering only, selecting Scene Only, and avoid mesh lights, and turn off camera headlamp, Architectural and Caustics in OFF, both.
Those suggestions will help whether the scene fits on the card or not...but really not much is going to fit in 2 GB, especially if that's the only card in the system.
How about a bit of chat about the render quality setting. All of the samples I've posted here have been at quality equal one and ther have been no questions / suggestions etc. I've compated renders at qual 1 to renders at qual 5 and can't see differences even when blown up to the point the image pixellates.
What's the seal since the docs have (qual=) 0 infor on what different settings do for this parameter.
The answer is simple. I have two slots on my motherboard for Ram and each is 2GB. You right it's DDR 2 800hz and they can't be 4Gb each only 2.
Yes it's kills my CPU during rendering 100% and can't even move mouse cursor sometimes.
I bought this video some months ago to handle this processor and Powersupply.
Do you think GT 740 will handle 300-350 WT Power supply or this CPU can run better Power supply? How much cuda cores it has? Can you tell me the full name of video with 4GB, Superclocked? or maybe there is something better and with better price? i mean less expensive or a little expensive but with better characteristics.
Thanks for advice so much.
By aovid mesh light you mean avoid light surfaces like planes? Never try HDR maps before just render with standart 3dlight. How to use them?
Thanks for advice so much.
I think my video have 96 cuda cores cause GDDR3, also I saw other model with 384 cores 2 GB but GDDR5 mine is 3.
Is that any posible way to know how much ram of video memory needs scene? And so if I understand right basic settings it's the lowest for tests. No possibility force the iteration process start faster with this harware right?
Are you using a laptop or desktop? If it's a desktop, do you know the make and model of your motherboard? (It will be printed on the motherboard, usually big enough to read easily)
In most cases you don't need more than 1. Setting it to 2 will help clear up some combinations of multiple reflecting/refracting surfaces. While I am sure there is a use for a higher than 2 setting, I have not had cause to need a higher setting yet.
I am on desktop. Motherboard is Gigabyte M61PME-S2P according AIDA64 Extreme. Asking here in Russia they don't have 740 GT 4GB but have 730 GT 4GB GDDR3 but I think it's 64 bit and really slow.
My video is 730 GT 2GB not 630 was a mistake. So solution is take something like Ti with new power supply but I can't right now it's pretty expensive.
It looks like your board will hold up to 8GB Ram so you can get 2x4GB sticks :)
Also, Support for Socket AM3/ AM2+/ AM2 processors: AMD Phenom™ FX processor/ AMD Phenom™ X4 processor/ AMD Phenom™ X3 processor/ AMD Athlon™ X2 processor/ AMD Athlon™ processor/ AMD Sempron™ X2 processor/ AMD Sempron™ processor
(Note) If you install AMD AM3/ AM2+ CPU on AM2 motherboard, the system bus speed will downgrade from HT3.0 (5200MT/s) to HT1.0 (2000 MT/s) spec; however, the frequency of AM3/ AM2+ CPU will not be impacted.
So it looks like you can still get better CPU with more cores (Supports AMD Phenom™II X6 processors) and I saw this with a quick search
http://pcwarehouseoutlet.com/index.php?route=product/product&product_id=8126
So you should be able to find something to improve things until you are able to do a better upgrade.
No my is not support DDR3 only DDR2 but DDR2 can't be more than 2GB. DDR3 have diffrent slots but I have only two separate in my motherboard and they are DDR2 support
Sure you can. DDR2 comes in 4GB size. Here's a whole page of them on newegg.
http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&IsNodeId=1&N=100007611 600006042 600006067
*edit*
there are even 16GB options for 2x8GB but your board will only see up to 8GB total
I tryed fiind DDR2 4GB in my town in Russia some years ago but couldn't find nothing more than 2GB. It's also really old I don't think they sell right now cause also it's pretty expensive.
By the way it's really strange I am curious if they really can be.
Anyway thanks for willing to help.