I have heard that DAZ versions 4.20 and 4.21 render much slower than prevous versions
kwannie
Posts: 869
in The Commons
In one of my dicussions someone suggested that DAZ versions 4.20 and 4.21 were much slower in rendering than previous versions. Now, I am curious which version of DAZ Studio yeilds the fastest render times. Any opinions?
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
This puzzles me. I've been using 4.21 for a while now and I've not noticed any difference in render times.
Cheers,
Alex.
I haven't noticed anything like that either. In any case it should be easy to test.
I heard the same thing before trying it. I am still on 4.15, but have been running the 4.21 beta for a few weeks and don't notice it being any slower.
I recently built a new system and only have DS ver 4.21 installed. On my old computer I had DS ver 4.10 and ver 4.16 so getting these older versions of DAZ installed on the new system requires a few hurdles. But a very reputable person in the forums commented that vers 4.20 and 4.21 could be rendering as much as 50% slower in some cases.
I'm on 4.21, didn't upgrade my drivers because I'm running pretty much running out of time these days (GTX960M), and it's rendering faster for me. Also rendering much faster on CPU for some reason.
I just tried out a RTX 3090 and an AMD 7950x and the GPU obviously can handle larger scenes now with 24 GB but it doesn't seem to render much faster than a RTX 3060. The CPU with 3DLight is much faster with 16 cores. but I have not tried out the RTX 3090 with previous versions of DAZ such as DS ver 4.16 and before, so I dow=n't know if those versions would be faster than ver 4.20 and above.
I think what you're describing is the changes made to Iray that forces non-RTX cards to emulate RTX features, although a) I believe that was introduced earlier than 4.20 and b) that affects the amount of available VRAM, not the speed of rendering. Most of the reports I've seen are that the recent versions render faster.
Well I won't readily identify the person that stated it, but he is a very trustworthy source in forum and here is what he stated. I do trust his input more often than others, but I would like more details about this if anybody has any. I quote him below.
The Iray in 4.20 and 4.21 is slow. It has been proven slow in the benchmark thread. The Iray Dev team has also publicly acknowledged this problem. Sometimes it can be over 50% slower.
Iray 4.20 fixed the glitch that allowed for ghost lights. So all older ghost lights do not work. New products have released, but they look different and are prone to fireflies if the ghost light is in the camera view.
Iray 4.20 changed how thin film works, so some water looks gray under certain conditions.
Iray 4.20 can also change how certain HDRIs look.
Iray 4.20 introduced volumetrics through VDB file support.
I've read much the same kwannie from a few different sources, not as much as 50% slower but definately slower than 4.15. It's the reason why I'm still on 4.15 if upgrading means slower rendering, particularly as I've got an older GPU.
..yeah I'm running a Maxwell Titan-X and render speed is fairly good, just don't like the hit on the VRAM. According to the logfile after WDDM a total VRAM of 11.9GB is indicated (meanwhile on MSI afterburner with Daz open only 286 MB if VRAM is shown to currently be in use (empty viewport) With Daz closed the VRAM usage drops to just over 200 MB (dual display system).
Running on W7 Pro SP-1
What I don't see is what the RTX emulation takes from the remaining VRAM. I looked at the start-up of a previous render session, but no indication there once Iray rendering is started.
RTX Emulation takes about a gigabyte
...thank you, I wish it would be confirmed in the logfile .
I have a 3060 but it's in the box as my MB's BIOS is too old to recognise it and there have been no further updates since 10 years ago. I'll have to wait until I can get the funds up to upgrade my system to W11 standards to use it, not an easy task on a fixed income.
Sometimes I become very confused because I guess I'm just dense when it comes to DS in as so much what will work and what won't with my PC setup. My PC is about three years old now and it's home built PC. It has an i7, six core CPU, 32 gigs of RAM a 650 watt PSU and I still use a GTX 1060 6 gig graphics card and use version of DS 4.12. I see so many folks on the forums using RTX cards and sometimes it make me wonder if I can even run 4.21. I know there is a beta 4.21 but I've never used a beta version of DS. The last version I have backed up is version 4.11. I want to update but if there is still issues with version 4.21 it is the reason, besides hardware, I ma dragging my feet.
I've noticed that lights do look different and it's harder to get lighting the way I like it but the trade off is I haven't had a single crash since using 4.21 when 4.15 would crash with one character and hair. I haven't noticed slower rendering times at all. I have an old 1080ti that I thought was obsolete until I tried 4.21. For me, it's MUCH better except for figuring out how to light it correctly.
I have tested 4.15 or 4.16 vs 4.20 on a RTX3070 laptop, a GTX1070 desktop and a RTX3060 desktop using the benchmark scene and in all cases 4.20 was slower. The results are in the benchmark thread starting on page 29.
http://www.daz3d.com/forums/discussion/341041/daz-studio-iray-rendering-hardware-benchmarking/p29
I don't have any recorded data that 4.21 renders slower than 4.12 (the last version I used), but it doesn't "feel" slower. I can say that 4.21 crashes way less for me than 4.12 did. I never let a render run longer than 30 minutes, and I haven't noticed a subjective quality difference with 4.21 with similar render settings. 2070 Super on my desktop and 3070 on my laptop.
4.20/21 has been consistently slower in almost every direct test I have done against 4.15/16. When I say that, it is not because I "felt" it was slower. I confirmed it was slower by rendering the same scene in both versions and recording render times from the log. Thus confirming my findings. I also looked at iteration counts, because these were very interesting. Sometimes 4.20 was running insane iteration counts without seeing a real improvement, which led to those longer render times.
I tried reinstalling my Daz exes. I tried different drivers. I observed this issue with two 1080tis, then a 3090, then a 3060. Four different GPUs in multiple combinations.
There were 2 scenes where 4.20 did render faster than 4.16. I do not know why, I tried to isolate whatever it might be to no avail. However, I want to point out that when 4.20 rendered faster, it did so because it ran a lot fewer iterations than 4.16 did, and the resulting image had a bit more noise. So I don't know if I would call that a win.
Now I did not try every setting. But I tried several. I tried using caustics, changing specular render settings, changing the quality settings, ect. I tried building a scene with lots of geometry, and a scene that had more complex shaders with less geometry. The results were pretty much the same, 4.20 was slower.
At any rate, your results may vary, after all, we all use Daz differently and build very different scenes. Still, I am not the only one who observed and reported this. You can see a very consistent trend in the Iray Benchmark thread where every user running 4.20 was rendering the benchmark scene slower than they did with 4.15/16. So at least in this scene, the trend was repeatable by everybody.
I do use 4.21 when I have to. It does have improvements for creating Daz content. But when I need to render, I go back to 4.16. I have 4.21 in beta form.
As for stability, I have had maybe one crash in my year+ with my 3090 and Daz 4.16. So 4.16 has been a rock for me. Interestingly, I had more crashes with my dual 1080ti setup before getting my 3090. I do not know if that is hardware or what, but my 1080tis gamed like a champ.
There is good news. When the Iray dev team finally acknowledged the performance regression, they announced a new version of Iray that supposedly fixes it. We just...have to wait on Daz to get that version down to us. Hopefully they don't take forever to push it out.
So whenever this comes, we hopefully will recover the speed that was lost. Until we actually see the data, this is all just talk, IMO.
There is more good news, ghost lights are coming back as an official Iray feature, not a glitch anymore. That is legit great news, though I wonder how the new ghost lights will look in action.
However, I still will keep my 4.16, because a number of scenes do look different in 4.20, and I value that consistency in some of the things I do. For those particular scenes, I have been unable to replicate my lighting in 4.20 the way I desire it. So 4.16 must stay for now on my machine.
As for the Ghost Lights, I've just tried to update the FG Bluson Library which is using these for some of its lights. It generally went fine just by increasing the Luminance values for the Emissive light shaders. There were however one thing that were different. If you look at the render below and compare it with the main promo on the product page, you'll see that that "lightball" effect on the Chandelier lights is missing. I'm not sure whether that has to do with the Ghost Light changes or what. Think I'm going try to render it in 4.15 to see how it looks there.
What is emissive? In the product shot there seems to be an near-spherical bulb that is glowing and then the cup shaped glass shade is around that - in your image the shade itself is bright as if glowing.
The Chandelier lamp units consist of the LampShade, the Light (bulb), and the LightPanel. See screenshot - I've colored the Light (bulb) blue and the LightPanel yellow here.
The LightPanel is the emissive, normally it has a Cutout Opacity of 0.00001, I've just changed it to 0.3 here to make it visible.
If I increase Luminance for all the Ghost Lights in the scene by a factor 100.000 which I understand is how you update to the new DS Ghost Light standard, I get the render I posted, which also seem to match the lighting intensity in the main product promo. So, apart from a slightly different color tone (a bit more reddish I think), that seems to be as it should be.
I just don't get the same look for the Chandelier lamps as in the main promo.
DS version is 4.21.0.5
At least some of that looks like bloom, based on the difference in the light on the lamps on the desk. If the hottest areas of the chandelier were smaller, with bloom, I would have guessed it would be closer to what was in the render in the store. I do not have the set to test this myself.
I have tried to play with the bloom settings, but can't find any way to produce this effect. When I get the time I'll try to revert back to 4.15 and see how it looks there, maybe that can give a clue about what it's about.
Which new GhostLight standard? Increasing luminance will compensate for the dimming of semi-transparent emissive surfaces, though at the exense of making them visible in reflections, but the beta has the offiical nVidia approach which is to provide a way to hide the light soruce, without needing transparency, from both direct view and reflections via new properties. The main issue is that it is an object-level setting, so you would need to in soem way (e.g. a Geometry Shell) double the object, making one visible but hiding the emissive plane, and the other mostly hidden, with the new properties, and with the plane visible and emitting.
OK, so they have updated it again. Will try the beta and see if I can make it work the new way. Thanks!
I have two RTX cards and it is painfully slow. And several of the functions are different. I can't left click and choose things in the figure anymore. I find the new version painful!
Make sure you are using a node selection tool (Node Selection itself or one of the pose adjustment tools) and in the Tool Settings pane, make sure the Node Selection Mode is either Node Only or Root Node first. Tool settings and the current tool are not shared between release channels
So the hit to vram--could that explain why I find myself often having to save scenes, close and reopen the program and reload the scene to get the vram to work? that sucks, if so. I would go back to 4.15, but I have so many scenes now saved at 4.21.
I mean I love the volumetric clouds, but it's been annoying that I used to be fine with vram even over 20 gigs of memory. Now it sometimes cuts to cpu at 16 o 17. Then if I close and reopen the exact same scene renders with gpu.
THat said, I can say that cpu renders seem faster than they used to be.
I really wish I could go back, but it would mean giving up some stuff I like.
CPU rendering actually is faster than before. That is one positive of the newer Iray. In testing between me and skyeshots, we found the CPU was roughly 18% faster than before. So people who are stuck CPU rendering can rejoice a little bit. But for everybody else on GPU, the rendering speed is indisputably slower.
There is one thing that the new Iray has improved, Guided Sampling. It is faster than the Caustic Sampler, and picks up some of the extra light paths that Caustic Sampling can. However, it is still slower than just rendering the scene without caustics. In some respects the Guided Sampler can have less noise than the using the default settings, but 4.16 is fast enough to where it can pump out more than enough iterations to make up for this.
A lot of stuff you save in 4.21 will work in 4.15. New features, like VBD of course will not be loaded. Genesis 9 will load as well. The only things that don't load for G9 are dev tool related shortcuts, and the "reset to zero" thing.
VRAM could be the result of extra subdivisions. Some new products load at very high subD levels. If these items are close to the camera, you can turn them down. Also, the scene in viewport takes VRAM as well, and some things load again at higher levels of detail than necessary.
Don't forget to check your Iray Advanced settings for compression. This setting is not saved with a scene, but is saved with the DS session. So if you change the compression settings and forget about them, you may have textures not being compressed enough.
I point these out because I haven't noticed 4.21 using more memory, though I don't use 4.21 that often because of its performance. I could look at it closer at some point.
if anything I've noticed the Iray preview mode renders and responds faster than other versions.