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
And yet another reason for not trying to buy an RTX 5090. Well, that's some money that I will save.
Could using Iray Server be a temporary fix to take advantage of a RTX 50XX ?
https://www.irayplugins.com/iray-server/
I did wonder about that, but don't have official word.
If anyone needs another reason to look at Blender, it has supported the 5090 all the way back since version 3.6, and a 5090 is 36% faster than a 4090.
DS supports the 50x0 cards - except in Iray.
Please, it's like having a piano with no strings; sure you can press the keys but you won't produce any music....
Oh my goodness Richard, thank you for the chuckle. It has been a long tough week. I get the point you may be making that DS 4 is supporting RTX 50xx cards, but since Iray and dForce are the main line items in most new products sold by DAZ over the last few years then basically you are saying if you use 3Delight your good to go.
When I read your comment and I thought good for Richard he got a job at the MadMen ad agency. They probably hired him because he is more debonair than that fellow Jon Hamm.
dForce should be ssupported, as far as I know (no 5090 here) - which was one of the aspects I was thinking of when I wrote that DS itself supports the new cards.
Yeah, would be very surprised if dForce isn't supported out-of-the-box. dForce uses a completely different / much more hardware agnostic processing pipeline than Iray (eg. no raytracing cores to contend with.)
And don't forget - Iray rendering will still technically be usable out-of-the-box in a 50XX build (via CPU rendering.) Just not particularly well.
So you would advice people to invest $2000+ in a GPU which makes your dForce simulation perhaps a minute or two faster - and adds litterally DAYS to your rendering?!
Honestly, this sounds like it's a good thing in the mid-term. Maybe not for 50 series owners in this very moment, but if the problem genuinely is unsolvable, it means that getting out at least a beta version of DS 5 has become an extremely high priority for Daz.
(Alternatively, it might be like it was with the early announcement of DS5 where they managed to fix what they initally thought was an unresolvable issue with DS4 and Macs. But getting a longer lifespan for DS4 so that it can continue to be used with newer hardware even after DS 5 comes out wouldn't be a terrible consolation prize if, indeed, DS4 plug-ins may not be able to be ported forwards).
Speakig as a person who is currently staring at a roughly $15,000.00 dual-use 3D rendering/generalized multimedia production PC build (cobbled together over many years, to be fair) as I am writing this, I would advise people to use their hard-earned money as they see fit.
What the actual... are you saying? So, you've spent $15.000 on a PC, congrats? - I assume you're "in the business" - and yet your advice is for people to waste their money (as they see fit) as opposed to advising against spending $2000+ on a GPU not suited for DS until version 5? Unless, ofcourse, you add a 50X0 to your current setup for dForce accelleration... (again, questionable advice unless money is no issue).
Business, as in (freelance) professional multimedia (audio, video and 3DCG - not only 3DCG) production? Yes.
My actual full advice would be that if you stand to make more money over a reasonable span of time from being able to incorporate one or more Blackwell generation GPUs in your workflow(s) than it would cost you to buy, you should make that purchase. Because that is simple logic.
You are assuming that reasonable availability for Blackwell generation GPUs is going to happen before a fully compatible version of Daz Studio becomes available. I would make no such assumptions.
Especially if - like me - you have succumbeed to the professional workstation GPU realm (where form factor and per unit power consumption are technically more important than performance) since the release cadence for those GPUs is 1-2 years after the initial gaming model releases.
I have zero doubts there will be a fully Blackwell compatible release of Daz Studio prior to that.
Just got a swanky system at work that we use for testing moulds, figured I'd give it a shot with Daz to see how it performs. Below are the tests.
System Configuration
System/Motherboard: Dell Precision 7960
CPU: Intel Xeon w9-375X
GPU: nVidia RTX 4000 Ada
System Memory: 384GB ECC @ 4400MT/s
OS Drive: 1TB SSD
Asset Drive: Same
Power Supply:1500W
Operating System: Windows 11 24H2
Nvidia Drivers Version: 32.0.15.7260/3D Visualization/ECC On
Daz Studio Version: 4.23
Benchmark Results
Total Rendering Time: 3 minutes 56.25 seconds
CUDA device 0 (NVIDIA RTX 4000 Ada Generation): 1800 iterations, 0.611s init, 234.317s render
Nvidia Drivers Version: 32.0.15.7260/IRayVR/ECC On
Total Rendering Time: 3 minutes 50.65 seconds
CUDA device 0 (NVIDIA RTX 4000 Ada Generation): 1800 iterations, 0.707s init, 228.132s render
Nvidia Drivers Version: 32.0.15.7260/3D Visualization/ECC off
Total Rendering Time: 2 minutes 58.90 seconds
CUDA device 0 (NVIDIA RTX 4000 Ada Generation): 1800 iterations, 1.094s init, 175.891s render
Nvidia Drivers Version: 32.0.15.7260/IRayVR/ECC off
Total Rendering Time: 2 minutes 58.83 seconds
CUDA device 0 (NVIDIA RTX 4000 Ada Generation): 1800 iterations, 0.667s init, 176.327s render
System Configuration
System/Motherboard: Asus Tuf Gaming Z690-Plus Wifi D4
CPU: i5-13600K
GPU: Nvidia RTX 3090 FE (undervolted to .825V, 1800MHz, max power 90% -MSI Afterburner)
System Memory: 2x32GB Mushkin Redline DDR4-3600 (18-22-22-42)
OS Drive: Solidigm P44 Pro 2TB
Asset Drive: WD Blue SN5000 4TB
Power Supply: SuperFlower Leadex III 850W
Operating System: Win 10 Pro 22H2 19045.5608
Nvidia Drivers Version: 572.60
Daz Studio Version: 4.23.0.1
Benchmark Results
DAZ_STATS
2025-03-15 17:27:12.838 [INFO] :: Total Rendering Time: 1 minutes 52.99 seconds
IRAY_STATS
025-03-15 17:27:32.995 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 0.981s init, 110.610s render
Iteration Rate: (DEVICE_ITERATION_COUNT / DEVICE_RENDER_TIME) iterations per second = 1800 / 110.610s = 16.36
Loading Time: ((TRT_HOURS * 3600 + TRT_MINUTES * 60 + TRT_SECONDS) - DEVICE_RENDER_TIME) seconds = 1*60+53-110.61 = 2.39
Benchmark results of a Nvidia ROG Astral GeForce RT 5090 32GB GDDR7 OC Edition in Daz Studio 2025 - 6.25.2025.11206 (64-bit) Public Build
NVIDIA display driver version: 576.02
Total Rendering Time: 49.48 seconds
CUDA device 0 (NVIDIA GeForce RTX 5090): 1800.0 spp, 460.400 ms init, 48.027 s render
Thanks. This seems to indicate around 33% faster rendering on the 5090 against the 4090. It doesn't differ much from gaming figures between the two cards either. I can see myself upgrading to a 5090 by the end of the year, provided that I can find one on sale
and that the price is not much above the MSRP, of course.
Thanks for that, 4090 ref card numbers, same version of DS and driver.
475.044 ms init, 0 01:01.445 render, 23.728 Msps
System Configuration
System/Motherboard: MSI PRO Z790-P
CPU: Intel Core I9 13900k
GPU: MSI NVIDIA GeForce RTX 5080 16G VENTUS 3X OC
System Memory: CORSAIR VENGEANCE RGB DDR5 RAM 32GB (2x16GB) 6000MHz CL36
Operating System: Windows 11
Nvidia Drivers Version: 572.83
Daz Studio Version: 2025 (6.25) Public Build
Benchmark Results
2025-04-23 09:39:18.322 [INFO] :: Total Rendering Time: 1 minutes 19.41 seconds
2025-04-23 09:39:17.738 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Received update to 01800.00 iterations after 01:17.385.
System Configuration
System/Motherboard: Asrock X570 Taichi
CPU: AMD Ryzen 9 5950X stock
GPU: NVIDIA GeForce RTX 4090 stock
System Memory: G.Skill Ripjaws 64GB (2x32GB, DDR4-3600 CL16-22-22-42)
OS Drive: Silicon Power P34A80 2TB
Asset Drive: Crucial MX500 4TB
Power Supply: MSI MEG Ai1300P PCIE5 (1300W)
Operating System: Windows 11 Pro (v23H2, build 22631.5189)
Nvidia Drivers Version: 572.83 (Studio)
Daz Studio Version: 2025 Alpha
Optix Prime Acceleration: N/A
Benchmark Results
2025-04-24 23:53:10.642 [INFO] :: Finished rendering.
2025-04-24 23:53:10.651 [INFO] :: Total Rendering Time: 1 minutes 58.46 seconds
2025-04-24 23:53:10.185 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4090): 3352.2 spp, 988.881 ms init, 0 01:55.621 render, 23.484 Msps
Iteration Rate: (3352.2 spp / 0 01:55.621 render) 28.9 iterations/second
Loading Time: ((0h * 3600 + 1m * 60 + 58.46 seconds) - DEVICE_RENDER_TIME) seconds
*As you can see, I couldn't calculate the loading time, since I don't know what to use for the DEVICE_RENDER_TIME. Also, I wasn't totally sure about the total iteration count, but I'm pretty sure I used the correct value there, since I was having a look at the iteration count as the render was made.
Right, a new batch of tests, as I'm swapping out my old GTX 1650 secondary/monitor card for an RTX 4060, so I'm doing a new set of benchmarks of everything:
GIGABYTE GTX 1650 OC Low Profile 4GB GDDR5
Note: This test has been reduced to 180 iterations rather than 1800, as I didn't see the sense in running the thing for most of an hour to make the number slightly more precise.
System Configuration
System/Motherboard: ASUS TUF Gaming X570-Plus
CPU: AMD Ryzen 7 5950x @ stock (Disabled for render)
GPUs:
ASUS RTX 4070 Ti Super ProArt OC 16GB @ Stock (Disabled for render)
GIGABYTE GTX 1650 OC Low Profile 4GB GDDR5 @ Stock
System Memory: 2x 32GB Corsair Vengeance DDR4 @ 3600 MHz
OS Drive: Samsung 970 EVO Plus 1TB M.2
Asset Drive: Junctioned - Geometry on 2TB Samsung 870 QVO, Textures on mirrored 2x14TB WD Gold pair
Power Supply: Corsair RM1000x
Operating System: Windows 10 Pro Build 19045
Nvidia Drivers Version: 576.02
Daz Studio Version: 4.24.0.3
Benchmark Results
2025-05-25 14:44:59.637 [INFO] :: Total Rendering Time: 4 minutes 17.18 seconds
2025-05-25 14:49:56.736 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce GTX 1650): 180 iterations, 1.087s init, 254.683s render
Iteration Rate: 0.71 iterations per second
Previous benchmark on 4.12.0.086 Beta (5th Dec 2019): 1161.116 seconds (@1800 samples) for 1.550 iterations/sec
This was an RTX version of Iray, so the performance difference isn't purely down to it being a GTX card. However, there are no other common components between the two systems (this is when I was stuck on a SFF build), so I don't know how much can be concluded from this halving of iteration rate.
~~~~~
ASUS GeForce RTX 4060 LP BRK OC Edition
System Configuration
System/Motherboard: ASUS TUF Gaming X570-Plus
CPU: AMD Ryzen 7 5950x @ stock (Disabled for render)
GPUs:
ASUS RTX 4070 Ti Super ProArt OC 16GB @ Stock (Disabled for render)
ASUS GeForce RTX 4060 LP BRK OC Edition @ Stock
System Memory: 2x 32GB Corsair Vengeance DDR4 @ 3600 MHz
OS Drive: Samsung 970 EVO Plus 1TB M.2
Asset Drive: Junctioned - Geometry on 2TB Samsung 870 QVO, Textures on mirrored 2x14TB WD Gold pair
Power Supply: Corsair RM1000x
Operating System: Windows 10 Pro Build 19045
Nvidia Drivers Version: 576.02
Daz Studio Version: 4.24.0.3
Benchmark Results
2025-05-26 12:18:56.386 [INFO] :: Total Rendering Time: 5 minutes 20.87 seconds
2025-05-26 12:22:04.139 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 4060): 1800 iterations, 0.827s init, 317.965s render
Iteration Rate: 5.66 iterations per second
No previous benchmark.
~~~~~
ASUS RTX 4070 Ti Super ProArt OC 16GB
System Configuration
System/Motherboard: ASUS TUF Gaming X570-Plus
CPU: AMD Ryzen 7 5950x @ stock (Disabled for render)
GPUs:
ASUS RTX 4070 Ti Super ProArt OC 16GB @ Stock
ASUS GeForce RTX 4060 LP BRK OC Edition @ Stock(Disabled for render)
System Memory: 2x 32GB Corsair Vengeance DDR4 @ 3600 MHz
OS Drive: Samsung 970 EVO Plus 1TB M.2
Asset Drive: Junctioned - Geometry on 2TB Samsung 870 QVO, Textures on mirrored 2x14TB WD Gold pair
Power Supply: Corsair RM1000x
Operating System: Windows 10 Pro Build 19045
Nvidia Drivers Version: 576.02
Daz Studio Version: 4.24.0.3
Benchmark Results
2025-05-26 10:15:48.569 [INFO] :: Total Rendering Time: 2 minutes 9.97 seconds
2025-05-26 10:16:00.764 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4070 Ti SUPER): 1800 iterations, 1.064s init, 126.644s render
Iteration Rate: 14.2 iterations per second
Previous benchmark on 4.22.1.58 Beta (1st Feb 2024): 98.888 seconds for 18.2 iterations/sec
There have been some other system configuration changes since (CPU upgraded from 5800X to 5950X, asset drives upgraded), but it seems there's been a notable iteration rate drop between these versions.
~~~~~
Assuming an equal 22% iteration rate drop to what's seen on the 4070TiS, this would suggest that my previous RTX 3060 12 GB would probably now clock in at about 5.2 iterations per second on DS4.24.0.3. This puts the 4060 8GB about 9% ahead of the 3060 12GB in terms of performance, but less so than its nominal compute performance would imply (which is about an 18% boost), so it looks like its narrow 128-bit memory bus compared to the 3060's 192-bit may be holding it back somewhat. (Previous testing of overclocking has shown that increasing memory speeds can result in quite notable performance gains). However, it's worth noting that it's in the secondary GPU slot (limited to 4x rather than 16x) whereas the 3060 wasn't, but I don't believe that's likely to have seriously impacted performance.
The 8GB 4060 is obviously however generally an inferior Iray card to the 12GB 3060 because of its lower VRAM. That said, this was not bought to be used for rendering (although it might see some use for testing in a second DS instance while the main card is working on a render). I have long wanted my secondary/monitor card to be a low profile design to allow good airflow into the main card, and the 4060 represents quite a considerable leap in performance to what's previously been available in half-height versions - there is a low profile 3050 6GB, but prior to that the last Nvidia card that existed in low profile versions was indeed the GTX 1650 4GB. And, as a pre-50-series card, it retains support for some features that Nvidia will be dropping going forwards, such as 32-bit PhysX, meaning that I should have a card that can handle older games and applications for a long time (being broadly equivalent to a 2080 Super), even as I have to think about upgrading the main card.
It does however have the drawback that with a 115 W TDP, it cannot be powered solely by the PCI slot and requires either a 6- or 8-pin (depending on manufacturer) power connection into the end, which may be a considerable issue with some small-form factor builds. (However, not an issue in my spacious case).
Hi
Can I ask, what's the current version of Daz that support most of the feature (let's say VDB and older) and still performant ? What's the performance difference for 4.22, 4.23, 4.24 or 2025 on a card like 4090 and 3090 ? I'm sorry I try to read the first page but I don't think it includes the number for each version of Daz. Maybe someone who has been following this thread can help me? Like is there a version that is slow that I should avoid? And what version is the fastest for rendering?
Thank you very much.
Best regards,
Minh
Even if the number existed, you shouldn't use it for that purpose. The fundamental problem is that iterations are not equal between different generations of Iray.
While you may see that one version has a slower iteration rate than others, this does not inherently mean that version is slower, because it may be that the changes that make an iteration take longer make the result converge proportionally more.
The tests that have been done here don't tell you which version of Iray is the quickest to converge, because the tests here only considered time to run a specific number of iterations, and final image convergence and noise were never factored in.
To meaningfully calculate something like that would first require coming up with a consistent metric for image convergence (you can't assume that Iray's convergence algorithm is the same across versions), and then quantifying results across multiple different scenes and probably graphics cards... and most likely, without going back far enough to completely change the capabilities of Iray, the results would probably come down to fairly small percentage differences.
My general advice is to stick on the newest version unless there's known to be something wrong with it, because the newer versions at least try to fix bugs that were in the old ones (for one example, Iray used to render refraction colours back to front, and thus generated rainbows that were the wrong way around) and Daz won't provide customer support for problems with the older versions.
Thank you very much. I'll stick to the latest version
Mini ITX build, case: Jonsbo Z20
System/Motherboard: Asrock A520m-itx
CPU: AMD RYZEN 7 5700X3D
GPU: Zotac RTX 5090 + UV
System Memory: 64 GB
OS Drive: GIGABYTE 2TB
Operating System: Windows 10 Pro
Nvidia Drivers Version: 576.80
Daz Studio Version: 6.25.2025.17807
UV core 0.890v 2752 Mhz, mem +2000, max power draw: 340W
2025-07-09 13:38:51.290 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics at rendering done:
2025-07-09 13:38:51.290 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 5090): 1800.0 spp, 902.168 ms init, 44.695 s render, 32.621 Msps
2025-07-09 13:38:51.796 [INFO] :: Finished rendering.
2025-07-09 13:38:51.807 [INFO] :: Total Rendering Time: 46.87 seconds
UV core 0.850v 2245 Mhz, max power draw: 260W
2025-07-09 14:10:05.395 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 5090): 1800.0 spp, 961.018 ms init, 51.358 s render, 28.389 Msps
2025-07-09 14:10:05.911 [INFO] :: Finished rendering.
2025-07-09 14:10:05.920 [INFO] :: Total Rendering Time: 53.60 seconds
Quoting this for when someone updates with the 5000 series.
https://www.daz3d.com/forums/discussion/comment/8285481/#Comment_8285481