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
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
If I remember correctly, the 1080ti was twice as fast as a 1060. The 2080ti can do about double what a 1080ti can. So in a very general sense you should see around a 4 to 5X jump in performance speed with a 2080ti.
BUT I will point out that the 3060 can basically match or beat a 2080ti with Iray, and is almost certainly cheaper. In fact my 3060 has a higher iteration rate in this benchmark than the 2080ti that is posted, or even the Titan RTX, which was a fully enabled 2080ti chip. So I would seriously consider the 3060, especially if budget is a concern. The 3060 is not as fast in other applications, though, if that means anything. The 3060 is only faster at rendering. I didn't see any recent benchmarks from the 2080ti here, or even the 1060 for that matter. But like I said, the Titan RTX has been benched, and that card is slightly faster or equal to a 2080ti, and my 3060 beat it.
Which on that note, would you mind running the benchmark scene with your 1060 and posting your result in this thread? That would be much appreciated. If you do get a 2080ti, it would also be nice if you posted a time for it as well. It is always great to see benchmarks for different hardware.
BTW, I have an idea. Maybe one of us can make a chart on an external page that has all the data collected into one thing, and link that to the thread. I think that would help a lot, and it can be controlled easier than trying to update the forum directly.
I still have 4.16.0.3 installed and I've just rerun the benchmark scene compared to the latest Beta, 4.20.0.12. (The GPU is a 3060 12GB, CPU is a 5800X.)
Slower by almost a whole iteration/s!
OHHHH...
I am assuming the pic is the render. Notice how dark the 4.20 version is?
THAT is the issue. There were some changes in how the strength of lighting is handled in 4.20's version of Iray. 4.20 essentially killed what we call "Ghost Lights" because of how it handles opacity of mesh surfaces that emit light. You can find me ranting harshly about this in a couple threads on the subject, along with Daz's anticonsumer policy of not providing downloads for previous versions.
Well that is what happened here. The glowing sphere has an opacity of 0.07. The previous Iray had the sphere output its light at 100% strength. The new Iray uses the opacity setting, and so with only 7% opacity, the light is now only 7% as strong, too.
The patch notes claim that the new Iray fixes a "glitch". But I do not understand how that was ever a glitch there is no real world connection between the opacity of an object and its light output. Regardless we do know one thing about Iray: It hates the dark. I have said that many times, darkness is Iray's enemy that can lead to long renders. I believe that is exactly what we are seeing here, the scene is taking longer because of the ghost light "fix" that Iray made, not because Iray is actually slower.
There is an easy way to test this. Simply change the opacity of the sphere the girl is holding. Bump it up all the way to the max of 1. Now render the scene. Do this in both versions of Daz just to double check.
Then we shall have our answer as to whether the new Iray is truly slower or not. However, this is why I did not upgrade. I use ghost lights, not just the product sold in the store, but I also make them myself and use a lot of mesh lighting. After all, I created the glowing sphere the girl is holding, the scene RayDAnt made is a mix of one I created previously. I have been using these kinds of light for a long time. Daz 4.20 straight breaks many of my scenes that I created, and I do not like that. So I rather prefer not to upgrade until I really have to. Sorry Daz, but you lose. While KA made a script to convert ghost lights to 4.20, it is not perfect and can leave bad artifacts where the light is in camera view.
If I am proven correct, then I suggest everyone change the opacity setting of the glowing sphere to 1 before running their benchmarks. That will level the playing field and make the scene render equally across each version.
Dang it!!! Where I said "2080" I meant 3080! My brain is gone. If I make that same mistake on the order form, I'll shoot myself;-)
I've gone ahead and modified the official benchmark download link at the start of this thread to point to a revision of the scene with opacity on the sphere light source set to 1.0 as suggested. Keep in mind that this makes the last 4 or so pages of benchmarks in this thread (anything posted using Daz Studio version 4.15.1.091 - when the opacity behavior change was first introduced) invalid for comparison purposes going forward.
Using the R4 benchmark, fixed the render visually, no real change in timing.
Just out of interest I rendered R4 in 4.16... it runs faster! The fastest run being 9.51 iterations/s
System Configuration:
Motherboard - Asus X99-E WS
Case - Antec 1200 V3; 5 X 120 mm fans (3 intake, 2 exhaust), 1 X 1200 mm top-mount exhaust
CPU - Intel I7 5930K 3.5 GHz
Memory - 64 GB
OS Drive - Crucial 512 GB SSD (power on hours 62,479 - 7 years 48 days (no, I don't normally shut it down)
Application Drives - 3 Crucial 2 TB SSD drives
PSU - Cooler Master V1200; 1200 W platinum
OS - Windows 7 Pro SP 1
Nvidia Driver version - 471.68
Daz Studio 4.20.0.12 64 bit (Public Beta)
GPU 1 - GTX 980 TI no longer used for rendering, just drives 2 27-inch 1920 X 1280 monitors
GPU 2 - GTX 1080 TI
GPU 3 - RTX 3060
Benchmark results - using current scene file in effect 2022-04-11
1080 TI -
2000-01-01 15:33:21.595 [INFO] :: Total Rendering Time: 7 minutes 21.4 seconds
2000-01-01 15:36:43.261 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce GTX 1080 Ti): 1800 iterations, 2.080s init, 435.562s render
Iteration Rate - 1800/435.6 = 4.13
3060 -
2000-01-01 17:05:18.946 [INFO] :: Total Rendering Time: 4 minutes 26.58 seconds
2000-01-01 17:06:31.538 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3060): 1800 iterations, 2.104s init, 261.147s render
Iteration Rate - 1800/261.1 = 6.89
For my own purpoes I ran one of my longer rendring scenes through comparison tests.
1080 TI - 45 minutes 22.12 seconds; 3060 - 24 minutes 55.83 seconds; Both - 16 minutes 23.89 seconds
Running both cards with GPU load 98% the maximum power draw reported by APC Powerchute monitor was 518 Watts.
System Configuration
Motherboard: GIGABYTE X570 AORUS ELITE
CPU: AMD Ryzen 7 5800X 3.8+ Ghz
GPU: RTX 2080 Ti
System Memory: 32 GB
OS Drive: Samsung SSD 860 EVO 1TB (SATA)
Asset Drive: Samsung SSD 850 EVO 500GB (SATA)
Power Supply: Corsair TX 750 Plus Bronze
Operating System: Windows 10 Professional (21H1)
Nvidia Drivers Version: 512.15
Daz Studio Version: Daz Studio 4.20.0.12 64 bit (Public Beta)
Benchmark Results
2022-04-12 23:52:33.675 [INFO] :: Total Rendering Time: 3 minutes 47.42 seconds
2022-04-12 23:52:44.919 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 2080 Ti): 1800 iterations, 0.838s init, 225.423s render
Iteration Rate - 1800/225.4 = 7.98
Thanks for a 2080ti benchmark! Nobody has benched one in ages. On the positive side this bench of 7.98 iterations is a little faster than the very old numbers we had before. Sadly we do not have a single 2080ti bench with any of the 4.14, 4.15, and 4.16 versions of Daz to compare it to. But the 2080ti is just a little faster than the 3060, at least in 4.20. So in pure speed the 2080ti is faster, but it is still remarkable that the 3060 is so close to it at all, especially when you factor in the prices of them (normal prices that is, but even now the 3060 is dropping below $500, and getting closer to $400 every day).
However these iteration counts are all over the place. That is concerning. None of the recent 3060 times match up with each other, even on the same 4.20 version.
The 3060 is no longer doubling up the 1080ti with 4.20 between two people. Most times with 4.14, 4.15, and 4.16 broke well above 8 iterations, easily doubling the 1080ti's 4 iterations. My 1080ti was hitting 4.6 iterations with 4.15, here namffuak got 4.13 iterations with their 1080ti, almost 0.5 iterations off my 1080ti bench.
My 3060 was running at 8.22 iterations with Daz 4.15. This is nearly identical to prixat's 8.23 with DS 4.16, with only a 0.01 difference, these benches are functionally the same. So that is good, that shows the bench is working. But then prixat scored 7.19 iterations with 4.20, while namffuak dropped all the way to 6.89 iterations with this Daz 4.20. Ouch.
To illustrate this difference over a longer render, just multiply by 10. A 218 second render becomes 2,180 seconds, or just over 36 minutes. A 261 second render becomes 2,610 seconds, which is 43.5 minutes.
That is 7.2 minutes, but I am sure many of you can create a scene that renders much longer than 45 minutes. This time adds up, and especially over cumulative renders day after day. But what is most disappointing here is that the hardware is the same. There really should not be this much of a difference just by upgrading Daz Studio. Considering all the other issues that 4.20 is having (the whole killing ghost lights is just one problem), also losing a decent of performance is yet another blow to 4.20.
Daz really needs to investigate what is going on here. I truly do not believe this was caused by Nvidia. Look, Daz was able to speed up Iray thanks to changing how it handled normal maps. I suspect that something similar is going on here with 4.20, where the render engine is not being allowed to run as efficiently as it should be. And hopefully Daz can figure out what it is. But we can confirm 100% that Daz 4.20 is without a doubt rendering slower than previous versions.
While the app may be more responsive, I don't think that offsets the slower rendering performance. It needs to be addressed. PBR rendering is brutal enough as it is. We should not be regressing on performance.
Didn't Richard say that the 'Caustics' option is being deprecated?
If you compare the 4-20 render to the 4-16 you can see the 4-20 is doing some caustics by default. That could be a source of the slow down.
System Configuration
Motherboard: GIGABYTE X570 AORUS ELITE
CPU: AMD Ryzen 7 5800X 3.8+ Ghz
GPU: RTX 2080 Ti
System Memory: 32 GB
OS Drive: Samsung SSD 860 EVO 1TB (SATA)
Asset Drive: Samsung SSD 850 EVO 500GB (SATA)
Power Supply: Corsair TX 750 Plus Bronze
Operating System: Windows 10 Professional (21H1)
Nvidia Drivers Version: 512.15
Daz Studio Version: Daz Studio 4.15.0.2 General Release
Benchmark Results
2022-04-15 16:13:42.896 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 2080 Ti)
2022-04-15 16:13:44.415 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Received update to 00001 iterations after 1.519s.
2022-04-15 16:17:10.026 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Received update to 01800 iterations after 207.130s.
2022-04-15 16:17:10.030 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Maximum number of samples reached.
2022-04-15 16:17:10.504 Total Rendering Time: 3 minutes 32.54 seconds
Iteration Rate - 1800/207.1 = 8.69
System/Motherboard: SuperMicro X12
CPU: 2x Xeon Gold 6348 @ Stock 3.5 GHZ
GPU: 5x A6000, 4x A5000
System Memory: 512 GB ECC @ 3200 mhz
OS Drive: WD SN850 1 TB NVMe
Asset Drive: 128 GB RAM DRIVE
Operating System: Win 11 Pro
Nvidia Drivers Version: 512.15 DCH
Daz Studio Version: 4.20
PSU: 3x Corsair AX1600
2022-04-16 00:57:45.146 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-16 00:57:45.146 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA RTX A6000): 227 iterations, 2.504s init, 13.245s render
2022-04-16 00:57:45.146 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA RTX A6000): 187 iterations, 3.051s init, 12.461s render
2022-04-16 00:57:45.146 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 2 (NVIDIA RTX A6000): 227 iterations, 2.503s init, 13.005s render
2022-04-16 00:57:45.146 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 3 (NVIDIA RTX A6000): 192 iterations, 3.023s init, 12.160s render
2022-04-16 00:57:45.146 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 4 (NVIDIA RTX A6000): 225 iterations, 2.346s init, 13.249s render
2022-04-16 00:57:45.147 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 5 (NVIDIA RTX A5000): 181 iterations, 2.751s init, 13.162s render
2022-04-16 00:57:45.147 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 6 (NVIDIA RTX A5000): 190 iterations, 2.970s init, 13.329s render
2022-04-16 00:57:45.147 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 7 (NVIDIA RTX A5000): 193 iterations, 2.478s init, 13.046s render
2022-04-16 00:57:45.147 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 8 (NVIDIA RTX A5000): 178 iterations, 2.919s init, 12.315s render
2022-04-16 00:57:45.463 [INFO] :: Finished Rendering
2022-04-16 00:57:45.514 [INFO] :: Total Rendering Time: 19.79 seconds
Loading Time: 6.545
Rendering Performance: 135.9 Iterations Per Second
Running kinda slow with the full GPU stack and the new .duf and Daz 4.20. I will rebench the 3090TI with your new scene at some point.
On another topic (performance related though) I have carefully migrated Daz Studio to RAM Disk & linked it to the Daz install Manager. Assets load fast and I can load/save scenes in just a few seconds. Reminds me of the day I tried out my first SSD. There is no going back. Perhaps 30 times faster in terms of disk access vs the NVMe. As a added plus, each time I save the RAM disk image, it creates an incemental copy of everything, including the CMS, cluster, assets and connect files. Wish I was doing this with 4.16.
Possibly. But turning on caustics talks rendering speeds beyond what 4.20 does. However, I do see the small white spot at the bottom right of the render image in 4.20, and this spot is present when caustics is turned on.
This is 4.15 with caustics ENABLED. I rendered each card alone for comparison. Also as a note I did NOT set the opacity of the sphere to 1 in these tests. This is straight comparison of 4.15 with caustics on and off.
2022-04-16 02:36:40.403 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-16 02:36:40.403 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 3060): 1800 iterations, 1.489s init, 439.800s render
---
2022-04-16 02:42:31.929 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-16 02:42:31.929 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 1.458s init, 182.176s render
-- This next test with the 3090 caustics OFF.
2022-04-16 02:51:52.370 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-16 02:51:52.370 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 1.405s init, 92.394s render
The render times basically double with casutics in 4.15. If 4.20 is turning caustics on by default, then it has to be some kind of simplified version of caustics to only hit performance by 13%. Another thing I noticed is that the light in the room with caustics enabled is noticeably brighter than with it off. More so than in 4.20. And if this is the case, it really invalidates all previous benchmarks compared to 4.20. Obviously you can't compare the different versions head to head anyway, but this punches a big hole in all the old marks.
Here are pics. First is with caustics OFF, which is the default we have been doing for years. I uploaded these to imgur so they are png instead of webp format.
Next is with caustics ON.
Notice that not only is the white spot visible, but the whole scene is brighter from more light and reflections. Again this is 4.15, not 4.20. But we see this spot in the 4.20 pics that were posted, and 4.20 is slightly brighter, though maybe not quite as a difference I see in 4.15.
System Configuration
System/Motherboard: ASUS ROG Strix Z-490-E Gaming
CPU: Intel Core i7-10700K stock
GPU: nVidia GTX 1080 TI stock
System Memory: Corsair Dominator Platinum 4x16GB (64GB total) DDR4 3200
OS Drive: PNY CS1311 960GB SSD
Asset Drive: Samsung SSD 970 EVO 1TB NVME
Power Supply: 750W
Operating System: Windows 10.0.19043 Build 19043
Nvidia Drivers Version: 511.79
Daz Studio Version: 4.20.0.2 Pro Version 64 BITS
Optix Prime Acceleration: NA
Benchmark Results
2022-04-16 13:17:30.868 [INFO] :: Finished Rendering2022-04-16 13:17:30.910 [INFO] :: Total Rendering Time: 7 minutes 49.56 seconds
2022-04-16 13:17:30.379 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Received update to 01800 iterations after 467.110s.
Iteration Rate: 3.8534 iterations per second
Loading Time: 2.45 seconds
Since I've got two cards installed and two versions of DS, I thought it might be worth running the benchmarks again.
System Configuration
System/Motherboard: ASUS X99-S
CPU: Intel i7 5960X @3GHz
System Memory: 32GB KINGSTON HYPER-X PREDATOR QUAD-DDR4
OS Drive: Samsung M.2 SSD 960 EVO 250GB
Asset Drive: 2TB WD CAVIAR BLACK SATA 6 Gb/s, 64MB CACHE (7200rpm)
Operating System: Windows 10 21H2 Build 19044.1645
Nvidia Drivers Version: 511.79
Using the updated scene for the test.
With DS 4.15.0.2:
3090+2080ti
Iteration Rate: 27.705 iterations per second
Loading Time: 6.730 seconds
2080ti:
Iteration Rate: 8.655 iterations per second
Loading Time: 5.460 seconds
3090:
Iteration Rate: 19.315 iterations per second
Loading Time: 5.090 seconds
With DS 4.20.02:
3090+2080ti
Iteration Rate: 24.321 iterations per second
Loading Time: 3.780 seconds
2080ti:
Iteration Rate: 8.072 iterations per second
Loading Time: 3.570 seconds
3090:
Iteration Rate: 16.544 iterations per second
Loading Time: 3.300 seconds
Raw data from logs:
4.15.02
2022-04-20 23:36:02.811 Total Rendering Time: 1 minutes 11.7 seconds
2022-04-20 23:36:02.835 Loaded image r.png
2022-04-20 23:36:02.880 Saved image: E:\Users\Joe\AppData\Roaming\DAZ 3D\Studio4\temp\RenderAlbumTmp\Render 4.jpg
2022-04-20 23:37:04.955 Saved image: F:\Windows Folders\Desktop\3090i+2080ti 2 4_15 Apr 2022.png
2022-04-20 23:37:04.968 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-20 23:37:04.968 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1252 iterations, 2.906s init, 65.382s render
2022-04-20 23:37:04.968 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2080 Ti): 548 iterations, 3.026s init, 65.701s render
2022-04-20 23:30:42.602 Total Rendering Time: 3 minutes 33.43 seconds
2022-04-20 23:30:42.626 Loaded image r.png
2022-04-20 23:30:42.671 Saved image: E:\Users\Joe\AppData\Roaming\DAZ 3D\Studio4\temp\RenderAlbumTmp\Render 2.jpg
2022-04-20 23:30:50.614 Saved image: F:\Windows Folders\Desktop\2080ti 4_15 Apr 2022.png
2022-04-20 23:30:50.627 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-20 23:30:50.627 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2080 Ti): 1800 iterations, 3.161s init, 207.971s render
2022-04-20 23:33:35.102 Total Rendering Time: 1 minutes 38.28 seconds
2022-04-20 23:33:35.126 Loaded image r.png
2022-04-20 23:33:35.172 Saved image: E:\Users\Joe\AppData\Roaming\DAZ 3D\Studio4\temp\RenderAlbumTmp\Render 3.jpg
2022-04-20 23:33:45.607 Saved image: F:\Windows Folders\Desktop\3090i 4_15 Apr 2022.png
2022-04-20 23:33:45.621 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-20 23:33:45.621 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 2.821s init, 93.185s render
4.20.02:
2022-04-20 23:44:18.819 [INFO] :: Total Rendering Time: 1 minutes 17.79 seconds
2022-04-20 23:44:18.843 [INFO] :: Loaded image: r.png
2022-04-20 23:44:18.887 [INFO] :: Saved image: E:\Users\Joe\AppData\Roaming\DAZ 3D\Studio4 Public Build\temp\RenderAlbumTmp\Render 2.jpg
2022-04-20 23:44:26.102 [INFO] :: Saved image: F:\Windows Folders\Desktop\3090+2080ti new april DS4_20_02 2.jpg
2022-04-20 23:44:26.113 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-20 23:44:26.113 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1224 iterations, 1.328s init, 74.182s render
2022-04-20 23:44:26.114 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2080 Ti): 576 iterations, 1.533s init, 74.011s render
022-04-20 23:49:40.413 [INFO] :: Total Rendering Time: 3 minutes 46.55 seconds
2022-04-20 23:49:40.436 [INFO] :: Loaded image: r.png
2022-04-20 23:49:40.481 [INFO] :: Saved image: E:\Users\Joe\AppData\Roaming\DAZ 3D\Studio4 Public Build\temp\RenderAlbumTmp\Render 3.jpg
2022-04-20 23:49:54.557 [INFO] :: Saved image: F:\Windows Folders\Desktop\2080ti new april DS4_20_02.jpg
2022-04-20 23:49:54.572 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-20 23:49:54.572 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2080 Ti): 1800 iterations, 1.500s init, 222.981s render
2022-04-20 23:52:28.419 [INFO] :: Total Rendering Time: 1 minutes 52.10 seconds
2022-04-20 23:52:28.442 [INFO] :: Loaded image: r.png
2022-04-20 23:52:28.488 [INFO] :: Saved image: E:\Users\Joe\AppData\Roaming\DAZ 3D\Studio4 Public Build\temp\RenderAlbumTmp\Render 4.jpg
2022-04-20 23:53:10.053 [INFO] :: Saved image: F:\Windows Folders\Desktop\3090 new april DS4_20_02.jpg
2022-04-20 23:53:10.066 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-04-20 23:53:10.066 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 1.237s init, 108.804s render
Very interesting, the 3090 lost proportionally more speed than the 2080ti did. The 3090 lost 14% while the 2080ti only lost 6%. That is very discomforting if this is a trend across other set ups. Just looking at the time, the 3090 and 2080ti both lost about 15 seconds...which is very strange. This is not making sense.
I do not have 4.20, so I have to ask this question. Can caustics be disabled in 4.20?
And if so, does the render image and performance return to pre 4.20 levels?
Well I've started doing digital art again and need to rebenchmark my 'puter. Where is the download for the current benchmark scene located? I also recently updatred to 4.20.0.17, Anything I need to know about settings? Here are my current specs:
System/Motherboard: Gigabyte X570 Aorus PRO WIFI
CPU: AMD Ryzen 7 3700X @ 3.6 (boosts to 4.3)
GPU: EVGA GTX 1070 GAMING @ 1506 MHz
System Memory: G.Skill 32 GB Trident Z Neo DDR-4 3200 memory @ SPEED
OS Drive: Samsung 860 EVO 1TB SSD
Asset Drive: 2 x Western Digital Blue WD20EZRZ 2TB HDD in mirrored array
Operating System: Windows 10 Professional 64 21H1
Nvidia Drivers Version: 456.38 Studio Driver
Daz Studio Version: 4.20.0.17
Optix Prime Acceleration: N/A
After I complete the benchmark with the GTX 1070, will be repalcing it with my brand new RTX 3060 12GB
The bench file is in the first post as always. Don't modify any settings.
If you don't mind, can you verify if there are any Iray settings for caustics? Previous versions have caustics OFF by default, and that has been the setting in the benchmark since the beginning. I want to know if the new 4.20 is changing caustics to always be ON, which is leading to the increase in render time with 4.20.
If there is a caustic setting in 4.20, you could run the bench with both settings to see what the difference is. I would assume that disabling caustics in 4.20 (if possible) would give us the same performance as 4.14 to 4.16. But I am not downloading 4.20 to find out.
I think you will enjoy that 3060 quite a bit.
Where do I find the caustics setting at? It has been a whiel since I opened DS and am a little fuzzy on some things.
4.20 may not even have it anymore. But it would be in the Render Settings, under the Optimization tab for previous versions of Daz. It is the same tab where Instancing Optimization and Max Path Length are located. If the setting is not there, then 4.20 has indeed removed this setting. I sure hope not, because I am not a fan of removing consumer choice.
There is a caustics sampler boolean switch in 4.20.0.12 and seems to be off by default
System Configuration
System/Motherboard: ASUS TUF GAMING X570-PLUS
CPU: AMD Ryzen 9 3900X @ 4650
GPU: GPU1 MSI GeForce RTX 2080 Super GPU2 MSI RTX 3070Ti
System Memory: G.Skill 64 GB DDR4 @ 1802 MHz
OS Drive: Samsung 960 EVO NVME
Asset Drive: Seagate Barracuda Internal Hard Drive 8TB
Power Supply: EVGA Supernova 1300 G2, 80+ Gold 1300W
Operating System: Microsoft Windows 11 Enterprise (x64) Build 22000.593
Nvidia Drivers Version: 472.12
Daz Studio Version: 4.20.0.12 64bit
Benchmark Results CPU + GPU All
Optix Prime Acceleration: NA
Daz__stats Total Rendering Time: 1 minutes 51.5 seconds
IRAY_STATS
CUDA device 1 (NVIDIA GeForce RTX 2080 SUPER): 630 iterations, 1.126s init, 108.604s render
CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 1070 iterations, 0.839s init, 108.522s render
CPU: AMD Ryzen 9 3900X: 100 iterations, 0.653s init, 108.463s render
Iteration Rate:
CUDA device 1 (NVIDIA GeForce RTX 2080 SUPER): 3.314
CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 9.8597
CPU: AMD Ryzen 9 3900X: 0.9219
Total 16.5739
Loading Time: 3 seconds
Benchmark Results 2080 + 3070Ti
DAZ_STATS Total Rendering Time: 1 minutes 47.44 seconds
IRAY_STATS
CUDA device 1 (NVIDIA GeForce RTX 2080 SUPER): 643 iterations, 1.064s init, 104.021s render
CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 1157 iterations, 0.812s init, 104.955s render
Rendering Performance: [DEVICE_ITERATION_COUNT(sum of all values) / DEVICE_RENDER_TIME_(largest value)] iterations per second
CUDA device 1 (NVIDIA GeForce RTX 2080 SUPER): 6.1814
CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 11.0195
Total 17.1502
Loading Time: 3 seconds
Benchmark Results 3070Ti Only
DAZ_STATS Total Rendering Time: 2 minutes 43.32 seconds
IRAY_STATS CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 1800 iterations, 0.845s init, 161.228s render
Rendering Performance: 11.1643 iterations per second
Loading Time: 2 seconds
Thanks to doing the benchmark I realized I never uncheck my CPU from render device, so my renders were bein slowed lol
Thanks for looking. This could be a situation where the switch is in Daz 4.20, but has no actual effect, a dummy switch. The switch may be "ON" even though it says off. Similar to when OptiX Prime Acceleration was phased out, the toggle was still present, but had no effect.
If you don't mind, you could try to toggle it to see if it impacts perrformance. If it doesn't, then caustics are probably enabled by default. If it does hit performance, look to see if the render image is different.
System Configuration
System/Motherboard: ASUS TUF GAMING X570-PLUS
CPU: AMD Ryzen 9 3900X @ 4650
GPU: GPU1 MSI GeForce RTX 2080 Super GPU2 MSI RTX 3070Ti
System Memory: G.Skill 64 GB DDR4 @ 1802 MHz
OS Drive: Samsung 960 EVO NVME
Asset Drive: Seagate Barracuda Internal Hard Drive 8TB
Power Supply: EVGA Supernova 1300 G2, 80+ Gold 1300W
Operating System: Microsoft Windows 11 Enterprise (x64) Build 22000.593
Nvidia Drivers Version: 472.12
Daz Studio Version: 4.15.0.30 64bit
Benchmark Results Both GPU
Optix Prime Acceleration: NA
Daz__stats Total Rendering Time: 1 minutes 35.72 seconds
IRAY_STATS
CUDA device 1 (NVIDIA GeForce RTX 2080 SUPER): 629 iterations, 1.821s init, 92.416s render
CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 1171 iterations, 1.602s init, 92.414s render
Iteration Rate:
CUDA device 1 (NVIDIA GeForce RTX 2080 SUPER): 6.8061
CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 12.6712
Total 19.4771
Loading Time: 3 seconds
Benchmark Results Both 3070Ti
Optix Prime Acceleration: NA
Daz__stats Total Rendering Time: 2 minutes 21.74 seconds
IRAY_STATS
CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 1800 iterations, 1.497s init, 138.945s render
Iteration Rate:
CUDA device 0 (NVIDIA GeForce RTX 3070 Ti): 12.9547
Loading Time: 3 seconds
Decided to test on 4.15 too for compare. Will try the switch in 4.20 now see what happens. I had to delete the girl, because I deleted the clothes before reading this and she is coming in naked lol. Switch off 1 minutes 35.83 seconds, caustics on 2 minutes 54.94 seconds. See a bit of difference in the lower right corner, and on the blue ball reflection with caustics on. Guess that rules out a false switch at least.
The switch works correctly in Daz 4.15. This we know. But 4.20 has changed something that is making render times increase, and notes say something about the option for caustics being depreciated.
Since you ran the test without the girl, may as well do the same in 4.20.
BTW, I suggest installing the 4.20 BETA instead of over writing your 4.15 install. Just in case you decide you don't like the performance loss, the change that breaks ghost lights, and other joyful things that 4.20 brought upon us.
System Configuration
System/Motherboard: ASUS H170 Pro Gaming
CPU: Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz
GPU: Nvidia GeForce GTX1070 @ 1,506 Mhz
System Memory: 16 GB 2400 Mhz DDR4 (MB only supports 2133 Mhz)
OS Drive: ScanDisk SSD Plus 480GB
Asset Drive: Same
Operating System: Microsoft Windows 10 Home (x64) Build 19043.1682
Nvidia Drivers Version: Studio 512.59
Daz Studio Version: 4.20.0.17
Optix Prime Acceleration: N/A
Benchmark Results
Total Rendering Time: 10 minutes 46.34 seconds
CUDA device 0 (NVIDIA GeForce GTX 1070): 1800 iterations, 1.681s init, 642.188s render
Iteration Rate: 2.80 iterations per second
Loading Time: 4.152 seconds
Well this is interesting, look at the difference in render time if I do the benchmark in Daz 4.15.
About 42 seconds faster overall.
System Configuration
System/Motherboard: ASUS H170 Pro Gaming
CPU: Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz
GPU: Nvidia GeForce GTX1070 @ 1,506 Mhz
System Memory: 16 GB 2400 Mhz DDR4 (MB only supports 2133 Mhz)
OS Drive: ScanDisk SSD Plus 480GB
Asset Drive: Same
Operating System: Microsoft Windows 10 Home (x64) Build 19043.1682
Nvidia Drivers Version: Studio 512.59
Daz Studio Version: 4.15.0.30
Optix Prime Acceleration: N/A
Benchmark Results
Total Rendering Time: 10 minutes 3.62 seconds
CUDA device 0 (NVIDIA GeForce GTX 1070): 1800 iterations, 4.172s init, 596.281s render
Iteration Rate: 3.02 iterations per second
Loading Time: 7.339 seconds
Oh sorry, my comment was confusing. The caustics switch test was in 4.20.0.12, just the render stats I posted was from 4.15