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
Compared to a 5700g it wouldn't make any difference. The 5800x3d has better single core, which could help Daz Studio itself run slightly better, but this would be so slight you might never notice. As for rendering, in other rendering application the 5800x3d actually performs worse than the 5800x. Look up content creation focused reviews for the 5800x3d and you will see performance drops a surprising amount for rendering tasks like Blender Cycles and Vray. Video games and single threaded tasks were the main beneficiaries of the 5800x3d's extra cache.
I wasn't talking about using it as a render device. I always keep CPU disabled for renders, I only used it for the benchmark.
What I meant is that it could possibly help with all the translation that has to happen between DS, Wine and Linux. It's a pretty CPU and memory intensive process, and I think that's the 5700G's weak point, since it has less pci-e bandwidth and less L3 cache than most of the Zen 3 line. But on searching for 5800X3D Wine performance I don't get anything, so I guess cache isn't much of an issue.
I guess if I really wanted to test the Wine performance hit I could install Windows and test on the same hardware, but I'm probably not doing that anytime soon.
The 5700g is not a piece of junk. If you were coming from something significantly older and weaker then maybe. But this upgrade is such a side upgrade I think you would be cheating yourself. It doesn't make sense. Few tasks can take full advantage of pcie, and if it mattered to Linux then somebody would be talking about it.
You also need to remember that Daz 4.0 released in 2011. Obviously it has been updated a lot over the decade, but the underlaying nuts and bolts are indeed a decade old. DS can lag at basic tasks with even the very fastest CPUs on the planet...I wouldn't blame any lag you may have so quickly on your translation layer. It is simply DS being DS. A CPU with good single core performance can help, but not cure the issue.
This has always been the talk of the town because Daz generally does not benefit from the CPU during renders. As outrider42 mentions, there are always things that can be done to improve render speeds via tweaking. These old school methods include downsizing textures, reducing vertices, capitulating denoiser functionality and recomposing the scene. While most users cannot step back to 4.16, we can all advance forward to the beta (4.21.1.13), while simultaneously retaining CMS libraries, directory structures and perpetual access to the release version of Daz (4.21.0.5).
In short, machines with top end processors may be able to recoup a percentage of the recently lost performance by using the current beta w/CPU enabled.
Big news, Iray 2022.1.0 is now out of beta and they even directly state the performance regression has been fixed for Daz Studio! I think this is the first time they have actually acknowledged this publicly, conveniently at the same time they claim to have fixed it. But hey, if they have indeed fixed, props to them for figuring it out and getting it done. This statement also finally puts the bed the idea this was done intentionally in the name of accuracy or something like that. I would love to hear what the problem actually was. But remember, we still have to wait for this to get added to Daz Studio. That may take some time.
But that is not all, how about "customizable ghostlight support"? When you look at this it really does seem like the Iray team took a look at our suggestions on their contact page or maybe even here. They even call them ghostlights. I do not believe this is an industry standard term.
There are several other bits and pieces here which could improve performance in other ways.
Here is what the dev team posted on their site:
Iray 2022.1.0 final released
New features:
Multi matte output buffer
Customizable ghostlight support (Hell yes)
LPE support for emissive volumes
Important improvements:
Deterministic guided sampling
Fixed some performance regressions (especially one that affected the Daz benchmark scene)
Improved hair BSDF sampling
Most important things already found in the Iray 2022.1.0 beta:
Spectral sky model
Guided sampling caching on camera movement
Guided sampling for the caustic sampler
Spheres/Particle support
Incremental material updates / less compilation overhead
Fully dynamic evaluation of environments (e.g. Sun&Sky, IBL)
Reduce VDB memory usage by a lot
Full OpenImageIO support
Improved guided sampling convergence + add support for volumes
Improved hair/fiber rendering performance on RTX capable cards
Daz developers have already updated to Iray 2022.1.0 beta in the private build channel fwiw.
Yes, but it looks like they saved the best for last if this all pans out. Pretty much all the new items in the Final version are exciting. Multimat could be really cool for people who like to composite images. Light Path Expressions might help make volumetric effects truly pop. Real official ghostlights are exactly what the doctor ordered. Fixing this performance regression that has vexed us for a while now is huge. Improved BSDF could make mesh hair more realistic. The guided sampling stuff can be interesting, too, as they are trying really hard to make caustics less resource expensive (they have been for a while, but maybe it is finally paying off now).
These sound like real upgrades, while the beta has nice improvements, they are more iterative. It will be nice to see to much faster the fiber/hair rendering is on RTX cards after this update. Like how fast is it in 4.16, how fast is it right now, and how fast is it in the future version that also gets our performance back. There are 3 different variables there.
The version before the regression. The version with the regression but has faster fiber rendering. The version that reverses the regression AND has the faster fiber rendering. That would be fun to see.
It may be time to break out the old "Show me the power" thread that had the lady with massive strand hair. Also intrigued by the "spheres/particle support" in the beta.
or the Daz developers made the case for ghostlights, and possibly explained the issue with the ebnchmark scene too.
Well, since Daz doesn't give us much transparancy this is really just a statement of speculation. At no point in time has Daz made a single official statement about these matters. They wouldn't even acknowledge there was any problem in the first place, so how can we assume there was any such discussion between these parties?
It also bears repeating that the benchmark scene was not the only thing rendering slower. If that had been the case there would not have been such an outcry. Practically everything I tested was slower in 4.20, not just the bench scene. Some scenes rendered drastically slower, even reaching double render times in the worst cases.
Any any case, even if Daz allegedly spoke to the Iray dev team, it would have been because of the complaints from its users prompting them to take action. If nobody had complained about ghost lights being totally changed, then they would never have been added as a new feature to Iray.
But I do hope Daz has people in constant contact with Iray to discuss its direction and upcoming features. They need to make sure the dev team is extremely aware that making big changes to how the renderer works can impact user workflows in negative ways. Accuracy is great, that is why people use Iray, but it still needs to have flexibility to allow people create fantasy. Hollywood lighting is mostly fantasy, even in the most down to earth movies use lighting that is impossible in real life. Iray still needs to be able to do that.
I must agree on that, and you can only collect so much data in 2 min @ 900:900 pixels. Most of my renders now are 4K with 1080p/720p for draft/animations only. The performance drop on larger scenes is probably closer to 50-60%.
Glad they are working on fixing it, but people would be MUCH happier if we could all just revert to 4.16 in the interim. It feels like we all got forced into Window 8 with no option to go back to Window 7.
I will keep my mouse over the download button for the update. T-minus...?
excerpt from DAZ changelog to show Iray 2022.1.0 with ghost light changes coming soon to DS (for those that don't follow changelog):
(may actually finally update from last 4.16 beta.)
Update to NVIDIA Iray 2022.1.0 (363600.1229)
Minimum driver is 526.67 (R525) on Windows
Minimum driver is 526.98 (R525) on Windows in order to retain OpenCL/dForce compatibility
Added support for controlling “Iray Ghost Light Factor” when NVIDIA Iray is the active renderer
Experimental
If a DzFloatProperty (user) property named “Iray Ghost Light Factor”, with the property group path of “Display/Rendering/Iray” and a default value of 1.0 is added to a node with geometry that is intended to act as a light source, the value of this property is used to control visibility in glossy interactions
See official documentation for more information
DAZ Studio : Incremented build number to 4.21.1.21
Do keep in mind that even with these big fixes that the current Iray is still different from 4.16 and previous. There are other changes that can impact how your render looks. Thin Film has been changed, and in some instances surfaces with thin film enabled may look darker or have a gray color to them. This is most visible with surfaces that try to replicate water. It depends on your scene and lighting, but I have observed tear props turning grayish or even black like they are getting mixed with black makeup. This is not to be confused with the old dark eye bug, though it can produce a similar result in some cases given that eyes often use thin film.
There are also other changes to environment lighting that can effect HDRI tone. And while they are bring back ghost lights, we still have to deal with figuring out how to replicate how ghost lights looked in 4.16 and prior. I doubt they will have the same tone, and it may be hard to replicate the look of a scene from 4.16 that contained ghost lights.
Of course I am still happy to see them return, but we still have to deal with replicating the tone of our old scenes from 4.16. So if you have any on going productions that depend on your renders keeping the same tone across a story, you will want to keep 4.16 for a while yet. I recommend everybody to install the beta first before updating Daz. I have kept 4.16 for quite a while now, and I only update the beta branch. I also keep my 4.16 installer so I can install it if I ever need it. Though people who have 4000 series GPUs cannot use 4.16 at all, so for those people this is not an option.
I've never posted on the forums before, but I recently built a new rig with a RTX4080 and figured that posting my benchmark results would be a good start:
System/Motherboard: ASROCK Z790 PG Lightning
CPU: Intel Core i7 13700k @ stock
GPU: Nvidia RTX4080 @ stock
System Memory: Corsair Vengeance 64GB (2x32GB) DDR5 @ 5600
OS Drive: WD Black SN850x, 1TB
Asset Drive: SAMSUNG 870 QVO SSD, 2TB
Power Supply: EVGA SuperNova G2 850W
Operating System: Windows 11 22H2, 22621.819
Nvidia Drivers Version: Nvidia Studio Driver v527.56
Daz Studio Version: 4.21.05, 64-bit
Benchmark Results:
Total Rendering Time: 1 minutes 55.28 seconds
IRAY_STATS
Iteration Rate: 16.565 iterations per second
Loading Time: 6.619 seconds
Welcome to the forums and thank you for sharing this! Your device time was 108.66 seconds. As you noted, this is at base clock speeds. Your benchmark result here gives us a perfect baseline for analysis.
This score would put the 4080 at about 73% of the 4090 in terms of iterations per second (16 vs 22 from prior benchmarks with similar drivers and Daz version), but it pushes past all the top tier cards of the prior generation, including 3090. At least in terms of raw speed. The 4080 has a boatload of new tech onboard too. My prediction on this GPU is that it has broader overlocking potential compared to a typical 4090. I have a feeling that with a push in the right direction, these cards will get very close to 20 IPS. Or even faster, especially once the IRAY fix drops.
However it is pretty insane that the 4080 is still slower than a 3090, when the 3090 is running 4.16 or older. Of course the 4080 will not work with 4.16, but it is just another demonstration of how poor Iray in 4.20 and 4.21 is.
Hopefully the new Iray will fix this. The new Iray 2022 final made it to the Daz Private Beta branch. As the name suggests this branch is not the public beta so we normal users cannot play with it just yet. But this means it should be coming soon, and we will finally be able to see just how well the 4000 series performs with Iray.
We will have to do all benchmarks again when this version drops.
What makes you say that 4000 cards don't work with Daz 4.16?
I've tested and it works just fine. And as you say everything is faster (loading time, etc.) in 4.16 as compared to 4.21.
System/Motherboard: Gigabyte Aorus Master X570
CPU: AMD 5950X @ stock
GPU: MSI RTX 4090 Suprim Liquid X + Asus RTX 3090 Rog Strix @ stock
System Memory: G.skill TridentZ 64GB (4x16GB) DDR4 3600
OS Drive: Sabrent 2TB Rocket 4 Plus NVMe Gen4
Asset Drive: Crucial MX500 4TB SATA
Power Supply: EVGA SuperNova T2 1600W
Operating System: Windows 11 22H2, 22621.963
Nvidia Drivers Version: Nvidia Game Ready Driver v527.56
Daz Studio Version(s): 4.20 (*and 4.21.1)
3090 vs 4090 with all else held equal, pending changes for 4000 series iray support. There's a lot of variance from different iray versions on differently configured pcs, sometimes preloading the scene in iray viewport, making it difficult to get a good A/B comparison. Scene preloaded for the following:
4.20
2022-12-18 00:33:54.275 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4090)
2022-12-18 00:35:03.099 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Received update to 01800 iterations after 68.824s.
2022-12-18 00:36:29.638 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 3090)
2022-12-18 00:38:12.487 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Received update to 01800 iterations after 102.849s.
both
2022-12-18 00:31:25.186 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 3090)
2022-12-18 00:31:25.186 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4090)
2022-12-18 00:32:07.795 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Received update to 01800 iterations after 42.609s.
4.21.1 was a bit worse of course
2022-12-18 00:41:07.722 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 3090)
2022-12-18 00:41:07.722 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4090)
2022-12-18 00:41:58.632 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Received update to 01800 iterations after 50.909s.
I've recently swapped a watercooled 3090 for a watercooled 4090.
I don't think there's that much more to come in terms of performance. The Iray devs said up to 2x in their blog. I'm seeing about a 70% uplift at the moment.
What is worthwhile is a lower power draw for similar workloads and ltherefore lower temps and noise. I'm in the high 40s core,hotspot and vram on the 4090.
The 3090 did mid 50s, mid 60s and high 70s/low 80s on the same fan profile and rads.
What you have here is a cool comparison. Trouble is the iteration count per GPU is missing. You should have this info from your log file, something like this:
2022-12-12 20:07:00.090 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-12-12 20:07:00.091 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA RTX A6000): 1792 iterations, 4.886s init, 1409.169s render
2022-12-12 20:07:00.091 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CPU: 208 iterations, 3.825s init, 1409.126s render
This would provide a measure of contribution each GPU is doing during the render. If the preload to IRAY happens (which does indeed spoil the results) perhaps try to open Daz first and set to texture view, then open the benchmark scene after the texture view setting is applied.
A couple things:
The whole "Iray preloading" thing is not actually a thing - at least not in the way people generally seem to think. Daz Studio's implementation of Iray (up to this point) does not, and I repeat, does not reuse texture/geometry/etc data loaded into VRAM from already active Iray liveview/render activities. Every single time you initiate a new Iray render in Daz Studio, all of the scene's data gets re-transmitted via the PCI-E bus to each GPU being utilized. The real reason why "preolading a scene into Iray" seemes to make a rendering performance difference (in the pre-SSD past especially) is due to hard disk access times.
Oftentimes, the process of going from previewing/manipulating a scene in Daz Studio to starting an Iray render of it results in a gap of direct disk access to the drive where the scene's 3D assets are located that is lengthy enough to trigger the typical Windows installations's power saving features regarding storage devices. Resulting in as much as a 6 second increase (iirc what I was seeing when initially doing research into this a good 3-4 years ago now) in rendering times. Which in a scenario (like this thread) where a benchmark is easily taking less than a minute to complete on llatest gen hardware, can seem to have a significant impact on measured rendering performance when going by the statistical data that Daz Studio/Iray spit out during/after the rendering process.
However, this is only the case if you are using (objectively speaking) the wrong rendering-related statistic reported by Daz Studio/Iray. Here's an excerpted logfile for a render of the benchmarking scene I just did on the latest stable version of Daz Studiio (4.21.0.005) after letting DS sit idle for ten minutes after initially loading the scene:
2022-12-19 23:20:22.475 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Maximum number of samples reached.
...
2022-12-19 23:20:22.973 [INFO] :: Finished Rendering
2022-12-19 23:20:23.007 [INFO] :: Total Rendering Time: 2 minutes 20.1 seconds
...
2022-12-19 23:20:26.559 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-12-19 23:20:26.559 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA RTX A5000): 1800 iterations, 1.330s init, 136.599s render
Here's the same again for another render of the benchmarking scene initiated immediately after the previous render had completed (ie. with the benchmarking scene having been "pre-loaded" into Iray):
2022-12-19 23:22:47.616 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Maximum number of samples reached.
...
2022-12-19 23:22:48.113 [INFO] :: Finished Rendering
2022-12-19 23:22:48.144 [INFO] :: Total Rendering Time: 2 minutes 19.9 seconds
...
2022-12-19 23:22:51.835 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-12-19 23:22:51.835 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA RTX A5000): 1800 iterations, 1.076s init, 136.596s render
Keep in mind that my system is using a very fast external SSD for asset storage (meaning that the variances observed here are going to be very small.) Notice how the values marked in red and yellow both shrink signficantly on a millesecond scale (100-200ms faster) going from the first run to the second, whereas the value marked in green stays virtually identical (3ms variance.) This is because the first two values are calculated internally by Daz Studio and Iray with a timer 0 starting point that takes place while the system is still working on initiating the hardware side of the rendering process. Whereas the third value only includes the time spent by that piece of hardware rendering. Meaning that it is a consistent, reliable measure of that one piece of information. Hence why this thread's instructions put such a heavy emphasis on using the green value rather than the other two. Because this is a rendering hardware benchmarking thread.
Something else to keep in mind is that multi-GPU benchmarks (ie. benchmark runs completed with more than just a single GPU/CPU device checked under advanced rendering settings) while useful for judging your system's real-world performance in actual use-case scenarios, isn't all that useful for making universal comparisons betweenyours/others' systems since it introduces too many potentially confounding variables into the equation.
Yeah, the "preload" only helps with the loading time, which can impact "total rendering time". But the render time that is reported in seconds has always been the true render time.
That is why it is important for people to report the number that raydiant highlighted in green. The total render time factors in loading the data into the GPU, and so is not the number we really want to see. Total render time can be useful, but it isn't nearly as important in the scope of things.
System/Motherboard: MSI B450 Tomahawk Max II
CPU: AMD Ryzen 2600 @ stock
GPU: MSI Suprium X RTX 3090 ti
System Memory: Crucial 64GB (2x32GB) DDR4 3600
OS Drive: Samsung M.2 970 Evo
Asset Drive: Same
Power Supply: Seasonic Prime GTX 1000 80 Plus Gold
Operating System: Windows 10 Build 19045
Nvidia Drivers Version: Nvidia Studio Driver v527.56
Daz Studio Version(s): 4.16.0.3
2022-12-21 09:41:06.205 Total Rendering Time: 1 minutes 34.53 seconds (true render time I'm sure is lower, but I'm running Daz now and can't look)
I'm bascially posting this to let people know the new studio driver works for 3090ti's and does not affect render times if you're still on 4.16 and want ghost lights.
This is the first driver I've tried since sticking with 512.96 because a friend told me it played well with 4.16 and there is no artifacting (line across the screen etc...)
Hey guys, just thought would mention did a test -- 4.16.1.43 vs just released 4.21.1.26. Posted it at DAZ beta thread as use my own scene (time thing) , so not really fitting here. But still the speed/mem footprint may be of interest to you guys.
Haven't done this myself in quite a while...
System Configuration
System/Motherboard: Gigabyte Z370 Aorus Gaming 7
CPU: Intel 8700K @ 4.7Ghz all-core (iGPU used for display)
GPU 0: Nvidia Titan RTX @ stock (custom watercooled)
GPU 1: PNY RTX A5000 #1 @ stock (custom watercooled)
GPU 2: PNY RTX A5000 #2 @ stock (custom watercooled)
System Memory: Corsair LPX 64GB (16x4) DDR4 @ 3000Mhz
OS Drive: Samsung 980 Pro 2TB
Asset Drive: Sandisk Extreme Pro v2 4TB
Power Supply: Corsair AX1500i 1500 watts
Operating System: Windows 10 Pro 22H2 19045.2364
Nvidia Drivers Version: 527.56 SD
Benchmark Results
Daz Studio Version: 4.21.1.013 Beta x64
2022-12-23 03:15:26.587 [INFO] :: Total Rendering Time: 2 minutes 22.39 seconds
2022-12-23 03:17:11.715 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA RTX A5000): 1800 iterations, 1.297s init, 138.998s render
Iteration Rate: 12.950 iterations per second
Loading Time: 3.39 seconds
Daz Studio Version: 4.21.1.026 Beta x64
2022-12-23 03:21:54.061 [INFO] :: Total Rendering Time: 2 minutes 8.70 seconds
2022-12-23 03:23:01.594 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA RTX A5000): 1800 iterations, 1.268s init, 125.350s render
Iteration Rate: 14.360 iterations per second
Loading Time: 3.35 seconds
So that works out to be a performance uplift of 10.9%
Pulling from the Iray developer info officially reposted here on the forums:
Iray 2022.1.0, build 363600.1229
Fixed Bugs
Seems to have been the root culprit all along.
Lastly, a quick Titan RTX run for the sake of completeness.
Daz Studio Version: 4.21.1.026 Beta x64
2022-12-23 03:51:54.239 [INFO] :: Total Rendering Time: 3 minutes 57.40 seconds
2022-12-23 03:53:18.737 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 2 (NVIDIA TITAN RTX): 1800 iterations, 9.371s init, 225.940s render
Iteration Rate: 7.967 iterations per second
Loading Time: 11.46 seconds
System Configuration
System/Motherboard: MSI MPG x570
CPU: AMD 5800x
GPU 0: EVGA 3060 Black @ stock
GPU 1: Nvidia Founder's 3090 @ stock
System Memory: GSkill 64GB (16x4) 3200Mhz
OS Drive: Inland M.2 2TB
Asset Drive: Samsung 870 EVO 4TB
Power Supply: EVGA 1000 GQ
Operating System: Windows 10 Pro 22H2 19045.2364
Nvidia Drivers Version: 527.56 SD
Benchmark Results
Daz Studio Version: 4.21.1.026 Beta x64
3090+3060
2022-12-27 21:22:02.313 [INFO] :: Total Rendering Time: 1 minutes 21.78 seconds
2022-12-27 21:22:19.260 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 3060): 502 iterations, 1.038s init, 78.865s render
2022-12-27 21:22:19.260 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1298 iterations, 0.940s init, 78.570s render
Iteration Rate: 22.824 per second
Loading Time: 2.915 seconds
3090 alone
2022-12-27 21:25:24.608 [INFO] :: Total Rendering Time: 1 minutes 49.90 seconds
2022-12-27 21:25:35.765 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 0.932s init, 107.757s render
Iteration Rate: 16.704 per second
Loading Time: 2.143 seconds
3060 alone
2022-12-27 21:31:01.549 [INFO] :: Total Rendering Time: 4 minutes 34.57 seconds
2022-12-27 21:34:21.293 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 3060): 1800 iterations, 0.690s init, 272.670s render
Iteration Rate: 6.601 per second
Loading Time: 1.9 seconds
So...I am not seeing any difference. I updated my driver and restarted my PC to double check it took. I verified that my Iray version is build 363600.1657, 14 Dec 2022.
I was not sure what is going on there, but then I compared my 4.21 time and found that WAS slower. I forgot that 4.21 was even slower than 4.20, yikes. My times in 4.21 were at 88 seconds for both cards, compared to 78 here. But I was getting about 78 in 4.20, so this improvement only gets us back to 4.20, NOT 4.16. My times in 4.16 are hitting 66 seconds. 4.20 was when the complaints started happening. So the newest Daz slots between. I will try some other tests, but not right now, I just don't feel that motivation. IMO it seams like the Iray Dev Team stretched the truth a bit. They may have improved the speed, but their claim they "fixed" a regression is not accurate if you consider 4.16, and that is the speed I want to see again.
We also have reports that some EXR based HDRIs are not working in this new version. I have not tested this.
Re-check your installed Daz Studio Beta version. The latest and only release so far to use Iray 2022.1.1 or newer (where the "fixes" happened) is DS 4.21.026 - you listed .013 above. Which is back on Iray 2022.0.1
Wow.. .this is a great thread, and I am learnig so much from it. I appreciate the respectful "back and forth" of the occasional different ideas that pop up, so I wanna thank all involved!
That was a copy-paste mistake to the forum.
*Which I have now fixed. I actually copied part of you post as a quick template, lol. I just forgot to change that line. I assure you is 4.21.1.026.
Maybe this isn't the right place to ask, but i didn't know you can use multiple GPU's like that, don't you have to have NVLINK for that?