Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.Daz 3D is part of
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
The tight content management system you're using sounds very efficient. Am not there yet as still setting things up. I do manually disable morphs though that am no longer using, and that has helped alot. When done setup will do same as you.
When i was saying interactivity, was thinking more along the lines of many characters, with wearable, and environments. That sitting in RAM (disabled ALL my windows harddrive caching), am curious how differnt RAM/CPU combos would affect that.
Am really looking forward to upgrading when 5*** series shows up with way better system-RAM then too. But won't have near same number of GPUS as you.
Could i bug you though to elaborate about chipset design maybe being more key? CPUs, unlike, GPUS, seem pretty slow relatively speaking to GPUs to improve speeds. Plus lots of software functions, like animations, are still by and large single core, whcih means cpu top speed is most important. How would chipset architecture be an additional component to that top speed and RAM access? My RAM speeed is currently 3600mhz (not smart sounding lol).
As you probably remember have a tower 900 too. Love the room, except for the power supply location. Nightmare that one. Would def be intrigued by a perf-based workstation writeup. Hope you do! Heck, i'm still air cooled on mine. One good reason stilll only 1 gpu. Airspace would be tight with 2+. Main GPU temp that MSI afterburner monitors never exceeds 68degC reagrdless what i throw at it, but the hot junction/hot spot, yeah one day looked and based on hwinfo64 that was around 88-90 degC. Not sure if liquid cooling can deal with that one?
My 4090 with +100 core +1000 vram and an 80% power target tends to run around 48c core, 50c vram & 56c hotspot with a fairly simple loop while using DS in a room with an ambient temp in the low 20s.
The only fans apart from the psu are the 5 attached to rads. The fans are typically between 400 - 1100 rpm depending on water temp, hitting about 1100 during an hour long render.
An hour of cyberpunk with everything up to 11 hits about 60/50s/72 with the fans at about 1250.
Hmmm. Cyberpunk dialed to max only yields you 72? Only have a 3090 though it is an Xtreme variant, and that would have that hotpsot at 88-90, though I can't obvi goes as high as you can.
Has water cooling really improved? Seems when I looked not too long ago the performance diff anecdotally was not quite as pronounced.
Thought i was doing well with 68degC (what most users report, or have been) vs that hotspot which is lot hotter.
So what are you getting for reg reported GPU temp? ie vs my 68degC (not junction nor hotspot ie. my 88-90).
Curious if your difference is less than mine. In that case rad cooling into GPU would be worthwhile for sure if want to keep for years and years.
lol, the way am updating GPUs, i update before they ever get heat stressed. But still. Like lower temps, esp. if lots lower. Then the whole risk of water cooler leak is more worthwhile, and doing reg maintenance.
You can't really directly compare 3090 and 4090 numbers because the 4090 is easier to cool because of the board layout but can draw more power. I had a watercooled 3090 FE and under a heavy gaming load I used to get high 50s on the core and high 70s on the ram with the hotspot somewhere between the two. Knock about 6c off if rendering in DS. Truthfully, if the 3090 temps bother you I'd sell it and get a 4090 rather than go to the expense of building a loop. You'd need a couple of decent sized rads, pump and reservoir, fittings , tube and other assorted parts. All that plus what you could sell a 3090 for gets you into 4090 territory.
Just tried re-saving the benchmark scene via the latest Beta, still "job rejected by the server".
Didn't know that about 4090 being easier to cool. Good to know! Have not looked at 4*** series, as had decided will wait for 5***series after having read some "hopium en mass" about what they may include, especially more VRAM.
Not worried about these temps, being under 90degC. For me again interesting part was the different board design. Thanks for saying! Guilty of not looking. Usually do. But as say, had decided to upgrade with 5*** series. Maybe in <1 year now? Look at water cooling then too. Rather spend time on DAZ content than PC building.
An upgrade to the server was available this week
Upgraded server to 2023.0.5 (build 367100.4957) however the benchmark scene was still rejected, with the same 'Integer out of range' error, when trying to queue it. ...but does works when streaming!
Unfortuantely I don't have access to an Iray Server instance to test againts at the moment. Have you filed a bug report with Daz about it? If not, I highly suggest doing so here.
I have some render servers, I can stand you up a test server if you want to do some testing.
System Configuration
System/Motherboard: SuperMicro X12
CPU: 2x Xeon Gold 6348
GPU: 4x RTX6000 ADA Generation
System Memory: 512 GB DDR4 ECC @ 3200 MHz
OS Drive: SK hynix Platinum P41 2TB PCIe NVMe Gen4
Asset Drive: 256 GB RAM DRIVE
Daz Studio: 4.22.0.1 Public Build
Operating System: Win 11 Pro
Nvidia Driver: 537.70
Benchmark Results
2023-11-14 20:48:50.056 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2023-11-14 20:48:50.056 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA RTX 6000 Ada Generation): 408 iterations, 1.245s init, 18.159s render
2023-11-14 20:48:50.057 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA RTX 6000 Ada Generation): 469 iterations, 1.270s init, 18.519s render
2023-11-14 20:48:50.057 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 2 (NVIDIA RTX 6000 Ada Generation): 462 iterations, 1.359s init, 18.431s render
2023-11-14 20:48:50.057 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 3 (NVIDIA RTX 6000 Ada Generation): 461 iterations, 1.345s init, 18.475s render
2023-11-14 20:48:50.391 [INFO] :: Finished Rendering
2023-11-14 20:48:50.457 [INFO] :: Total Rendering Time: 22.97 seconds
Iteration Rate: 97.20 iterations per second
Loading Time: 4.451 seconds
Current public build and updated drivers.
System Configuration
System/Motherboard: ASUS TUF GAMING X570-PLUS
CPU: AMD Ryzen 9 3900X
GPU: MSI RTX 4090 Gaming X Trio
System Memory: corsaire 16gb x 4 64gb total
OS Drive: samsung 960 EVO
Asset Drive:
Power Supply: EVGA SuperNOVA 1300W
Operating System: Windows 11, Build 22621, Enterprise
Nvidia Drivers Version: 531.79
Daz Studio Version: 4.21.1.108
Optix Prime Acceleration: No
Benchmark Results
DAZ_STATS Total Rendering Time: 1 minutes 12.91 seconds
IRAY_STATS 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4090): 1800 iterations, 1.792s init, 69.006s render
Iteration Rate: 26.0846
Loading Time: 3.904 seconds
System Configuration
System/Motherboard: ASUS Pro WS WRX80E-SAGE-SE WiFi
CPU: AMD Ryzen Threadripper Pro 5975WX
GPU: (3x)PNY GeForce RTX 4090 XLR8
System Memory: (8x) Samsung 3200 MHZ Registered RDIMM M393AG40AB2-CWE (512 MB)
OS Drive: SAMSUNG SSD 990 Pro M.2 4TB
Asset Drive: WD_Black SN850X 2TB
Power Supply: Seasonic Prime ATX 3.0 1600 watt
Operating System: Windows 11 Pro Build 22621
Nvidia Drivers Version: 31.0.15.4601 Studio
Daz Studio Version: 4.22
Benchmark Results
2023-11-17 16:06:15.414 [INFO] :: Finished Rendering 2023-11-17 16:06:15.444 [INFO] :: Total Rendering Time: 25.81 seconds
IRAY_STATS
Iteration Rate: CUDA device 2 (NVIDIA GeForce RTX 4090): 600 iterations, 0.694s init, 21.930s render: CUDA device 1 (NVIDIA GeForce RTX 4090): 579 iterations, 0.701s init, 22.875s render: CUDA device 0 (NVIDIA GeForce RTX 4090): 621 iterations, 0.719s init, 22.714s render
Loading Time: 2.935 seconds
Thanks, would you mind doing the bench with one 4090? I don't think we have a 4.22 test with one 4090. I think the speed is pretty much the same, but it would be nice to verify.
here are the results for one:
System Configuration
System/Motherboard: ASUS Pro WS WRX80E-SAGE-SE WiFi
CPU: AMD Ryzen Threadripper Pro 5975WX111111111
GPU: (3x)PNY GeForce RTX 4090 XLR8
System Memory: (8x) Samsung 3200 MHZ Registered RDIMM M393AG40AB2-CWE (512 MB)
OS Drive: SAMSUNG SSD 990 Pro M.2 4TB
Asset Drive: WD_Black SN850X 2TB
Power Supply: Seasonic Prime ATX 3.0 1600 watt
Operating System: Windows 11 Pro Build 22621
Nvidia Drivers Version: 31.0.15.4601 Studio
Daz Studio Version: 4.22
Benchmark Results
2023-11-18 04:58:07.144 [INFO] :: Total Rendering Time: 1 minutes 4.23 seconds
IRAY_STATS
2023-11-18 04:59:12.446 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4090): 1800 iterations, 0.503s init, 62.449s render
Loading time: 1.781 seconds
Thank you for taking the time to do that. That comes to 28.8 iterations per second, which is spot on with where the 4090 has been.
BTW, Iray 2023.0.6 is now in the Daz PRIVATE Beta branch. This surprised me a bit, because for one, this is the latest version that just dropped on November 7. So Daz just skipped all the others, lol.
Hopefully we will get Iray 2023 released to the public beta soon. This is a big update, with lots of promised performance improvements. Not only do they claim Iray 2023 runs better on RTX 4000, but they say it improves speeds for ALL hardware because of other fixes they have done. They optimized how Iray calculates certain geometry, for example, so in some scenes we may see improvements not observed in other scenes. They also revamped the caustic sampler and guided sampler, making them faster. They also claim many bug fixes, and who knows, these fixes might be fixes for some problems we have had with Iray 2022.
Of course, that is what they claim to have done. We cannot be certain until we get our hands on it. Considering it is only in the private beta, it may be a couple months before it goes public.
Since I need to test that my new GeForce RTX 4070 is properly working before the rest of my new PC parts come in, I decided to do this DAZ Studio iRay benchmark again.
System Configuration
System/Motherboard: Gigabyte B450M DS3H WIFI (AMI UEFI BIOS F66B)
CPU: AMD Ryzen 7 5700G with Radeon Graphics / 3.80 GHz
GPU: PNY GeForce RTX 4070 GPU 12GB XLR8 Gaming VERTO EPIC-X RGB / 1920MHz
System Memory: 2x16GB Patriot Premium / 3200MHz
OS Drive: PNY CS2130 2 TB M.2-2280 PCIe 3.0 X4 NVME SSD
Asset Drive: Crucial 2 TB Sata III SSD
Power Supply: SeaSonic 750W Gold
Operating System: Windows 11 Pro 23H2 22631.2715
Nvidia Drivers Version: 546.01 Studio
Daz Studio Version: DAZ Studio Pro Public Beta 4.22.0.7 64-bit
Optix Prime Acceleration: N/A
Benchmark Results
DAZ_STATS: 2023-12-03 15:20:22.737 [INFO] :: Total Rendering Time: 2 minutes 21.87 seconds
IRAY_STATS: 2023-12-03 15:20:22.202 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Received update to 01800 iterations after 139.706 s.
Iteration Rate: 1800 iterations/139.706 seconds = 12.88+ iterations/second
Loading Time: 261.87 seconds TRT - 139.706 seconds rendering time = 122.164 seconds loading time
I don't know if this was supposed to be obvious but I can confirm it. Even using only 1 monitor at fullhd resolution, running my RTX 4090 as the display card my bechmark results were always 25.6~ iterations per second in DAZ 4.22 BETA, no matter what I did I could not get it to 28 IPS. (I kept the card itself at stock speeds, I just mean other stuff like closing all background process, unplugging the internet, removing other gpu, unplugging slow storage devices... I know these aren't supposed to affect the rendering once it's loaded in the GPU but was worth a try I guess).
RTX 4090 ASUS ROG STRIX @ stock speeds
Now I did the bench again with another card as the display card, selected only the RTX 4090 and got the 28.3~ IPS in DAZ 4.22 BETA. Will go over the thread later to check other results.
I know the benchmark isn't absolute, but considering a RTX 3060 can do the avarage of 6.6 IPS, isn't 3 iterations too much of a loss only by using the card as the display for 1 monitor?
Imo people tend to hugely overlook how much of a processing load driving windows graphics is on a GPU. Here's a fun little demosntration you can do: Pop open Task Manager to the tab showing the 3D Activity graph for the GPU you are using to drive your display(s). Then start moving a window (not the Task Mnaager window) as rapidly as you can all around your screen(s), and watch as the activity graph starts to go up. Now consider the fact that that is just the processing it takes for a single window. Every window, every status box, every system overlay is an additional piece of stuff that the GPU is having to contend with while you do whatever else it is you think you are actually using the GPU to do. The WDDM (Windows Display Driver Model) is essentially like having a game running constantly in the background whenever it's being used to drive displays. An extremely well optimized one (especially at idle), but one whose overall load on system resources is still there regardless.
Remember this test is not an absolute. The results could also be influenced by how short the test is now. A delay of just a few seconds can translate to multiple iterations being lost. Thus the penalty gets higher for the fastest GPUs. A better way is to test on some larger scenes and observe how much impact it has.
That said, it is also possible the issue comes Iray 2022, which is still what Daz uses. The 4090 can achieve 28+ iterations with Iray 2021. At some point, one day, who knows when, Daz Studio may get Iray 2023. Iray 2023 showed up in the PRIVATE beta channel recently. So hopefully it will make its way to the public beta eventually. It did not make it into the beta that just dropped which contains Face Transfer 2. You just have to keep asking Daz when we might finally get to play with this new Iray.
Been a while, i've assembled a new computer to replace my 3-year-old Alienware m17 R4 laptop.
System Configuration
System/Motherboard: MSI B550 Tomahawk Max WIFI
CPU: AMD Ryzen 7 5700X @ stock (Power Limited to 65W)
GPU: MSI Ventus 2X OC RTX 4060Ti 16GB @ stock (Power Limited to 100W)
System Memory: Kingston Fury BEAST @ 3200 MT/s
OS Drive: Kingston NV2 Gen4 NVME 2TB
Asset Drive: Same
Power Supply: MSI A850GF 850W 80+ Gold PSU
Operating System: Windows 10 Home 22H2
Nvidia Drivers Version: 546.33
Daz Studio Version: 4.22.01
Optix Prime Acceleration: Enabled
Benchmark Results
2023-12-16 08:19:56.185 Iray [INFO] - IRAY:RENDER :: 1.2 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4060 Ti): Scene processed in 0.979s
2023-12-16 08:19:56.186 Iray [INFO] - IRAY:RENDER :: 1.2 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4060 Ti): Allocated 15.450 MiB for frame buffer
2023-12-16 08:19:56.191 Iray [INFO] - IRAY:RENDER :: 1.2 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4060 Ti): Allocated 1.969 GiB of work space (2048k active samples in 0.005s)
2023-12-16 08:19:56.191 Iray [INFO] - IRAY:RENDER :: 1.2 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 4060 Ti): Optimizing for cooperative usage (performance could be sacrificed)
2023-12-16 08:19:56.443 Iray [INFO] - IRAY:RENDER :: 1.2 IRAY rend info : Allocating 1-layer frame buffer
2023-12-16 08:23:36.013 [INFO] :: Finished Rendering
DAZ_STATS
2023-12-16 08:23:36.035 [INFO] :: Total Rendering Time: 3 minutes 50.21 seconds
IRAY_STATS
Iteration Rate: (1800 / 220.322s) 8.1698 iterations per second
Loading Time: ((230.21) - 220.322) 9.8888 seconds
What version of iray does Daz Studio 4.16.0.3 use? (I can only find info for 4.16.0.003 which says iray 2020.1.6, but I'm not sure if that extra zero is a typo). I'm looking into buying a new GPU but I don't want to upgrade studio any further and I want to make sure I get a GPU compatible with what I currently have installed.
It certainly looks like a typo, and if the version of interest doesn't specify a version it will be using the one that was last abbounced.
Iray 2020.1.6
The "extra" zeros seen eg. here are on purpose for formatting reasons (they make scanning up and down through version lists at a glance easier to do.)
System Configuration
System/Motherboard: Supermicro X12
CPU: 2x Xeon Gold 6348
GPU: 4x RTX6000 ADA Generation
System Memory: 512 GB DDR4 ECC @ 3200 MHz
OS Drive: SK Hynix Platinum P41 2TB PCIe Gen4
Asset Drive: 256 GB RAM DRIVE
PSU: 2x Corsair AX1600i
Operating System: Win 11 Pro
Nvidia Driver: 537.99
Daz Studio: 4.22.0.15
Benchmark Results with 4x RTX6000 ADA
2023-12-27 13:26:09.696 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2023-12-27 13:26:09.696 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA RTX 6000 Ada Generation): 414 iterations, 1.220s init, 17.262s render
2023-12-27 13:26:09.696 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA RTX 6000 Ada Generation): 461 iterations, 1.130s init, 17.041s render
2023-12-27 13:26:09.696 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 2 (NVIDIA RTX 6000 Ada Generation): 451 iterations, 1.172s init, 17.015s render
2023-12-27 13:26:09.696 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 3 (NVIDIA RTX 6000 Ada Generation): 474 iterations, 1.147s init, 17.752s render
2023-12-27 13:26:10.519 [INFO] :: Finished Rendering
2023-12-27 13:26:10.585 [INFO] :: Total Rendering Time: 22.45 seconds
Iteration Rate: 101.4 iterations per second
Loading Time: 4.698 seconds
Benchmark Results with 1x RTX6000 ADA
2023-12-27 23:27:59.624 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2023-12-27 23:27:59.624 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 3 (NVIDIA RTX 6000 Ada Generation): 1800 iterations, 1.157s init, 63.457s render
2023-12-27 23:28:00.293 [INFO] :: Finished Rendering
2023-12-27 23:28:00.360 [INFO] :: Total Rendering Time: 1 minutes 7.15 seconds
Iteration Rate: 28.3657
Loading Time: 3.603
Daz 4.22.0.15 appears to be faster. Following the conversation, I tried a few runs with just a single 'non-display' GPU. The results are similar to a 4090. When I tested 3x GPUs, skipping the one being used for display, it was just under 80 iterations per second; (79.87 / 3 = 26.62 average). I think if something else was pushing the display(s), I would land right around 105 iterations per second.
Anyone have luck running the benchmark on the Beta 4.22.1.41 with new IRAY version?
Having issues? It's working for me, but I had to go and manually download/update my graphics drivers for it to work (Iray 2023.1.0 requires driver version 545.84 as mininum, and GeForce Experience doesn't seem to be pushing out the latest Studio driver release past that version for some unknown reason. Maybe server update issues.)
Since it was asked if anyone had run the benchmark with the public beta 4.22.1.41 I have run it here.
Hardware Notice:
a) New MSI motherboad replacing the Gigabyte motherboard
b) New Fikwot 2TB USB M.2 NVMe SSD replacing the 2TB Crucial Sata III SSD
Software Config Notice:
a) 1st quoted test was ran without CPU fallback while the below test was ran with CPU fallback and that seems to have allowed DAZ Studio to shell off part of the rending to the CPU. At least that's what I think has happened. That might explain the longer render time. Correct me if I'm wrong and you know why.
b) The nVidia driver and Windows 11 were also both updated. I am using the stable "Studio" driver.
System Configuration
System/Motherboard: MSI MAG Tomahawk B550M WIFI
CPU: AMD Ryzen 7 5700G with Radeon Graphics / 3.80 GHz
GPU: PNY GeForce RTX 4070 GPU 12GB XLR8 Gaming VERTO EPIC-X RGB / 1920MHz
System Memory: 2x16GB Patriot Premium / 3200MHz
OS Drive: PNY CS2130 2 TB M.2-2280 PCIe 3.0 X4 NVME SSD
Asset Drive: Fikwot FN501 Pro 2TB NVMe SSD
Power Supply: SeaSonic 750W Gold
Operating System: Windows 11 Pro 23H2 22631.2861
Nvidia Drivers Version: 546.33 Studio
Daz Studio Version: DAZ Studio Pro Public Beta 4.22.1.41 64-bit
Optix Prime Acceleration: N/A
Benchmark Results
DAZ_STATS: 2023-12-03 15:20:22.737 [INFO] :: Total Rendering Time: 2 minutes 31.79 seconds
IRAY_STATS: 2023-12-03 15:20:22.202 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Received update to 01800 iterations after 148.816 s.
Iteration Rate: 1800 iterations/148.816 seconds = 12.095+ iterations/second
Loading Time: 151.79 seconds TRT - 148.816 seconds rendering time = 2.974 seconds loading time
Re: CPU fallback, you should see the CPU listed in the render log output.
System/Motherboard: MSI MPG Z490 Carbon EK X
CPU: I9-10850K @ 3.6 ghz
GPU: RTX 4000 ADA (20 GB - Single Slot GPU)
System Memory: 64 GB Corsair Dominator Platinum DDR4-3466
OS Drive: Samsung 970 EVO SSD 1TB – M.2 NVMe
Asset Drive: Same
Operating System: Win 11 Pro
Nvidia Drivers Version: 546.01
Daz Studio Version: 4.22.1.54 (Public Build)
2024-01-12 23:02:11.312 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2024-01-12 23:02:11.312 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA RTX 4000 Ada Generation): 1646 iterations, 3.135s init, 168.622s render
2024-01-12 23:02:11.312 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CPU: 154 iterations, 1.830s init, 168.312s render
2024-01-12 23:02:11.851 [INFO] :: Finished Rendering
2024-01-12 23:02:11.942 [INFO] :: Total Rendering Time: 2 minutes 59.41 seconds
Loading Time: 10.788 seconds
GPU Iteration Rate: 9.76 iterations per second
System/Mothboard : Gigabyte Z790 Gaming X AX
CPU : I9 13900K (no overclock)
GPU : RTX 4070 SUPER 12 Go Gigaoctet Windforce oc
System Memory : 32 Go DDR5 Kingston Fury X
OS Drive : Samsung 990 PRO 2TO -nvme
OS : W11
Nvidia Driver : 546,45
Daz Studio 4.22 (Last Public Build...)
Dont know render infos, but she render at 1800 itérations in 2 minutes, 10 secondes. GPU Only
GPU + CPU : 1800 ité at 1 minute 57 sec
/Google Traduction