Loading...
> >

Daz Studio Iray - Rendering Hardware Benchmarking

1…2324262829…45

Comments

  • skyeshotsskyeshots Posts: 148
    edited November 2021


    System/Motherboard: Gigabyte X299X
    CPU: I9-10920X @ 3.5 GHZ
    GPU: PNY RTX A5000 x4 (+1150 Mem)
    System Memory: 32 GB Corsair Dominator Platinum DDR4-3466
    OS Drive: Intel 670p M.2 1 TB NVMe
    Asset Drive: (Same) Intel 670p M.2 1 TB NVMe
    Operating System: Win 10 Pro, 21H1
    Nvidia Drivers Version: 472.47 DCH/Branch Drivers
    Daz Studio Version: 4.15

    2021-11-11 23:55:21.492 Finished Rendering
    2021-11-11 23:55:21.548 Total Rendering Time: 32.70 seconds
    2021-11-11 23:55:24.330 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-11-11 23:55:24.330 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 2 (NVIDIA RTX A5000):      451 iterations, 2.133s init, 27.416s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA RTX A5000):      451 iterations, 2.047s init, 27.148s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA RTX A5000):      445 iterations, 2.167s init, 27.145s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 3 (NVIDIA RTX A5000):      434 iterations, 2.236s init, 27.030s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      19 iterations, 1.604s init, 27.906s render
    Loading Time: 4.794
    Rendering Performance: 1800/27.416 = 65.66 iterations per second

    Edit: Should be: 1800/27.906 (the CPU time) = 64.50

    Looks like about 85% the speed of the 4x A6000 for about half the cost. Great for smaller scenes and linkable if you gotta go big.

    Post edited by skyeshots on
  • RayDAntRayDAnt Posts: 1,135

    skyeshots said:


    System/Motherboard: Gigabyte X299X
    CPU: I9-10920X @ 3.5 GHZ
    GPU: PNY RTX A5000 x4 (+1150 Mem)
    System Memory: 32 GB Corsair Dominator Platinum DDR4-3466
    OS Drive: Intel 670p M.2 1 TB NVMe
    Asset Drive: (Same) Intel 670p M.2 1 TB NVMe
    Operating System: Win 10 Pro, 21H1
    Nvidia Drivers Version: 472.47 DCH/Branch Drivers
    Daz Studio Version: 4.15

    2021-11-11 23:55:21.492 Finished Rendering
    2021-11-11 23:55:21.548 Total Rendering Time: 32.70 seconds
    2021-11-11 23:55:24.330 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-11-11 23:55:24.330 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 2 (NVIDIA RTX A5000):      451 iterations, 2.133s init, 27.416s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA RTX A5000):      451 iterations, 2.047s init, 27.148s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA RTX A5000):      445 iterations, 2.167s init, 27.145s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 3 (NVIDIA RTX A5000):      434 iterations, 2.236s init, 27.030s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      19 iterations, 1.604s init, 27.906s render
    Loading Time: 4.794
    Rendering Performance: 1800/27.416 = 65.66 iterations per second

    Looks like about 85% the speed of the 4x A6000 for about half the cost. Great for smaller scenes and linkable if you gotta go big.

    I'd recommend running the benchmark again - this time with the CPU disabled from taking part in the render. Past testing has shown that Iray rendering performance actually degrades if you mix one or more high efficiency rendering devices (ie. high performance Nvidia GPUs) with a single much slower device like virtually any CPU or a lower end GPU.

  • skyeshotsskyeshots Posts: 148

    RayDAnt said:

    I'd recommend running the benchmark again - this time with the CPU disabled from taking part in the render. Past testing has shown that Iray rendering performance actually degrades if you mix one or more high efficiency rendering devices (ie. high performance Nvidia GPUs) with a single much slower device like virtually any CPU or a lower end GPU.

    Great point. I was putting this build together late last night and CPU was enabled by default wtih a fresh Daz install.

    Performance statistics should have been based on the CPU as it reflected the greatest render time. It looks like CPU was at 4.3 GHz as well. Despite its abysmal render speeds, the 10920x did not reduce the overall system performance.

    System/Motherboard: Gigabyte X299X
    CPU: I9-10920X @ 4.3 Ghz
    GPU: PNY RTX A5000 x4 (+1150 Mem)
    System Memory: 32 GB Corsair Dominator Platinum DDR4-3466
    OS Drive: Intel 670p M.2 1 TB NVMe
    Asset Drive: (Same) Intel 670p M.2 1 TB NVMe
    Operating System: Win 10 Pro, 21H1
    Nvidia Drivers Version: 472.47 DCH/Branch Drivers
    Daz Studio Version: 4.15

    With CPU Enabled:
    2021-11-11 23:55:21.548 Total Rendering Time: 32.70 seconds
    2021-11-11 23:55:24.330 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-11-11 23:55:24.330 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 2 (NVIDIA RTX A5000):      451 iterations, 2.133s init, 27.416s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA RTX A5000):      451 iterations, 2.047s init, 27.148s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA RTX A5000):      445 iterations, 2.167s init, 27.145s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 3 (NVIDIA RTX A5000):      434 iterations, 2.236s init, 27.030s render
    2021-11-11 23:55:24.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      19 iterations, 1.604s init, 27.906s render
    Loading Time: 4.794
    Rendering Performance: 1800/27.906 = 64.50


    Without CPU Enabled:
    2021-11-12 20:13:49.739 Total Rendering Time: 33.16 seconds
    2021-11-12 20:14:02.492 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-11-12 20:14:02.492 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 2 (NVIDIA RTX A5000):      457 iterations, 2.311s init, 27.185s render
    2021-11-12 20:14:02.492 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA RTX A5000):      457 iterations, 2.054s init, 27.220s render
    2021-11-12 20:14:02.492 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA RTX A5000):      470 iterations, 2.146s init, 27.892s render
    2021-11-12 20:14:02.492 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 3 (NVIDIA RTX A5000):      416 iterations, 2.156s init, 27.895s render
    Loading Time: 5.265
    Rendering Performance: 1800/27.895 = 64.53

  • takezo_3001takezo_3001 Posts: 1,979

    Welp, I'm gonna post another testing session using the newest beta and studio drivers...

    My system:
    WIN 10 21H2
    AMD RYZEN 3900x
    SAMSUNG 1Tb SSD EVO 850
    64Gb DDR4 2400
    RTX 3090 vrs 472.39 STUDIO DRIVER

    SPECULAR RENDER MODE:

    2021-11-13 08:23:39.866 Finished Rendering
    2021-11-13 08:23:39.913 Total Rendering Time: 1 minutes 50.0 seconds
    2021-11-13 08:24:07.848 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 0.868s init, 107.413s render

    NORMAL RENDER MODE:

    2021-11-13 08:35:39.676 Finished Rendering
    2021-11-13 08:35:39.712 Total Rendering Time: 1 minutes 43.69 seconds
    2021-11-13 08:35:53.846 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 0.876s init, 101.153s render

    Whelp, specular rendering is at least 6 seconds slower than the default rendering mode, but still, I am sold on it especially with character realism! Though to be honest, it doesn't display much of a difference with this lighting/scene...

  • chrislbchrislb Posts: 100

    skyeshots said:

    And at a considerably smaller power budget too - 230 vs. 280 watts. Which might not sound like all that much of a difference by itself. But when you're talking about having too or more of these in a system together... that can easily be difference between needing a new power supply or not (my 750 watt EVGA G2 is still going strong with this setup.)

    ETA: That's another thing I'm really appreciating about this A5000 right now: All that performance - just a single 8-pin power connector.

    Jumping back to this very important point. The cards and EPS cables are truely amazing in terms of effeciency. The A6000s use the same hookups as the A5000s. It makes cable management easier with better airflow. When I ordered custom cables for my A6000 cards, they were also cheaper than 2 or 3 groups of PCIe.

    The real issue with the gaming cards though, in my opinion, is the thermal issues - inside and outside the case. Even with the full water loop on my tripple 3090 rig the heat into the room is virtually identical to the therms from the quad A6000 setup. I know this from the power management software in thier respective Corsair PSUs. This is 20% + wasted energy in therms. And you pay for those therms twice: once to create them and then again to cool them with your AC. 

    I will throw together a quad A5000 system later this week for some comparisons.

    GPUs have been that way for several generations where there are diminishing returns of adding more power at normal ambient room temperatures.

    For example In one of the ray tracing graphics benchmarks, the score improvement from 500 watts power consumption ot 700 watts power consumption per card is about 1,300 points difference with the 500 watt power limit being a score around 29,700 and 700 watts power consumption being a score a little over 31,000.  At about a 400 watt power limit per card, the score is a little under 28,000. 

  • skyeshotsskyeshots Posts: 148
    edited November 2021

    Just looking back at this, I should clarify on the cables. The A6000 cards use a single 8-pin CPU/EPS-12V cable. The A5000 cards use a single PCIe 8 Pin. The single cable hookup is much cleaner, but these are different cables.

     

    A6000 notes.PNG
    649 x 746 - 129K
    Post edited by skyeshots on
  • RayDAntRayDAnt Posts: 1,135

    skyeshots said:

    Just looking back at this, I should clarify on the cables. The A6000 cards use a single 8-pin CPU/EPS-12V cable. The A5000 cards use a single PCIe 8 Pin. The single cable hookup is much cleaner, but these are different cables.

    Very interesting - hadn't caught that about the A6000! Don't think I've heard of anything other than a CPU that uses EPS. Add that to the list of things that make the A6000 not just simply a more fully populated PCB variant of its cheaper brerthren (still surprised by Nvidia's decision to use the 3090 Founders Edition V-style blow through PCB layout on it despite the overall card having the same footprint as the 3080 Reference Edition style A5000.)

    Picked up a second A5000 this weekend. Expect to see some watercooled A5000 benchmarking in the near future. crying

  • skyeshotsskyeshots Posts: 148

    Daz Studio Version 4.16 - Power draw analysis for the (4) card A5000, A6000 and 3090 setups w/CPU enabled:

    System/Motherboard: Gigabyte X299X
    CPU: I9-10920X @ 4.6 Ghz
    GPU: MSI Ventus RTX 3090 x4 (+1190 Mem)
    System Memory: 64 GB Corsair Dominator Platinum DDR4-3466
    OS Drive: Intel 670p M.2 1 TB NVMe
    Asset Drive: (Same) Intel 670p M.2 1 TB NVMe
    Operating System: Win 10 Pro, 21H1
    Nvidia Drivers Version: 496.76 Game Ready Drivers
    Daz Studio Version: 4.16

    2021-11-26 21:21:46.308 Total Rendering Time: 27.73 seconds
    2021-11-26 21:21:50.586 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-11-26 21:21:50.586 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 2 (NVIDIA GeForce RTX 3090):      450 iterations, 2.148s init, 22.724s render
    2021-11-26 21:21:50.586 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 3090):      446 iterations, 2.007s init, 22.696s render
    2021-11-26 21:21:50.586 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA GeForce RTX 3090):      446 iterations, 2.158s init, 22.548s render
    2021-11-26 21:21:50.587 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 3 (NVIDIA GeForce RTX 3090):      440 iterations, 2.205s init, 22.482s render
    2021-11-26 21:21:50.587 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      18 iterations, 1.431s init, 22.913s render
    Loading Time: 27.73 - 22.913 = 4.817
    Rendering Performance: 1800/22.913 = 78.86 iterations per second
    Power Draw: 1789 Watts in, 1649 Watts out at PSU 92% effeciency

    System/Motherboard: Gigabyte X299X
    CPU: I9-10980XE @ 4.6 GHZ
    GPU: PNY RTX A6000 x4 (+1250 Mem)
    System Memory: 64 GB Corsair Dominator Platinum DDR4-3466
    OS Drive: Intel 670p M.2 1 TB NVMe
    Asset Drive: Intel Optane 380 GB (905p)
    Operating System: Win 10 Pro, 21H1
    Nvidia Drivers Version: 472.47 DCH/Branch Drivers
    Daz Studio Version: 4.16

    2021-11-25 21:31:15.604 Total Rendering Time: 28.88 seconds
    2021-11-25 21:31:19.798 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-11-25 21:31:19.799 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 2 (NVIDIA RTX A6000):      432 iterations, 2.181s init, 22.842s render
    2021-11-25 21:31:19.799 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA RTX A6000):      448 iterations, 1.943s init, 23.443s render
    2021-11-25 21:31:19.799 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA RTX A6000):      437 iterations, 2.064s init, 22.843s render
    2021-11-25 21:31:19.799 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 3 (NVIDIA RTX A6000):      460 iterations, 1.803s init, 23.994s render
    2021-11-25 21:31:19.799 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      23 iterations, 1.411s init, 23.761s render
    Loading time: 28.88-23.994 = 4.886
    Rendering Performance: 1800/23.994 = 75.02 iterations per second
    Power Draw: 1687 Watts in, 1566 out at PSU 93% effeciency

    System/Motherboard: Gigabyte X299X
    CPU: I9-10920X @ 4.6 Ghz
    GPU: PNY RTX A5000 x4 (+1175 Mem)
    System Memory: 64 GB Corsair Dominator Platinum DDR4-3466
    OS Drive: Intel 670p M.2 1 TB NVMe
    Asset Drive: (Same) Intel 670p M.2 1 TB NVMe
    Operating System: Win 10 Pro, 21H1
    Nvidia Drivers Version: 496.76 Game Ready Drivers
    Daz Studio Version: 4.16

    2021-11-26 22:28:44.056 Total Rendering Time: 33.68 seconds
    2021-11-26 22:29:22.563 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-11-26 22:29:22.563 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 3 (NVIDIA RTX A5000):      436 iterations, 2.261s init, 28.153s render
    2021-11-26 22:29:22.564 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA RTX A5000):      430 iterations, 2.425s init, 27.467s render
    2021-11-26 22:29:22.564 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA RTX A5000):      437 iterations, 2.414s init, 28.232s render
    2021-11-26 22:29:22.564 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 2 (NVIDIA RTX A5000):      473 iterations, 2.427s init, 28.313s render
    2021-11-26 22:29:22.564 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      24 iterations, 1.448s init, 28.868s render
    Loading time: 4.812
    Rendering Performance: 1800/28.868 = 62.35 iterations per second
    Power Draw 1195 Watts in. 1118 Watts out at PSU 94% effeciency

    Planning out a 4x 3090 water cooled setup. These are modest scores here, but for this round I was focusing on system power draw.

    The 4x 3090 setup (no fans or pump) pegged out a Corsair AX1600i. To add water to the setup, it looks like I will need a 2nd PSU.

  • RayDAntRayDAnt Posts: 1,135

    Fwiw added a "Power Supply: BRAND MODEL WATTAGE" line to the self-reporting instructions for the thread since imo it's becoming an increasingly useful thing to know.

  • skyeshotsskyeshots Posts: 148

    RayDAnt said:

    Fwiw added a "Power Supply: BRAND MODEL WATTAGE" line to the self-reporting instructions for the thread since imo it's becoming an increasingly useful thing to know.

    This is a great idea for HEDT users. It is good to know your draw and see the opportunities for more build out. The Ventus 3090 cards I tested above don't pull much in terms of power. In contrast, 3090s with 3 PCIe connectors and/or unlocked bios, etc usually draw considerably more, up to 450 watts per card. From the wall this may be closer to 500 watts depending on the PSU. Try this x4 with a high end CPU, water cooling, etc and you have load issues for a residential 20 AMP (which should really be kept under 1920W). Thinking outside the box here, it is a good idea if you can put your rendering PC(s) on dedicated circuits for safety and stability.

    Even if you don't have fancy equipment to test the power draw, just watch your lights. If the lights flicker when you hit the render button, its time to branch the PC onto another circuit. And whatever you do, no matter how tempting it is, don't plug the Xmas lights into the same outlet as the Daz PC.

  • skyeshotsskyeshots Posts: 148

    Also, have been running some benchmarks on Windows 11 Pro 21H2. OOTB Windows 11 is consistantly performing 2-3% slower on my 4x A6000 build. There is a long list of potential culprits, but something worth mentioning.

  • guska84guska84 Posts: 4

    Decided to see how my Ryzen 9 stacked up in order to objectively justify a second GPU just for rendering, and figured I'd add my results in here

    System Configuration
    System/Motherboard: ASUS ROG Strix B550-E
    CPU: AMD Ryzen 9 3900XT @ 3793MHz
    GPU: AMD Radeon RX 5700 XT @ Stock
    System Memory: Corsair Vengeance 2x16GB @ 1799MHz
    OS Drive: Samsung 860 EVO 1TB
    Asset Drive: Same)
    Operating System: Win 11 Pro 10.0.22000
    Nvidia Drivers Version: N/A
    Daz Studio Version: 4.16.0.3 64bit
    Optix Prime Acceleration: N/A

    Benchmark Results

    DAZ_STATS
    2021-12-15 14:33:14.947 Finished Rendering
    2021-12-15 14:33:14.991 Total Rendering Time: 30 minutes 11.5 seconds
    IRAY_STATS
    2021-12-15 14:53:31.986 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-12-15 14:53:31.986 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU: 1800 iterations, 2.104s init, 1806.319s render


    Iteration Rate: (1800 / 1806.319) = 0.996 iterations per second
    Loading Time: ((0 * 3600 + 30 * 60 + 11.5) - 1806.319) = 5.181 seconds

    There's no existing results table for Ryzen 9 up there, so here's a new one

     

    Contributor            OS           DS Version              OptiX     Total Rendering Time         Device Render     Iteration Rate     Loading Timetristan.lovett84       W11 22000    4.16.0.3 x64            NA        30 minutes 11.5 seconds      1806.319          0.996              5.181

     

    BenchmarkResult.png
    900 x 900 - 1M
  • RayDAntRayDAnt Posts: 1,135
    edited December 2021

    Added a 2nd RTX A5000 (plus a fancy 1500 watt PSU) to the mix and commenced on a fresh round of benchmarking before going on to the next step (see below.)

    System Configuration
    System/Motherboard: Gigabyte Z370 Aorus Gaming 7
    CPU: Intel 8700K @ stock (MCE enabled) - onboard graphics used for all display functionality
    GPU: Nvidia RTX A5000 @ stock
    GPU: Nvidia RTX A5000 @ stock
    GPU: Nvidia Titan RTX @ stock (custom watercoooled)
    System Memory: Corsair Vengeance LPX 32GB DDR4 @ 3000Mhz
    OS Drive: Samsung Pro 980 512GB NVME SSD
    Asset Drive: Sandisk Extreme Portable SSD 1TB
    Power Supply: Corsair AX1500i 1500 watts
    Operating System: Windows 10 Pro version 21H2 build 19044.1415
    Nvidia Drivers Version: 472.12 SRD
    Daz Studio Version: 4.16.0.3 64-bit
    Optix Prime Acceleration: N/A

    Benchmark Results - RTX A5000 #1 (PCI-E slot 2):
    WDDM Driver Mode
    Total Rendering Time: 1 minutes 53.12 seconds
    CUDA device 0 (NVIDIA RTX A5000): 1800 iterations, 2.033s init, 108.846s render
    Iteration Rate: 16.537 iterations per second
    Loading Time: 4.274 seconds

    TCC Driver Mode
    Total Rendering Time: 1 minutes 50.90 seconds
    CUDA device 0 (NVIDIA RTX A5000): 1800 iterations, 2.630s init, 106.002s render
    Iteration Rate: 16.981 iterations per second
    Loading Time: 4.898 seconds

     

    Benchmark Results - RTX A5000 #2 (PCI-E slot 5):
    WDDM Driver Mode
    Total Rendering Time: 1 minutes 52.78 seconds
    CUDA device 1 (NVIDIA RTX A5000): 1800 iterations, 2.151s init, 108.364s render
    Iteration Rate: 16.611 iterations per second
    Loading Time: 4.420 seconds

    TCC Driver Mode
    Total Rendering Time: 1 minutes 50.73 seconds
    CUDA device 1 (NVIDIA RTX A5000): 1800 iterations, 1.949s init, 106.669s render
    Iteration Rate: 16.875 iterations per second
    Loading Time: 4.061 seconds

     

    Benchmark Results - Titan RTX (PCI-E slot 7):
    WDDM Driver Mode
    Total Rendering Time: 3 minutes 18.91 seconds
    CUDA device 2 (NVIDIA TITAN RTX): 1800 iterations, 2.160s init, 194.503s render
    Iteration Rate: 09.254 iterations per second
    Loading Time: 4.407 seconds

    TCC Driver Mode
    Total Rendering Time: 3 minutes 14.91 seconds
    CUDA device 2 (NVIDIA TITAN RTX): 1800 iterations, 2.174s init, 190.556s render
    Iteration Rate: 09.446 iterations per second
    Loading Time: 4.354 seconds

     

    And a couple card combo benchmarks for good measure.

    Benchmark Results - RTX A5000 #1 (PCI-E slot 2) + RTX A5000 #2 (PCI-E slot 5):
    WDDM Driver Mode
    Total Rendering Time: 59.60 seconds
    CUDA device 0 (NVIDIA RTX A5000): 902 iterations, 2.094s init, 55.138s render
    CUDA device 1 (NVIDIA RTX A5000): 898 iterations, 2.139s init, 54.625s render
    Iteration Rate: 32.645 iterations per second
    Loading Time: 4.470 seconds

    TCC Driver Mode
    Total Rendering Time: 57.99 seconds
    CUDA device 0 (NVIDIA RTX A5000): 902 iterations, 2.435s init, 53.312s render
    ​CUDA device 1 (NVIDIA RTX A5000): 898 iterations, 1.957s init, 52.919s render
    Iteration Rate: 30.764 iterations per second
    Loading Time: 4.678 seconds

     

    Benchmark Results - RTX A5000 #1 (PCI-E slot 2) + RTX A5000 #2 (PCI-E slot 5) + Titan RTX (PCI-E slot 7):
    WDDM Driver Mode
    Total Rendering Time: 48.81 seconds
    CUDA device 0 (NVIDIA RTX A5000): 712 iterations, 2.110s init, 44.221s render
    CUDA device 1 (NVIDIA RTX A5000): 707 iterations, 2.077s init, 44.043s render
    CUDA device 2 (NVIDIA TITAN RTX): 381 iterations, 2.159s init, 43.585s render
    Iteration Rate: 40.705 iterations per second
    Loading Time: 4.589 seconds

    TCC Driver Mode
    Total Rendering Time: 47.21 seconds
    CUDA device 0 (NVIDIA RTX A5000): 708 iterations, 2.032s init, 42.157s render
    ​CUDA device 1 (NVIDIA RTX A5000): 705 iterations, 1.851s init, 42.324s render
    CUDA device 2 (NVIDIA TITAN RTX): 387 iterations, 1.942s init, 42.892s render
    Iteration Rate: 41.966 iterations per second
    Loading Time: 4.318 seconds

    Given the Iray growing pains currently playing out in the beta channel, I decided to go with the current official release of Daz Studio for this testing regimen.

    You will notice in the data above that A5000 #2 is slightly but consistently slower at rendering than A5000 #1. This is because the card starts to thermal limit after around 2 minutes of continuous load due to there only being an 8mm gap between it and the backplate of the watercooled Titan RTX next to it (locations dictated by the motherboard design):

    Coming probably not very soon - a new motherboard/CPU/RAM combo.

    From studying these two cards' clock speeds I'm pretty sure that A5000 #2 is - in actuality - a slightly better performer since it initially reaches higher clock speeds before thermal loading of the heat sink causes them to drop. Somethhing I find interesting given that A5000 #1 was purchased allegedly brand new (the box was missing its standard complement of connector cables...) whereas A5000 #2 came used (with a full complement of cables...) ripped from a video production rig. 

    Both A5000's definitely have additional performance headroom that the stock cooler design simply isn't capable of accommodating. The only way to know how much would be to watercool them. Speaking of which...

    Coming very soon - watercooled RTX A5000 benchmarks (already have the blocks and fittings - just waiting on an NVLink bridge and the courage to be custom modding $5k+ of professional computer gear to arrive to get started.)

    case_config.jpg
    3024 x 4032 - 4M
    Post edited by RayDAnt on
  • TheMysteryIsThePointTheMysteryIsThePoint Posts: 2,948

    RayDAnt said:

    Coming very soon - watercooled RTX A5000 benchmarks (already have the blocks and fittings - just waiting on an NVLink bridge and the courage to be custom modding $5k+ of professional computer gear to arrive to get started.)

    That's funny. When I built my first TR, the stress of slotting a thousand dollar CPU was daunting. Physically mounting 4 $1500 GPUs and hoping, just hoping to god that everything just worked the first time, was not fun. Especially when they didn't.

  • outrider42outrider42 Posts: 3,679
    edited December 2021

    A crazy thing happened, out of the blue I got an email from EVGA saying that the 3060 I signed up for way back in FEBURARY was in the que and available for purchase. I had completely forgotten that I had signed up for it! While I had managed to somehow get a 3090 in October, I could not possibly pass this up, so heck yeah I grabbed the 3060. It was $400, which is what EVGA's MSRP is now. I wasn't super happy about that because it was $330 when I signed up, but that is still like HALF what 3060s go for. So by some miracle I have secured both a 3090 and 3060 at (technically) MSRP prices in less than a 3 month span. I feel extremely fortunate to have these, its funny, because previously I had never bought a brand new GPU in my life. Every single GPU I had ever purchased, including my two 1080tis, was second hand off ebay.

    So here is a new test, I believe a new combo here, a 3090+3060 test. I am still using 4.15 at this time.

    Windows 10  20H2

    Ryzen 5800X

    GPU 1 3090 Founder's Edition

    GPU 2 EVGA 3060 Black

    64GB RAM

    Asset Drive Samsung 4TB 870 EVO

    OS Drive 2TB M.2 Inland Platinum

    Daz 4.15.1.72

    GPU Driver 472.12

    2021-12-15 21:20:39.776 Total Rendering Time: 1 minutes 9.90 seconds

    2021-12-15 21:21:00.518 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:

    2021-12-15 21:21:00.518 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA GeForce RTX 3060): 520 iterations, 1.653s init, 65.435s render

    2021-12-15 21:21:00.518 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1280 iterations, 1.493s init, 66.128s render

    That comes to 27.50 iterations per second.

    This test is very interesting compared to the 3090+1080ti test I did before. These are the numbers from that test.

    2021-10-24 20:47:30.055 Total Rendering Time: 1 minutes 19.7 seconds

    2021-10-24 20:47:43.123 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1469 iterations, 1.399s init, 75.498s render

    2021-10-24 20:47:43.123 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA GeForce GTX 1080 Ti): 331 iterations, 1.628s init, 75.867s render

    That is 23.84 iterations per second.

    So...only 9 seconds faster? But that 9 seconds also translates to 3.66 extra iterations per second, which is not such a small number. As the numbers get lower you get what appears to be diminishing returns. I imagine this number would be felt better in more complex renders. What really stood out to me was the sound. The old EVGA 1080ti I had in there could get loud. I would have kept the larger MSI 1080ti with the 3090, except it is too large to fit in my case next to the freakin beast 3 slot 3090. The EVGA is just a 2 slot card, and so with a smaller cooler it gets warmer. That also means a louder fan. The 3060 is soooo much quieter, even though this is just a Black model with fairly basic cooler. It is also comical to look at this 3060 installed next to the Founder's 3090. The 3060 Black looks so cute, LOL. Together these 2 cards are much quieter than my old two 1080tis, and I am getting far more performance. My two 1080tis were running this bench in around 3:30, so hitting 1:10 is exactly a third of the time required. That is quite a drastic upgrade, and this is just one bench scene.

    For fun this pic shows the 3060 and 3090 in my rig. Take a wild guess which is which. The difference in size is staggering.

    Also, this makes me think how Nvidia really shouldn't be putting the power connector in the middle of the card, and the 12 pin adapter stands out so much. I think adding some form of cable management into the card design would be the next step for GPUs. As it is, trying to keep the cables away from the fans means running the cables right over Nvidia's RTX branding on the card. I bet they didn't consider that in their design at all. The EVGA stands out because you can actually see its branding. I plan on using a different cable on the 3060, but this is where the 1080ti was with its 2 8pins, and I would like to get a 12 pin cable for the 3090 that splits at the power supply rather than the 3090.

    Next up I plan on making a special comparison between the 1080ti and the 3060. I was originally going to do the 1080ti VS the 3090, but I think the 3060 is a much more interesting card to compare the 1080ti to. The 3060 is roughly equal to the 1080ti in gaming, so this is a good opportunity to really experiment and try to figure out just how much these ray tracing cores really bring to the table. I want to attempt to answer the question of how well the performance gaps we see in the bench scene hold up in other tests. I will make a thread dedicated to what I find.

    Post edited by outrider42 on
  • RayDAntRayDAnt Posts: 1,135
    edited December 2021

    outrider42 said:

    So...only 9 seconds faster? But that 9 seconds also translates to 3.66 extra iterations per second, which is not such a small number. As the numbers get lower you get what appears to be diminishing returns. I imagine this number would be felt better in more complex renders.

    Yeah - always keep in mind that the "iterations per second" stat this benchmark revolves around is a proportion. Meaning that smaller reductions in rendering time translate to larger performance gains as rendering time overall gets shorter.

    Also, this makes me think how Nvidia really shouldn't be putting the power connector in the middle of the card, and the 12 pin adapter stands out so much. I think adding some form of cable management into the card design would be the next step for GPUs.

    A series cards have their power connectors on the butt end (opposite the display connecttors) of the card. There are so many little things like this about these cards that make them so damn usable... I really can't recommend them enough - especially during an era when price inlfation of their consumer brethren has made them price comparable.

     

    Post edited by RayDAnt on
  • skyeshotsskyeshots Posts: 148

    outrider42 said:

    A crazy thing happened, out of the blue I got an email from EVGA saying that the 3060 I signed up for way back in FEBURARY was in the que and available for purchase.

    Grats on the added boost in that setup! It is kinda funny how that 3090 dwarfs the 3060 in your pic. Iteration per dollar wise the 3060 is still the gold medal winner.

  • outrider42outrider42 Posts: 3,679
    edited December 2021

    RayDAnt said:

    outrider42 said:

    So...only 9 seconds faster? But that 9 seconds also translates to 3.66 extra iterations per second, which is not such a small number. As the numbers get lower you get what appears to be diminishing returns. I imagine this number would be felt better in more complex renders.

    Yeah - always keep in mind that the "iterations per second" stat this benchmark revolves around is a proportion. Meaning that smaller reductions in rendering time translate to larger performance gains as rendering time overall gets shorter.

    Also, this makes me think how Nvidia really shouldn't be putting the power connector in the middle of the card, and the 12 pin adapter stands out so much. I think adding some form of cable management into the card design would be the next step for GPUs.

    A series cards have their power connectors on the butt end (opposite the display connecttors) of the card. There are so many little things like this about these cards that make them so damn usable... I really can't recommend them enough - especially during an era when price inlfation of their consumer brethren has made them price comparable.

    Yeah, the iteration count will make a difference at long renders for sure. The 3.66 iteration gap actually makes sense. One 1080ti does right about 4 iterations per second, the 3060 can do over 8. So the 3.66 gap fits in there, though it falls short of scaling like I would hope, I am using cards with wide gaps in performance levels.

    I forgot to post the 3060 alone, though several others have posted theirs. More data is always nice.

    3060 alone, all specs the same as previous post.

    2021-12-15 21:29:59.217 Total Rendering Time: 3 minutes 41.78 seconds

    2021-12-15 21:30:08.962 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:

    2021-12-15 21:30:08.962 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA GeForce RTX 3060): 1800 iterations, 1.342s init, 218.887s render

    8.22 iterations per second

    I can understand how the A series may be more appealing under current market conditions, the A4000 and newer A4500 are pretty compelling. So compelling that even gamer focused Hardware Unboxed decided to review the A4000 for gaming because of how crazy things are right now. But I do feel they are still out of reach for most people. I got lucky and managed to get my 3090 at $1500. If I wasn't able to score one I would just never have bought a new card at all. Plus I do plan on buying an OLED TV to use as a monitor, and since the A series only has Display Port, that just wouldn't work for me. I will need HDMI for the TV. Those 42" LGs cannot release soon enough! Though now Samsung is getting involved with their own version of OLED which they are calling WOLED, and they may be releasing a 32" size screen. That would be the perfect size, but I do not want to buy a first generation product (these are not the same as AMOLED used in their phones and tablets). I know how LG OLEDs perform, and how LG handles their warranties. But I don't want to stray off topic.

    Post edited by outrider42 on
  • RayDAntRayDAnt Posts: 1,135
    edited December 2021

    The GPU buying market is so bizarre right now. Through some judicious waiting I was able to buy both of my A5000s in Ebay auctions for just slightly over their original MSRP ($2,350) prices - both ludicrous events if you consider the going rates of consumer level cards with even vaguely similar performance specs. There seems to be a blind spot in the current GPU buying/selling craze for upper mid-tier workstation graphics cards (my theory for why being that there really isn't an effective way for scalpers to market them as something desirable to consumers.) For anyone still in the hunt for high performance GPU muscle, I can't recommend enough going after A series cards despite their hefty prices (even in normal times.) Unlike any consumer level card they will continue to retain their value (due to their workstation/pro features) to the right people even long after consumer GPU prices return to normal. whenever that is. 

     

    outrider42 said:

    Plus I do plan on buying an OLED TV to use as a monitor, and since the A series only has Display Port, that just wouldn't work for me. I will need HDMI for the TV.

    Keep in mind that all it takes is a passive cable to feed an HDMI device from a DisplayPort output (the opposite - HDMI out to DispalyPort in - isn't possible.) Although I'm not sure if things like variable/60+ refresh rates or HDR is possible under such configurations.

    Post edited by RayDAnt on
  • skyeshotsskyeshots Posts: 148

    System/Motherboard: Gigabyte X299X
    CPU: I9-10980XE @ 4.4 GHZ
    GPU: PNY RTX A6000 x4 & A5000 x1 (+1150 Mem)
    System Memory: 96 GB Corsair Dominator Platinum DDR4-3466
    OS Drive: Intel 670p M.2 1 TB NVMe
    Asset Drive: Intel Optane 380 GB (905p)
    Operating System: Win 11 Pro, 21H2
    Nvidia Drivers Version: 472.47 DCH
    Daz Studio Version: 4.16

    2021-12-15 23:44:02.071 Finished Rendering
    2021-12-15 23:44:02.192 Total Rendering Time: 25.47 seconds
    2021-12-15 23:44:05.126 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2021-12-15 23:44:05.126 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA RTX A6000):      362 iterations, 2.029s init, 19.473s render
    2021-12-15 23:44:05.126 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 3 (NVIDIA RTX A6000):      364 iterations, 2.223s init, 19.637s render
    2021-12-15 23:44:05.126 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 4 (NVIDIA RTX A5000):      327 iterations, 1.707s init, 19.603s render
    2021-12-15 23:44:05.126 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 2 (NVIDIA RTX A6000):      365 iterations, 2.322s init, 19.699s render
    2021-12-15 23:44:05.127 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA RTX A6000):      364 iterations, 1.827s init, 19.470s render
    2021-12-15 23:44:05.127 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      18 iterations, 1.422s init, 20.092s render

    Loading Time: 5.378
    Performance: 89.59 Iterations Per Second

    Baseline test here using PCIe bifercation w/5 GPUs in the case. Auxilary 750 watt Mini-ITX PSU for the added juice. It all fits in the tower well enough. I guess I will have to share pics...Planning to replace the A5000 for one that matches the rest of the setup though and clean it up. With CPU disabled here I am dropping down to about 80.5 IPS or 8% slower. This defies the general rule of thumb by a long shot. Could be OS/Win 11 related, or perhaps in part due to the GPU mismatch. IDK though.

  • outrider42outrider42 Posts: 3,679

    RayDAnt said:

    The GPU buying market is so bizarre right now. Through some judicious waiting I was able to buy both of my A5000s in Ebay auctions for just slightly over their original MSRP ($2,350) prices - both ludicrous events if you consider the going rates of consumer level cards with even vaguely similar performance specs. There seems to be a blind spot in the current GPU buying/selling craze for upper mid-tier workstation graphics cards (my theory for why being that there really isn't an effective way for scalpers to market them as something desirable to consumers.) For anyone still in the hunt for high performance GPU muscle, I can't recommend enough going after A series cards despite their hefty prices (even in normal times.) Unlike any consumer level card they will continue to retain their value (due to their workstation/pro features) to the right people even long after consumer GPU prices return to normal. whenever that is. 

     

    outrider42 said:

    Plus I do plan on buying an OLED TV to use as a monitor, and since the A series only has Display Port, that just wouldn't work for me. I will need HDMI for the TV.

    Keep in mind that all it takes is a passive cable to feed an HDMI device from a DisplayPort output (the opposite - HDMI out to DispalyPort in - isn't possible.) Although I'm not sure if things like variable/60+ refresh rates or HDR is possible under such configurations.

    You struck gold on those A5000s!

    The A and Quadro series for whatever reason just haven't appealed to gamers even in the current market. Many don't seem to even know that these cards can play games. After all, Hardware Unboxed video on the A4000 was one of the few gaming focused outlets to ever make a full gaming bench set with such a card. Steve did not really know what to expect out of it, and this is a guy who has done GPU hardware for years. It is kind of weird how well Nvidia has segmented these markets. I recall Linus Tech Tips covering Quadro a few years ago, but that card was $5000. He was not impressed, at least for gaming. He did explain why it was $5000, and showed what it could do with the software Quadro is designed for compared to using gaming cards.

    A reason the A series could make sense is how bizzarre Nvidia has been with VRAM. The 3080 has been knocked for having just 10gb, and most of their cards have 8. The 3090 stands out not because of its performance, but more so because of its VRAM, as does the 3060. If the 3080 actually matched its AMD competition with 16gb, boy I wonder how things might have gone. I would have totally signed up for such a 3080, and certainly the almost released 3080 20gb that was rumored. That card was planned, it was very real. They really were going to make that, specifically to compete with AMD, but changed their minds when they realized everything was selling. This ridiculous market has cost us more than just high prices and constant out of stock...it cost us some really cool products, too.

    Display Port to HDMI adaptors sadly do not support the game functions. As far as I can tell, they do not have a way to do that yet. You can get a lot of other features, but not VRR. VRR is pretty important to gaming, and though I game a lot less than I used to, I still do on occassion.

  • outrider42outrider42 Posts: 3,679

    skyeshots said:

    outrider42 said:

    A crazy thing happened, out of the blue I got an email from EVGA saying that the 3060 I signed up for way back in FEBURARY was in the que and available for purchase.

    Grats on the added boost in that setup! It is kinda funny how that 3090 dwarfs the 3060 in your pic. Iteration per dollar wise the 3060 is still the gold medal winner.

    I am pretty stoked, and very fortunate. It is hard to believe this same case once housed a GTX 970.

    I signed up for the 3060 in Febuary believing that I would never get a 3090. I also rebuilt my computer with a Ryzen 5800X thinking I would probably not be getting a new GPU anytime soon. So I figured I would at least do something as my CPU was getting long in the tooth.

    The 3060 and 3090 two were the only cards I ever really targeted. I did not want anything less than 12GB, and the 3080ti was just too close to the 3090 to make much sense to me considering the VRAM. And I wasn't going to ever make a scalper happy. The only other card that tempted me was the A4000, and I was looking into buying one, until I saw the lack of HDMI. I was so disappointed when I saw that...but not long after that I struck gold with the 3090. What a turn around!

    So 2021 has been a wild ride in that regard. I have never owned a computer that was this state of the art in every aspect. Almost everything is entirely new. I even bought a 4TB SSD for Daz and a M.2 for the OS. Only the case and power supply remain from my former PC now, along with a couple SSDs. I just realized that all my internal drives are SSD now, LOL. I have some externals that HDD.

  • RayDAntRayDAnt Posts: 1,135

    Anyone have an M1 Mac handy?

  • Silas3DSilas3D Posts: 565

    RayDAnt said:

    Anyone have an M1 Mac handy?

    I too would be curious to know results for the M1 Macs - Especially for the M1 Pro and M1 Max variants.

    My current iMac has done very well to survive for seven years but its now time for an upgrade, so waiting patiently to see what comes of the iMac Pro rumoured for Spring 2022.

  • leemoon_c43b45a114leemoon_c43b45a114 Posts: 866

    The M1 Macs are still limited to CPU rendering, even with the ARM-native Iray renderer in the current DS beta.

    I did render a simple character full body portrait using my default lighting scene.

    DS 4.16.1.17 beta on M1 MacBook Air: 2 min 48 sec

    DS 4.15.0.30 on M1 iMac: 2 min 20 sec

    DS 4.12.0.86 on i9 MacBook Pro: 1 min 32 sec

    I also started the Iray Benchmark scene on the M1 MacBook Air, but could see that it would take hour(s) to complete, so I stopped the render.  I needed to continue working on my ghostlights setup with the newer version of Iray.

    The M1 GPUs were not used, which we knew would be the case, and the 4 efficiency cores and 4 performance cores on the M1 were at 100% for my personal test renders.

    I'm looking forward to more powerful Mac desktops next year.

    Oh, if I get some time, I'll let the M1 MacBook Air run the Iray Benchmark scene to completion and post the results.  I'd also like to see M1 Max and M1 Pro results, for those who have those Macs. :)

    Lee

  • leemoon_c43b45a114leemoon_c43b45a114 Posts: 866
    edited December 2021

    Ran the benchmark on the M1 iMac and it just completed a few minutes ago.

    M1 iMac, 8 core CPU, 8 core GPU, 16GB RAM, 2TB SSD, DAZ Studio 4.15.0.30, and the forum Iray benchmark scene.

    From the log file:

    2021-12-18 11:21:26.807 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      1800 iterations, 2.479s init, 5710.788s render

    2021-12-18 11:20:30.387 Total Rendering Time: 1 hours 35 minutes 16.38 seconds

    Hope this is an informative data point.

    Lee

    Benchmark-1h35m17sec.png
    900 x 900 - 1M
    Post edited by leemoon_c43b45a114 on
  • RayDAntRayDAnt Posts: 1,135
    edited December 2021

    leemoon_c43b45a114 said:

    Ran the benchmark on the M1 iMac and it just completed a few minutes ago.

    M1 iMac, 8 core CPU, 8 core GPU, 16GB RAM, 2TB SSD, DAZ Studio 4.15.0.30, and the forum Iray benchmark scene.

    From the log file: 2021-12-18 11:20:30.387 Total Rendering Time: 1 hours 35 minutes 16.38 seconds

    Hope this is an informative data point.

    Lee

    Could you grab the Iray statistics line from your log file (described in detail here under step 3) and post it as well? The total rendering time by itself isn't actually useful in measuring rendering performance (counter-intuitive I know, but that's just the way Daz Studio + Iray works.)

    Post edited by RayDAnt on
  • leemoon_c43b45a114leemoon_c43b45a114 Posts: 866

    I edited my post above to include the Iray statistics, RayDAnt.

    I would have included them initially, but the log file wasn't showing them until after I saved the rendered image. Then that info was written to the log file. :)

    Is that everything you need?

    Lee

  • outrider42outrider42 Posts: 3,679

    Yeah, the log file does not have all the info until the image is either saved or closed. You do not actually have to save the image if you do not want to, you can select to close it.

    I rather doubt the M1 will ever get full GPU support for Daz Iray. Iray requires CUDA to run on a GPU. Only Nvidia GPUs can run this code. So no matter how fast or amazing AMD, Intel and Apple make their GPUs, they will never be able to run Iray. Any devices that have GPUs from them will always be forced to render on CPU. And in case somebody asks, no Rosetta cannot be used to run CUDA applications. I have seen the topic come up, but I have not seen a case of it actually working.

    I am sorry, but that is just how it is. If Nvidia allows them to use CUDA that could change, but it seems extremely unlikely that will ever happen. CUDA is Nvidia's single greatest asset in the industry. Additionally, Nvidia and Apple had a major falling out between each other a few years ago. Regardless of whose fault it might be, there are no Nvidia drivers for any Apple products for Turing and Ampere. So you cannot even plug a 2000 or 3000 series GPU up to a Mac with an external enclosure.

    So lets look at the benchmarks:

    Apple M1........1.5 HOURS    That equates to 0.315 iterations per second.

    RTX 3090.......1.5 MINUTES   That is 19.623 iterations per second (based on my personal 3090 time).

    That is a factor of 60! Not 2 or 3 times faster, not even 10 times faster. The 3090 is over 60 times faster than the Apple M1 for rendering Iray. And while it may be seem unfair to compare the 3090 to what is essentially a mobile CPU rather than the GPU, the Apple GPU is never going to run Iray. So it is in fact totally fair to make that comparison because the CPU is all that M1 owners will ever get to use on Iray.

    So Apple users will probably want to either use 3DL or Filament to render or export to other software that can render on the M1 GPU.

  • Navin JNavin J Posts: 5

    Finally, I was able to get the RTX A5000 after a very long wait. I'm super excited and happy on the results.

    DAZ 3D 4.16.0.3 Pro x64 - RTX A5000 - Windows 11 - Latest NVIDIA drivers 511.09

    System Configuration

    System/Motherboard:  MSI Z370 PC PRO 
    CPU:  Intel i7-8700 3.20 GHz @ stock 
    GPU:  NVIDIA RTX A5000 (CORE: -300 mhz | MEM: +1500 mhz | Power: 99%) - [Afterburner screenshot] 
    System Memory: 1x CORSAIR VENGEANCE 16GB DDR4 (CMK16GX4M1B3000C15) @ stock 
    OS Drive:  Samsung PM981 NVMe 1024 GB 
    Asset Drive:  Seagate 2TB (ST2000DL003-9VT116)  Backup:  Samsung 960 Evo 250 GB 
    Operating System:  Windows 11 Home Single 21H2 (22000.434) 
    Nvidia Drivers Version:  511.09 
    Daz Studio Version:  4.16.0.3 Pro x64 
    Power Supply: Corsair VS450 (450 watts) 
     

    Benchmark Results

    DAZ_STATS

    2022-01-17 14:30:31.989 Finished Rendering
    2022-01-17 14:30:32.032 Total Rendering Time: 1 minutes 47.47 seconds

    IRAY_STATS

    2022-01-17 14:30:42.026 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2022-01-17 14:30:42.026 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA RTX A5000):      1800 iterations, 2.524s init, 103.064s render


    Iteration Rate:   1800/103.064 = 17.4648 
    Loading Time:   107.47 - 103.064 = 4.406

     

    Plugged into my office system to test it out (I had to remove the HDD cage to accommodate the gpu - hence, dirty n bad cable management)
    Until my new power supply arrives - the card is rock steady with a lower tier 450w psu!

    Previous benchmark (RTX 1660 Super)
    https://www.daz3d.com/forums/discussion/341041/daz-studio-iray-rendering-hardware-benchmarking/p17#Comment_6404861

    RTX A5000 - VS450w.jpg
    2268 x 3300 - 3M
    MSI Afterburner - A5000 - Daz3D.png
    800 x 550 - 186K
Sign In or Register to comment.