Daz Studio Iray - Rendering Hardware Benchmarking

1235714

Comments

  • DustRiderDustRider Posts: 2,344
    edited November 2019

    Thought it would be interesting to see how my new laptop with an RTX 2070 and a i9-9900K performed compared to the desktops. Looks like it did pretty well for a laptop. It's an MSI GT76 Titan DT 9SF-091.

    Hope I did everything properly!!

    System Configuration
    System/Motherboard: MSI (Laptop)
    CPU: Intel i9-9900K stock
    GPU: Nvidia RTX 2070 stock
    System Memory: 128GB (4x32GB) DDR4 @ 2666MHz
    OS Drive: 2TB XPC certified M.2 PCIE NVME SSD (Read 2000MB/S | Write 1750MB/S)
    Asset Drive: Same
    Operating System: Windows 10 Pro Build 18363.476
    Nvidia Drivers Version: 441.20
    Daz Studio Version: DS Beta V14.12.1..16

    Benchmark Results RTX 2070
    DAZ_STATS:17:43:45.880 Finished Rendering
                         17:43:45.906 Total Rendering Time: 7 minutes 1.21 seconds
    IRAY_STATS: CUDA device 0 (GeForce RTX 2070): 1800 iterations, 5.601s init, 413.007s render
    Iteration Rate: (1800/421.21) = 4.273 iterations per second
    Loading Time: ((0 +7 * 60 + 1.21 = 421.21) - 413.007) = 8.203 seconds

    Benchmark Results i9-9900K
    DAZ_STATS:20:02:26.835 Finished Rendering​
                         Total Rendering Time: 51 minutes 48.32 seconds​
    IRAY_STATS: CPU: 1800 iterations, 2.161s init, 3103.563s render
    Iteration Rate: (1800/ 3,109.32) = 0.5789 iterations per second
    Loading Time: ((0 +51 * 60 + 49.32 = 3,109.32) - 3103.563) = 5.757 seconds​

    Post edited by DustRider on
  • RayDAntRayDAnt Posts: 807

    Updated tables with all the latest results. 

    @TheMysteryIsThePoint out of curiosity, how did you manage to cram 4 2080Ti's onto a Gaming 7 board?

  • KharmaKharma Posts: 2,978
    edited November 2019

    Hope I did this correctly, I couldn't find anything saying Load Time though:

    System Configuration
    System/Motherboard: ASUS X99-A USB 3.1
    CPU: Intel I7-5820K 3.3 GHz LGA 2011 v3 stock 
    GPU: EVGA NVIDIA GTX 970 4GB
    System Memory: G,SKILL Ripjaws 32 GB (4 x 8 GB) DDR4 2400
    OS Drive: Samsung 850 EVO 2.5 SSD SATA III 500 GB
    Asset Drive: WD Black 2TB 7200 RPM
    Operating System: Windows 10 HOME 17763.864
    Nvidia Drivers Version: 430.39
    Daz Studio Version: DS Beta V14.12.1.16
    Optix Prime Acceleration: STATE (omit with Daz Studio 4.12.1.x or later)

    Benchmark Results
    DAZ_STATS  2019-11-16 10:09:22.244 Finished Rendering
                          2019-11-16 10:09:22.285 Total Rendering Time: 17 minutes 34.97 seconds
    IRAY_STATS

    Iteration Rate:  1800 / 1044.410 = 1.724 iterations per second
    Loading Time: ((17 * 60) + 34.97) - 1044.410 = 10.560 seconds

    Post edited by Kharma on
  • RayDAntRayDAnt Posts: 807
    Kharma said:

    Hope I did this correctly, I couldn't find anything saying Load Time though:

    System Configuration
    System/Motherboard: ASUS X99-A USB 3.1
    CPU: Intel I7-5820K 3.3 GHz LGA 2011 v3 stock 
    GPU: EVGA NVIDIA GTX 970 4GB
    System Memory: G,SKILL Ripjaws 32 GB (4 x 8 GB) DDR4 2400
    OS Drive: Samsung 850 EVO 2.5 SSD SATA III 500 GB
    Asset Drive: WD Black 2TB 7200 RPM
    Operating System: Windows 10 HOME 17763.864
    Nvidia Drivers Version: 430.39
    Daz Studio Version: DS Beta V14.12.1.16
    Optix Prime Acceleration: STATE (omit with Daz Studio 4.12.1.x or later)

    Benchmark Results
    DAZ_STATS 2019-11-16 10:09:22.244 Finished Rendering
                          2019-11-16 10:09:22.285 Total Rendering Time: 17 minutes 34.97 seconds
    IRAY_STATS
    Iteration Rate: 1800 iterations, 5.227s init, 1044.410s render
    Loading Time: no info shown 

    Here's what your final results evaluate to:

    Iteration Rate:  1800 / 1044.410 = 1.724 iterations per second
    Loading Time: ((17 * 60) + 34.97) - 1044.410 = 10.560 seconds

  • KharmaKharma Posts: 2,978
    RayDAnt said:
    Kharma said:

    Hope I did this correctly, I couldn't find anything saying Load Time though:

    System Configuration
    System/Motherboard: ASUS X99-A USB 3.1
    CPU: Intel I7-5820K 3.3 GHz LGA 2011 v3 stock 
    GPU: EVGA NVIDIA GTX 970 4GB
    System Memory: G,SKILL Ripjaws 32 GB (4 x 8 GB) DDR4 2400
    OS Drive: Samsung 850 EVO 2.5 SSD SATA III 500 GB
    Asset Drive: WD Black 2TB 7200 RPM
    Operating System: Windows 10 HOME 17763.864
    Nvidia Drivers Version: 430.39
    Daz Studio Version: DS Beta V14.12.1.16
    Optix Prime Acceleration: STATE (omit with Daz Studio 4.12.1.x or later)

    Benchmark Results
    DAZ_STATS 2019-11-16 10:09:22.244 Finished Rendering
                          2019-11-16 10:09:22.285 Total Rendering Time: 17 minutes 34.97 seconds
    IRAY_STATS
    Iteration Rate: 1800 iterations, 5.227s init, 1044.410s render
    Loading Time: no info shown 

    Here's what your final results evaluate to:

    Iteration Rate:  1800 / 1044.410 = 1.724 iterations per second
    Loading Time: ((17 * 60) + 34.97) - 1044.410 = 10.560 seconds

    Thnak you, updated on my original stat post

  • System Configuration
    System/Motherboard: Gigabyte x470 Aorus Gaming 5 Wifi (Bios F50a latest)
    CPU: Ryzen 7 2700X/stock
    GPU: Asus RTX 2080ti Turbo
    System Memory: Ripjaws V DDR4-3600 4x16GB (64GB) Manually Clocked to 3200 for stability
    OS Drive: Samsung SSD 950 PRO 512GB
    Asset Drive: WD My Passport 0827 USB Device 3TB
    Operating System: Windows 10 Pro Version 1809 OS build 17763.529
    Nvidia Drivers Version: 26.21.14.4112(NVIDIA 441.12) Creators
    Daz Studio Version: 4.12.1.16

    GPU ONLY
    Benchmark Results
    2019-11-16 16:03:58.453 Finished Rendering
    2019-11-16 16:03:58.491 Total Rendering Time: 4 minutes 9.3 seconds
    2019-11-16 16:04:18.609 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2019-11-16 16:04:18.609 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2080 Ti):      1800 iterations, 2.429s init, 243.297s render
    Iteration Rate: (1800 / 243.297) 7.3983649613 iterations per second
    Loading Time: ((0 + 4 * 60 + 9.3) - 243.297) 6.003 seconds

    GPU+CPU
    Benchmark Results
    2019-11-16 16:32:38.447 Finished Rendering
    2019-11-16 16:32:38.488 Total Rendering Time: 4 minutes 8.84 seconds
    2019-11-16 16:32:51.366 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2080 Ti):      1661 iterations, 2.348s init, 243.250s render
    2019-11-16 16:32:51.373 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      139 iterations, 2.156s init, 243.298s render
    Iteration Rate: (1800 / 243.298) 7.3983345527 iterations per second
    Loading Time: ((0 + 4 * 60 + 8.84) - 243.298) 5.542 seconds

    CPU Only
    Benchmark Results
    2019-11-16 17:31:58.638 Finished Rendering
    2019-11-16 17:31:58.678 Total Rendering Time: 49 minutes 21.96 seconds
    2019-11-16 17:48:07.474 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2019-11-16 17:48:07.482 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CPU:      1800 iterations, 2.127s init, 2956.609s render
    Iteration Rate: (1800 / 2956.609) 0.6088055607 iterations per second
    Loading Time: ((0 + 49 * 60 + 21.96) - 2956.609) 5.351 seconds

  • BejaymacBejaymac Posts: 1,441
    edited November 2019

    No idea about most of the stuff like Brand/ Model etc as this is an ASUS pre-built, whatever is in it is what they stuck in it, even System has very little info on what is in here.

    CPU - Intel i5 9400 2.9

    GPU - GTX 1660 Ti

    RAM - pretty sure it's one 8Gb stick of DDR4 2666

    OS  - Win 10 home 64

    driver - 436.48

    DS - 4.12.0.86 (64)

    OptiX on

    2019-11-19 22:19:18.316 Finished Rendering

    2019-11-19 22:19:18.336 Total Rendering Time: 10 minutes 55.95 seconds

    2019-11-19 22:19:25.077 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:

    2019-11-19 22:19:25.077 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce GTX 1660 Ti): 1800 iterations, 3.520s init, 649.555s render

    OptiX off

    2019-11-19 22:50:25.379 Finished Rendering

    2019-11-19 22:50:25.398 Total Rendering Time: 10 minutes 57.59 seconds

    2019-11-19 22:50:32.032 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:

    2019-11-19 22:50:32.032 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce GTX 1660 Ti): 1800 iterations, 3.488s init, 651.231s render

     

    Little difference wether OptiX is on or off

     

    EDIT

    It's an ASUS ROG STRIX GL10CS for anyone that knows wtf that means

    Post edited by Bejaymac on
  • outrider42outrider42 Posts: 2,674

    Thanks for posting. I think that's the first 1660ti time posted, so its always good to get numbers for new cards. So it provides a noticeable bump over the 1660, but there is a big gap between it and the 2060, and especially the 2060 Super. Notably it beats the 1070, and basically ties the 1070ti.

    I would expect ASUS to list a model for the overall system, which would probably give most of the missing info. ASUS makes their own GPUs, so we at least know it is an ASUS Strix branded card. They have lots of different models, and they can have variants with 1, 2 or even 3 fans. In general, the more fans probably means faster factory overclocks. The single fan one has a "gaming mode" of 1785 Mhz, while the fastest triple fan (that I found) has a gaming mode set to 1860 Mhz. For Daz Iray the cards should render at around those clockspeeds. The card would shoot up its advertised boost clock at the start of a render, then as temps go up the clocks would drop to the advertised gaming clock for the rest of the render as long as temps stay within spec. So, if you have a hardware monitor running, you can probably figure out which ASUS you have by observing the clockspeeds it hits.

    BTW, if you have 8GB of RAM, I would think about adding another 8 fairly soon. It can work, but you may find yourself running out of RAM. My laptop has 8, and I can run it up just with browsing tabs and video. Daz can eat up memory, too, and if you use all the 6GB VRAM you have on your GPU, you may well exceed your 8GB RAM in the process, and your computer would run terrible. If your hard drive or SSD is constantly getting pegged, that could indicate you are using a lot of memory from paging files because the RAM is full. Plus if that is an SSD, that could be bad for its longevity. So the extra RAM would not only speed up the system, it would help extend the life of the SSD. You could probably contact ASUS about the exact memory they used if you decide you would like to add more.

  • RayDAnt said:

    Updated tables with all the latest results. 

    @TheMysteryIsThePoint out of curiosity, how did you manage to cram 4 2080Ti's onto a Gaming 7 board?

    Sorry, didn't see this post.

    The fourth one indeed doesn't fit into the second 8x slot because of other jumpers on the edge of the motherboard. But, and I cannot claim foresight for this as it didn't even occur to me that they might not physically fit, my full tower case happened to have a double width slot mounted horizontally just above where the other three cards are, so the 4th card almost lays flat along the case door. Then this:

    Phanteks PH-CBRS_FL22 – 220mm Pci-E X16 Riser Cable, 90 Degree Adapter, EMI Shielded, Slim Design

    did the trick.

    The cards are all blowers, so having them stacked so closely together is not a problem for the fans and the hot air is vented outside of the case.

    Incidentally, though, is that with all four cards installed, all the SATA ports stopped working. In any case, I think I've bought my last Gigabyte motherboard. I had two Titan RTXs that I returned because the cards worked great but I got no video signal. I had to remote into it to run Blender. The machine POSTed, and booted fine, just no video. Either card worked fine alone in either slot, though, with no problems at all. And Gigabyte was of absolutely no help at all.

  • droidy001droidy001 Posts: 189
    edited November 2019
    Managed to shave off about 30s with a modest oc of the 2069 super. I got +140 on the clock and +700 on the memory before it started to throttle. Although I will be putting it back to stock. My ears just can't take it. Gpu fans running at 100% + having to up the case fans.
    Post edited by droidy001 on
  • outrider42outrider42 Posts: 2,674

    A 2069 huh, well now, that really must be super. devil

    But yeah, that kind of overclock is probably not advisable. Maybe you kept it from throttling but I wonder how that would really play out after an hour or so of rendering. Not to mention extra stress on the fans. I also kind of doubt the memory overclock made any difference. Iray is not swapping data in and out like a game does (due to the entire scene loading onto VRAM), so it is not pushing VRAM like it is the GPU core. In fact, I guarantee if you tried playing a game with those same overclock settings (and especially a game benchmark like TimeSpy,) it wouldn't hold because the games will run a lot hotter than Iray due to the additional stress on memory. I can run 10C or more hotter with a game VS Iray. The memory OC might help the initial loading time, so you might see a slight reduction in time just due to that, but it wont actually speed up rendering itself.

    If Iray behaved differently, like say how Octane can run "out of core rendering" with a CPU, then maybe memory speed would matter more. But the way Iray is designed I don't see any benefit to memory speed other than very minor and specific instances. We don't even see a benefit going from PCIe2 to 3. It is not something Iray relies on.

  • RayDAnt said:

    Updated tables with all the latest results. 

    @TheMysteryIsThePoint out of curiosity, how did you manage to cram 4 2080Ti's onto a Gaming 7 board?

    Sorry, didn't see this post.

    The fourth one indeed doesn't fit into the second 8x slot because of other jumpers on the edge of the motherboard. But, and I cannot claim foresight for this as it didn't even occur to me that they might not physically fit, my full tower case happened to have a double width slot mounted horizontally just above where the other three cards are, so the 4th card almost lays flat along the case door. Then this:

    Phanteks PH-CBRS_FL22 – 220mm Pci-E X16 Riser Cable, 90 Degree Adapter, EMI Shielded, Slim Design

    did the trick.

    The cards are all blowers, so having them stacked so closely together is not a problem for the fans and the hot air is vented outside of the case.

    Incidentally, though, is that with all four cards installed, all the SATA ports stopped working. In any case, I think I've bought my last Gigabyte motherboard. I had two Titan RTXs that I returned because the cards worked great but I got no video signal. I had to remote into it to run Blender. The machine POSTed, and booted fine, just no video. Either card worked fine alone in either slot, though, with no problems at all. And Gigabyte was of absolutely no help at all.

    Any chance you could post a pic of your card setup within the case?

  • Jason GalterioJason Galterio Posts: 1,671
    edited November 2019

    System Configuration
    System/Motherboard: ASRock Z97 Extreme6
    CPU: Intel Core i7-4790 @ 3.6 GHz
    GPU: EVGA RTX GeForce 2070 Super
    System Memory: Corsair Vengence 8GB DDR3 x4
    OS Drive: SanDisk Ultra II 960GB SSD
    Asset Drive: Western Digital WD30EZRX 2.8TB
    Operating System: Windows 10 Pro
    Nvidia Drivers Version: 441.20
    Daz Studio Version: 4.12.0.86

    Benchmark Results
    2019-11-22 11:16:21.026 Finished Rendering
    2019-11-22 11:16:21.068 Total Rendering Time: 5 minutes 1.95 seconds

    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070 SUPER):      1800 iterations, 6.677s init, 290.227s render

    Iteration Rate: 6.2 iterations per second
    Loading Time: 12 seconds

    Test Render.jpg
    900 x 900 - 473K
    Post edited by Jason Galterio on
  • Any chance you could post a pic of your card setup within the case?

    I'd rather not disconnect everything and pull it out of the corner just to take a picture of the fourth card on top of and obscuring the other three, as well as my bird's nest wiring job. In any case, there're really no decisions to be made; if you're going to do it, your particular setup is fully determined by where your case has put the extra double slot bracket. I think this is becoming more common because a dinky little $79 case I built for a NAS had one, too.

    I can say that it works perfectly, though, and heat does not appear to be a problem, even with them crammed in there with so little space between them, especially the first three directly on the motherboard. I haven't rendered frames overnight with them yet, but I don't expect any problems. With the dual 1080 setup I upgraded from, they would emit an odour of hot metal that was kind of concerning. Now it's worse, but everything works.

     

  • A 2069 huh, well now, that really must be super. devil

    But yeah, that kind of overclock is probably not advisable. Maybe you kept it from throttling but I wonder how that would really play out after an hour or so of rendering. Not to mention extra stress on the fans. I also kind of doubt the memory overclock made any difference. Iray is not swapping data in and out like a game does (due to the entire scene loading onto VRAM), so it is not pushing VRAM like it is the GPU core. In fact, I guarantee if you tried playing a game with those same overclock settings (and especially a game benchmark like TimeSpy,) it wouldn't hold because the games will run a lot hotter than Iray due to the additional stress on memory. I can run 10C or more hotter with a game VS Iray. The memory OC might help the initial loading time, so you might see a slight reduction in time just due to that, but it wont actually speed up rendering itself.

    If Iray behaved differently, like say how Octane can run "out of core rendering" with a CPU, then maybe memory speed would matter more. But the way Iray is designed I don't see any benefit to memory speed other than very minor and specific instances. We don't even see a benefit going from PCIe2 to 3. It is not something Iray relies on.

    The oc adds a 9 to it. Yeah, I don't fancy running it at full whack on the verge of throttling for any length of time, just to save 5 minutes on the hour. And ear defenders can get quite annoying over that time.
  • outrider42outrider42 Posts: 2,674

    System Configuration
    System/Motherboard: ASRock Z97 Extreme6
    CPU: Intel Core i7-4790 @ 3.6 GHz
    GPU: EVGA RTX GeForce 2070 Super
    System Memory: Corsair Vengence 8GB DDR3 x4
    OS Drive: SanDisk Ultra II 960GB SSD
    Asset Drive: Western Digital WD30EZRX 2.8TB
    Operating System: Windows 10 Pro
    Nvidia Drivers Version: 441.20
    Daz Studio Version: 4.12.0.86

    Benchmark Results
    2019-11-22 11:16:21.026 Finished Rendering
    2019-11-22 11:16:21.068 Total Rendering Time: 5 minutes 1.95 seconds

    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070 SUPER):      1800 iterations, 6.677s init, 290.227s render

    Iteration Rate: 6.2 iterations per second
    Loading Time: 12 seconds

    Thanks for this. Now we have a 2070 Super to add to the list, which gives us at least one bench time for all of the Super series of cards (unless I overlooked one). And this one looks quite interesting. The 2070 Super is quite a bit faster than a 2070 (a solid 2 minutes faster), and actually it comes in faster than the 2080 posted here as well. It is not too far off from a 2080 Super we have posted, which is surprising. That is pretty impressive considering the 2070 Super is around $200 less than the 2080 Super.

    Also of note is that this is on an older Z97 motherboard with an i7-4790. Further demonstrating how little those matter.

  • Any chance you could post a pic of your card setup within the case?

    I'd rather not disconnect everything and pull it out of the corner just to take a picture of the fourth card on top of and obscuring the other three, as well as my bird's nest wiring job. In any case, there're really no decisions to be made; if you're going to do it, your particular setup is fully determined by where your case has put the extra double slot bracket. I think this is becoming more common because a dinky little $79 case I built for a NAS had one, too.

    I can say that it works perfectly, though, and heat does not appear to be a problem, even with them crammed in there with so little space between them, especially the first three directly on the motherboard. I haven't rendered frames overnight with them yet, but I don't expect any problems. With the dual 1080 setup I upgraded from, they would emit an odour of hot metal that was kind of concerning. Now it's worse, but everything works.

     

    I gotcha. I was curious about using those and after watching few videos about how most users are mounting the cards seems like it would take some Macgyvering to get it to physically work. Seems like those adapter cable lengths are capped depending on the pci gen speed you want to operate at. 

  • marblemarble Posts: 4,905
    edited November 2019

    System Configuration
    System/Motherboard: ASRock Z97 Extreme6
    CPU: Intel Core i7-4790 @ 3.6 GHz
    GPU: EVGA RTX GeForce 2070 Super
    System Memory: Corsair Vengence 8GB DDR3 x4
    OS Drive: SanDisk Ultra II 960GB SSD
    Asset Drive: Western Digital WD30EZRX 2.8TB
    Operating System: Windows 10 Pro
    Nvidia Drivers Version: 441.20
    Daz Studio Version: 4.12.0.86

    Benchmark Results
    2019-11-22 11:16:21.026 Finished Rendering
    2019-11-22 11:16:21.068 Total Rendering Time: 5 minutes 1.95 seconds

    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070 SUPER):      1800 iterations, 6.677s init, 290.227s render

    Iteration Rate: 6.2 iterations per second
    Loading Time: 12 seconds

    Thanks for this. Now we have a 2070 Super to add to the list, which gives us at least one bench time for all of the Super series of cards (unless I overlooked one). And this one looks quite interesting. The 2070 Super is quite a bit faster than a 2070 (a solid 2 minutes faster), and actually it comes in faster than the 2080 posted here as well. It is not too far off from a 2080 Super we have posted, which is surprising. That is pretty impressive considering the 2070 Super is around $200 less than the 2080 Super.

    Also of note is that this is on an older Z97 motherboard with an i7-4790. Further demonstrating how little those matter.

    I was looking at the price of GPU cards here in NZ yesterday and it looks like the 2080 ti is twice the price of the 2070 Super here. Looking at these numbers, I'm quite surprised to see that the 2070 Super is only a minute slower than the 2080 ti (just over 5 mins vs just over 4). Is that worth paying double? Admittedly, the 2080 ti has 11GB VRAM which is looking more important these days as NVidia & DAZ between them seem to be finding ways to use up all that VRAM leaving less scope for adding characters to the scenes.

    Post edited by marble on
  • System Configuration
    System/Motherboard: ASRock Z97 Extreme6
    CPU: Intel Core i7-4790 @ 3.6 GHz
    GPU: EVGA RTX GeForce 2070 Super
    System Memory: Corsair Vengence 8GB DDR3 x4
    OS Drive: SanDisk Ultra II 960GB SSD
    Asset Drive: Western Digital WD30EZRX 2.8TB
    Operating System: Windows 10 Pro
    Nvidia Drivers Version: 441.20
    Daz Studio Version: 4.12.0.86

    Benchmark Results
    2019-11-22 11:16:21.026 Finished Rendering
    2019-11-22 11:16:21.068 Total Rendering Time: 5 minutes 1.95 seconds

    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070 SUPER):      1800 iterations, 6.677s init, 290.227s render

    Iteration Rate: 6.2 iterations per second
    Loading Time: 12 seconds

    Thanks for this. Now we have a 2070 Super to add to the list, which gives us at least one bench time for all of the Super series of cards (unless I overlooked one). And this one looks quite interesting. The 2070 Super is quite a bit faster than a 2070 (a solid 2 minutes faster), and actually it comes in faster than the 2080 posted here as well. It is not too far off from a 2080 Super we have posted, which is surprising. That is pretty impressive considering the 2070 Super is around $200 less than the 2080 Super.

    Also of note is that this is on an older Z97 motherboard with an i7-4790. Further demonstrating how little those matter.

    Way to make me feel good and bad at the same time. Old hardware indeed...  Just kidding. I thought about investing in updating these, but decided not to bother.

    I am wondering if part of the difference has to do with the SSD or the power supply. I have an 850 in there, which should have been enough for the two 970s, but I question if that was the case. I know available power has little to do with the GPU, but I wonder if the excess keeps it from throttling down when it is under load.

  • outrider42outrider42 Posts: 2,674

     

    I was looking at the price of GPU cards here in NZ yesterday and it looks like the 2080 ti is twice the price of the 2070 Super here. Looking at these numbers, I'm quite surprised to see that the 2070 Super is only a minute slower than the 2080 ti (just over 5 mins vs just over 4). Is that worth paying double? Admittedly, the 2080 ti has 11GB VRAM which is looking more important these days as NVidia & DAZ between them seem to be finding ways to use up all that VRAM leaving less scope for adding characters to the scenes.

    To be fair, the x80ti has historically been a lot more expensive than other cards. Both it and the Titan have never been something you could describe as "price to performance leaders". The 1080ti released at $700, which was roughly twice the cost of the 1070. So the 2080ti costing twice the price of the 2070 and its Super is not so surprising when you think about it. However it cannot be denied that one of the biggest complaints about Turing is how the prices have inflated dramatically over Pascal across the whole lineup. Before Turing launched, I was considering buying a 2080ti, as I owned a 1080ti...until I saw the price announced. Though technically the base price was $1000, Nvidia had the Founder's Edition for $1200, and I instantly knew that most 3rd party vendors would follow closer to the higher $1200 than $1000. So it would be rare to find a 2080ti at that lower price. Either way, even the lower end of that was more than I was willing to spend, and I decided not to buy one. I picked a 2nd used 1080ti instead, which I snagged for about $450. So I was pretty happy with that, and by the time the 3000 series comes along I'll have got my money's worth out of my two 1080tis.

     

    System Configuration
    System/Motherboard: ASRock Z97 Extreme6
    CPU: Intel Core i7-4790 @ 3.6 GHz
    GPU: EVGA RTX GeForce 2070 Super
    System Memory: Corsair Vengence 8GB DDR3 x4
    OS Drive: SanDisk Ultra II 960GB SSD
    Asset Drive: Western Digital WD30EZRX 2.8TB
    Operating System: Windows 10 Pro
    Nvidia Drivers Version: 441.20
    Daz Studio Version: 4.12.0.86

    Benchmark Results
    2019-11-22 11:16:21.026 Finished Rendering
    2019-11-22 11:16:21.068 Total Rendering Time: 5 minutes 1.95 seconds

    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2019-11-22 11:16:31.826 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070 SUPER):      1800 iterations, 6.677s init, 290.227s render

    Iteration Rate: 6.2 iterations per second
    Loading Time: 12 seconds

    Thanks for this. Now we have a 2070 Super to add to the list, which gives us at least one bench time for all of the Super series of cards (unless I overlooked one). And this one looks quite interesting. The 2070 Super is quite a bit faster than a 2070 (a solid 2 minutes faster), and actually it comes in faster than the 2080 posted here as well. It is not too far off from a 2080 Super we have posted, which is surprising. That is pretty impressive considering the 2070 Super is around $200 less than the 2080 Super.

    Also of note is that this is on an older Z97 motherboard with an i7-4790. Further demonstrating how little those matter.

    Way to make me feel good and bad at the same time. Old hardware indeed...  Just kidding. I thought about investing in updating these, but decided not to bother.

    I am wondering if part of the difference has to do with the SSD or the power supply. I have an 850 in there, which should have been enough for the two 970s, but I question if that was the case. I know available power has little to do with the GPU, but I wonder if the excess keeps it from throttling down when it is under load.

    Modern PSUs do a good job of regulating power. Its not a factor unless somebody buys one that is not nearly adequate for their needs. Even then, most power supplies can actually push well beyond their advertised Wattage, that 850 is simply the guaranteed Wattage and a they keep this conservative for safety. You could probably push over a 1000 Watts on that system, though I certainly would never suggest attempting it. The drives only effect the loading speed, so if your assets were also on a SSD like your OS they would shave a few seconds off the load time...but that is all. So they would not effect rendering either. Simply put, all that matters is your GPU. If you use multiple GPUs, THEN and only then do other components start to matter a little bit. So if your i7 is doing fine for you with how you use your PC, then why bother? That i7 can still play games pretty well too without being too much of a bottleneck on high end GPUs, if that matters to you.

    And I know this because I have an i5 4690. So your old CPU still beats mine! I play everything at 1440p with no issues. I do plan on getting a Ryzen sometime. But I am not in a rush. I was going to get a 3900X, but then I bought a house. That kind of changed things, LOL.

    The more likely factors would be the card's cooling and boost clocks. If the 2080ti or 2080 Supers that were benched ran hot or are low end models, they could downclock more during the bench. If your 2070 Super runs really cool, it could maintain its boost for longer, letting it catch up. Sadly we can't factor all of these variables here, it would just be impossible. But even with different cooling we should still see clear trends. Like all the 2080tis tend to bench very similarly, even though each system is wildly different and the 2080tis can be clocked differently. Its also possible you just won the silicone lottery and got a really good performing chip that will hold its boost clock better.

  • I gotcha. I was curious about using those and after watching few videos about how most users are mounting the cards seems like it would take some Macgyvering to get it to physically work. Seems like those adapter cable lengths are capped depending on the pci gen speed you want to operate at. 

    No, not at all... no Macgyvering. I would not have attempted the kind of surgeries I've read about on a $1300 card. It should be very easy, if your case has the extra double bracket. Be sure to get the riser with the 90 degree connector and it is really easy. No guesswork because there's really only one way to do it. But blowers are essential.

    As far as speed vs length, I guess for 8x, 220mm is OK.

    Good luck, and let us know how it turns out.

  • RayDAntRayDAnt Posts: 807
    edited November 2019

    Just finished updating, the tables for those interested. And yeah, that 2070 SUPER score is very interesting considering how close it is to the 2080 SUPER. Although I suspect that has more to do with the current 2080 SUPER score being under-representative of its true relative performance due to software differences (our only 2080 SUPER score so far was supplied by @TheKD running 436.30 drivers on a Windows 7 system.)

    Post edited by RayDAnt on
  • Matt_CastleMatt_Castle Posts: 1,214

    I know you've already got a sample for this card on this version of Iray, but it doesn't hurt to shoot a pattern, and it will give me something to refer back to should I upgrade the system at any point*.

    System Configuration
    System/Motherboard: Hewlett Packard IPM87-MP
    CPU: Intel Core i7 4790 @ stock
    GPU: MSI GTX 1050 Ti LP @ stock
    System Memory: 2x Micron PC3-12800U-11-11-B1 8GB DDR3 @ 1600 MHz
    OS Drive: Samsung 850 EVO 250GB mSATA
    Asset Drive: Seagate Barracuda Green ST2000DL003 2 TB
    Operating System: Windows 8.1 Home 64-bit, Version 6.3, Build 9600
    Nvidia Drivers Version: 441.12
    Daz Studio Version: Daz Studio 4.12.0.086
    Optix Prime Acceleration: On

    Benchmark Results:
    2019-11-27 16:02:52.961 Total Rendering Time: 29 minutes 8.24 seconds
    2019-11-27 16:29:25.477 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce GTX 1050 Ti): 1800 iterations, 4.631s init, 1740.110s render

    Iteration Rate: 1.0344 iterations per second
    Loading Time: 8.13 seconds

     * Unfortunately, for practical reasons I'm limited to low-profile/half-height cards, which means the only viable upgrade right now is the GTX 1650, and making the assumption that (without RT cores) the performance is going to be roughly proportional to gaming benchmarks, that's a disappointingly small margin of improvement (apparently only about 30-40%).

    I had been hoping that Nvidia would announce the GTX 1650 Ti such that I could upgrade to a low profile version of that, but with the announcement of the 1650 Super as a 100 Watt card, it seems unlikely that's going to exist as a low profile card. (Which is disappointing, as with the gaming benchmarks being nearly twice the 1050 Ti, it would have been a wonderful boost).

    So, if there's a passable Black Friday deal on a 1650, I might end up taking it. I know it's not going to be a great Iray card, but the possibility of cutting four hour renders down to three hour renders is still somewhat tempting.

  • timon630timon630 Posts: 33
    edited November 2019

    System Configuration
    System/Motherboard: ASUS LGA1150 B85-PLUS B85
    CPU: Intel Core i5-4570 3.2GHz stock
    GPU: MSI GeForce RTX 2070 ARMOR OC stock
    System Memory: 2x DDR3 8192MB PC12800 1600MHz Kingston HyperX CL10-10-10
    OS Drive: SATA-3 1Tb WD Blue 7200rpm
    Asset Drive: -
    Operating System: Windows 10 Pro 1903
    Nvidia Drivers Version: 441.20
    Daz Studio Version: Daz Studio 4.12.0.086
    Optix Prime Acceleration: On

    Benchmark Results

    First result
    2019-11-30 23:25:30.780 Total Rendering Time: 7 minutes 0.17 seconds
    2019-11-30 23:26:17.194 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070): 1800 iterations, 8.446s init, 407.286s render

    Second result
    2019-11-30 23:49:12.633 Total Rendering Time: 6 minutes 47.96 seconds
    2019-11-30 23:50:03.059 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070): 1800 iterations, 8.192s init, 396.644s render

    #1
    Iteration Rate: 4.419 iterations per second
    Loading Time: 12.884 seconds

    #2
    Iteration Rate: 4.538 iterations per second
    Loading Time: 11.316 seconds

    Post edited by timon630 on
  • timon630timon630 Posts: 33
    edited November 2019

    System Configuration
    System/Motherboard: ASUS LGA1150 B85-PLUS B85
    CPU: Intel Core i5-4570 3.2GHz stock
    GPU: MSI GeForce RTX 2070 ARMOR OC stock
    System Memory: 2x DDR3 8192MB PC12800 1600MHz Kingston HyperX CL10-10-10
    OS Drive: SATA-3 1Tb WD Blue 7200rpm
    Asset Drive: -
    Operating System: Windows 10 Pro 1903
    Nvidia Drivers Version: 441.20
    Daz Studio Version: Daz Studio 4.12.1.40
    Optix Prime Acceleration: On

    Benchmark Results

    First result
    2019-12-01 00:06:32.661 Total Rendering Time: 6 minutes 52.28 seconds
    2019-12-01 00:06:52.530 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070): 1800 iterations, 11.753s init, 396.143s render

    Second result
    2019-12-01 00:16:05.454 Total Rendering Time: 6 minutes 47.23 seconds
    2019-12-01 00:17:23.333 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070): 1800 iterations, 6.808s init, 397.307s render

    #1
    Iteration Rate: 4.544 iterations per second
    Loading Time: 16.137 seconds

    #2
    Iteration Rate: 4.531 iterations per second
    Loading Time: 9.923 seconds

    Post edited by timon630 on
  • outrider42outrider42 Posts: 2,674
    Thanks for posting. So Timon630's 2070 pretty much matches Drip's 2070 with 4.12, which is good, we have some consistent data.

    Thinking about the 2070 Super and 2080 Super, I remember some reviewers not being very impressed by the 2080 Super. I also recall that Puget's testing came up with some very weird results for 2070 and 2080 with Vray. The 2070 actually beat the 2080, and they tested it multiple times. Then when the 2070 Super released, this anomaly repeated itself. The 2070 Super beat the 2080 Super in the Vray bench. https://www.pugetsystems.com/labs/articles/V-Ray-Next-GPU-Roundup-NVIDIA-GeForce-RTX-SUPER-Performance-1548/

    Obviously Vray is not Iray, and these benches really made me question Puget's methods, but there is some precedent for the 2070 and 2080 having odd performance in a rendering engine. I think the general Takeaway is the 2080 Super was not as enhanced as the 2060 and 2070 Supers were compared to their non Super versions. And maybe there is a quirk with the 2080 Super that is keeping it from performing as well as it should for this type of rendering.
  • timon630timon630 Posts: 33
    edited December 2019

    System Configuration
    System/Motherboard: ASUS LGA1150 B85-PLUS B85
    CPU: Intel Core i5-4570 3.2GHz stock
    GPU: Asus GeForce GTX 1070 ROG STRIX strix stock
    System Memory: 2x DDR3 8192MB PC12800 1600MHz Kingston HyperX CL10-10-10
    OS Drive: SATA-3 1Tb WD Blue 7200rpm
    Asset Drive: -
    Operating System: Windows 10 Pro 1903
    Nvidia Drivers Version: 441.20
    Daz Studio Version: Daz Studio 4.12.0.086
    Optix Prime Acceleration: On

    Benchmark Results

    First result
    2019-12-02 06:08:14.454 Total Rendering Time: 12 minutes 16.16 seconds
    2019-12-02 06:09:52.190 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (GeForce GTX 1070): 1800 iterations, 0.735s init, 731.831s render

    Second result
    2019-12-02 06:23:19.469 Total Rendering Time: 12 minutes 16.20 seconds
    2019-12-02 06:23:43.821 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (GeForce GTX 1070): 1800 iterations, 0.656s init, 732.465s render

    #1
    Iteration Rate: 2.460 iterations per second
    Loading Time: 4.329 seconds

    #2
    Iteration Rate: 2.457 iterations per second
    Loading Time: 3.735 seconds

    The 'Iray Viewport' was enabled, so 'Loading Time' is low.

    Post edited by timon630 on
  • timon630timon630 Posts: 33
    edited December 2019

    System Configuration
    System/Motherboard: ASUS LGA1150 B85-PLUS B85
    CPU: Intel Core i5-4570 3.2GHz stock
    GPU: MSI GeForce RTX 2070 ARMOR OC stock + Asus GeForce GTX 1070 ROG STRIX strix stock
    System Memory: 2x DDR3 8192MB PC12800 1600MHz Kingston HyperX CL10-10-10
    OS Drive: SATA-3 1Tb WD Blue 7200rpm
    Asset Drive: -
    Operating System: Windows 10 Pro 1903
    Nvidia Drivers Version: 441.20
    Daz Studio Version: Daz Studio 4.12.0.086
    Optix Prime Acceleration: On

    Benchmark Results

    First result
    2019-12-02 06:33:03.348 Total Rendering Time: 4 minutes 47.93 seconds
    2019-12-02 06:33:13.742 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070): 1145 iterations, 7.761s init, 276.849s render
    2019-12-02 06:33:13.742 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (GeForce GTX 1070): 655 iterations, 8.248s init, 276.268s render

    Second result
    2019-12-02 06:39:01.892 Total Rendering Time: 4 minutes 47.9 seconds
    2019-12-02 06:39:12.325 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (GeForce RTX 2070): 1145 iterations, 7.891s init, 275.921s render
    2019-12-02 06:39:12.325 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (GeForce GTX 1070): 655 iterations, 8.541s init, 275.113s render

    #1
    Iteration Rate: 6.502 iterations per second
    Loading Time: 11.081 seconds

    #2
    Iteration Rate: 6.523 iterations per second
    Loading Time: 11.979 seconds

    Post edited by timon630 on
  • LenioTGLenioTG Posts: 2,069
    edited December 2019

    I'm deciding if I should buy a RTX 2060 Super or a RTX 2070 Super, and since here there are some results for both of them, I'd like to see how my 2060 behaves in this test.
    I've installed this Public Build just to test this out, to compare the two Daz Studio versions.

    Congratulations on this thread RayDAnt, it looks very professional :)
    Thanks for allowing us to gather this data.

    ____________________________________________________________________________

    System Configuration
    System/Motherboard: Gigabyte B450M AORUS M
    CPU: Ryzen 5 3600 @ Stock speed
    GPU: RTX 2060 EVGA Xc @ Stock speed
    System Memory: 32Gb (4x8) HyperX Fury @ 3000Mhz
    OS Drive: Samsung 970 Evo 250Gb
    Asset Drive: Samsung 860 1TB
    Operating System: Windows 10 build 18362
    Nvidia Drivers Version: 441.28
    Daz Studio Version: 4.12.0.86 Pro Edition (64-bit)
    Optix Prime Acceleration: OFF (Daz Studio 4.12.1.086 or earlier only)

    Benchmark Results
    DAZ_STATS
    2019-12-04 12:01:18.051 Finished Rendering
    2019-12-04 12:01:18.093 Total Rendering Time: 8 minutes 2.89 seconds
    IRAY_STATS
    Sorry, can't find that, but I've found this:

    2019-12-04 12:01:17.444 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Received update to 01800 iterations after 479.971s.
    2019-12-04 12:01:17.450 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Maximum number of samples reached.

    Iteration Rate: 3,750
    Loading Time: around 3.074s (1st iteration finished) I guess, sorry

    ____________________________________________________________________________

    System Configuration
    System/Motherboard: Gigabyte B450M AORUS M
    CPU: Ryzen 5 3600 @ Stock speed
    GPU: RTX 2060 EVGA Xc @ Stock speed
    System Memory: 32Gb (4x8) HyperX Fury @ 3000Mhz
    OS Drive: Samsung 970 Evo 250Gb
    Asset Drive: Samsung 860 1TB
    Operating System: Windows 10 build 18362
    Nvidia Drivers Version: 441.28
    Daz Studio Version: 4.12.1.40 Pro Edition Public Build (64-bit)
     

    Benchmark Results
    DAZ_STATS
    2019-12-04 12:25:53.113 Finished Rendering
    2019-12-04 12:25:53.151 Total Rendering Time: 8 minutes 8.28 seconds
    IRAY_STATS
    Again, I can't manage to find that

    2019-12-04 12:17:53.531 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Received update to 00001 iterations after 6.425s.
    2019-12-04 12:25:52.524 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Received update to 01800 iterations after 485.418s.
    2019-12-04 12:25:52.530 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Maximum number of samples reached.
    Iteration Rate: 3,708
    Loading Time: around 6.425s

    ____________________________________________________________________________

    So, in my test, the 2 versions are basically identical, with the only difference being that the Beta started to render 3 seconds after.

    ____________________________________________________________________________

    This is basically the only way to understand Iray GPU performance.

    So, compared to the existing test, my 2060 was 7% faster (3,750 vs 3,505 iterations per second) in the stable build.

    Yet, a 2060 Super achieved 4,444, +18,5% compared to my RTX 2060.
    And the 2070 Super achieved 6,202, +65% compared to my RTX 2060, and +40% compared to the 2060 Super.
    The RTX 2080 and 2080 Super didn't go much faster.

    Post edited by LenioTG on
  • RayDAntRayDAnt Posts: 807
    edited December 2019

    System Configuration
    System/Motherboard: Microsoft Surface Book 2
    CPU: Intel i7-8650U @ stock
    GPU: Nvidia GTX 1050 2GB @ stock
    System Memory: 16GB DDR3 @ 1867Mhz
    OS Drive: Samsung OEM 512GB NVME SSD
    Asset Drive: Sandisk Extreme 1TB External SSD
    Operating System: W10 1903
    Nvidia Drivers Version: 441.41
    Daz Studio Version: 4.12.1.040 Beta x64

    Benchmark Results (GTX 1050 2GB)
    Total Rendering Time: 38 minutes 41.83 seconds
    CUDA device 0 (GeForce GTX 1050): 1800 iterations, 7.762s init, 2310.598s render
    Iteration Rate: (1800 / 2310.598) = 0.779 iterations per second
    Loading Time: ((0 + 2280 + 41.83) - 2310.598) = 11.232 seconds

     


     

    System Configuration
    System/Motherboard: Gigabyte Z370 Aorus Gaming 7
    CPU: Intel i7-8700K @ stock (MCE enabled)
    GPU: Nvidia Titan RTX @ stock (watercooled)
    System Memory: Corsair Vengeance LPX 32GB DDR4 @ 3000Mhz
    OS Drive: Samsung Pro 970 512GB NVME SSD
    Asset Drive: Sandisk Extreme Portable SSD 1TB
    Operating System: Windows 10 Pro 1903
    Nvidia Drivers Version: 441.41
    Daz Studio Version: 4.12.1.040 Beta x64
     

    Benchmark Results (Titan RTX: TCC mode)
    Total Rendering Time: 3 minutes 49.13 seconds
    CUDA device 0 (TITAN RTX): 1800 iterations, 2.180s init, 224.419s render
    Iteration Rate: (1800 / 224.419) = 8.021 iterations per second
    Loading Time: ((0 + 180 + 49.13) - 224.419) = 4.711 seconds

    Benchmark Results (Titan RTX: TCC mode + i7-8700K) - ignore, needs retesting
    Total Rendering Time: 3 minutes 51.59 seconds
    CUDA device 0 (TITAN RTX): 1700 iterations, 2.258s init, 225.493s render
    CPU: 100 iterations, 2.131s init, 226.884s render
    Iteration Rate: (1800 /226.884) = 7.934 iterations per second
    Loading Time: ((0 + 180 + 51.59) - 226.884) = 4.706 seconds

    Benchmark Results (i7-8700K) - ignore, needs retesting
    Total Rendering Time: 1 hours 25.19 seconds
    CPU: 1800 iterations, 2.375s init, 3620.264s render
    Iteration Rate: (1800 / 3620.264) = 0.497 iterations per second
    Loading Time: ((3600 + 0 + 25.19) - 3620.264) = 4.926 seconds

     

    Post edited by RayDAnt on
Sign In or Register to comment.