Iray Starter Scene: Post Your Benchmarks!

1171820222349

Comments

  • Nyghtfall3DNyghtfall3D Posts: 760

    Until yesterday, I've been running dual 980 Ti's since last November.  I swapped one of them out for a new 1080 Ti, which is also now my display driver.  These are my new numbers, without having switched to Iray Draw Mode in the Viewport first:

    980 Ti: 2 min, 48 sec
    1080 Ti: 2 min, 4 sec
    1080 Ti + 980 Ti: 1 min, 23 sec
    CPU + 1080 Ti + 980 Ti: 1 min, 28 sec

    The last benchmark doesn't surprise me.  I only ran it to be thorough.  My CPU is fairly old by computing standards - Core i7-4770K - but since I don't use it for Iray, the numbers are irrelevant to me.

    Suffice it to say, I am very pleased with my purchase.  Iray now has ample VRAM to work with, and my art will render even faster. :)

  • prixatprixat Posts: 1,585

     

    Nyghtfall said:
    Just so I'm clear, did you first have Iray turned on in your viewport? Or was your viewport switched to the basic shader? These tests should be done with Iray active in the viewport first. 

    There's no mention of that requirement in SY's original post.  Why would it make a difference on render time?

     

    It depends if you're measuring the performance of the system architecture or trying to isolate the performance of the GPU.

    If you want to include the pre-load time you should also include your motherboard, memory and overclock etc. specs.

    Changing the sampling rate and save interval also made a significant improvement on my times, for the same reason... as it sent back its results less often, there were fewer accesses to main memory and all the overhead that involved.

     

    try-this.jpg
    564 x 582 - 188K
  • PA_ThePhilosopherPA_ThePhilosopher Posts: 1,039
    edited June 2017
    Nyghtfall said:
    Just so I'm clear, did you first have Iray turned on in your viewport? Or was your viewport switched to the basic shader? These tests should be done with Iray active in the viewport first. 

    There's no mention of that requirement in SY's original post.  Why would it make a difference on render time?

    It is closer to real-world.

    It is common practice, when working with Iray, to switch it on in the viewport prior to render, in order to punch out renders more quickly. It saves a good amount of time. Most power-users do this unconsciously, without even thinking about it.

    -P

    Post edited by PA_ThePhilosopher on
  • BobvanBobvan Posts: 2,652
    edited June 2017

    i7-6950X (10-core CPU) 1080ti

    1 Min 59 sec GPU 1 Min 53 Sec GPU plus CPU

     

    5 Mins 18 seconds on my ROG Asus laptop (specs attached) GTX980M

    Untitled.jpg
    543 x 271 - 26K
    Post edited by Bobvan on
  • Nyghtfall3DNyghtfall3D Posts: 760
    Nyghtfall said:
    Just so I'm clear, did you first have Iray turned on in your viewport? Or was your viewport switched to the basic shader? These tests should be done with Iray active in the viewport first. 

    There's no mention of that requirement in SY's original post.  Why would it make a difference on render time?

    It is closer to real-world.

    It is common practice, when working with Iray, to switch it on in the viewport prior to render, in order to punch out renders more quickly. It saves a good amount of time. Most power-users do this unconsciously, without even thinking about it.

    -P

    Well, then, 1 min, 4 sec with my 1080 Ti + 980 Ti.

  • BobvanBobvan Posts: 2,652

    Nyghtfall said:

    Just so I'm clear, did you first have Iray turned on in your viewport? Or was your viewport switched to the basic shader? These tests should be done with Iray active in the viewport first. 

    There's no mention of that requirement in SY's original post.  Why would it make a difference on render time?

    It is closer to real-world.

    It is common practice, when working with Iray, to switch it on in the viewport prior to render, in order to punch out renders more quickly. It saves a good amount of time. Most power-users do this unconsciously, without even thinking about it.

    -P

    Well, then, 1 min, 4 sec with my 1080 Ti + 980 Ti.

    You got your 1080 ti cool beans

  • SedorSedor Posts: 1,764

    With active Iray at a  viewport: 21seconds

  • Nyghtfall3DNyghtfall3D Posts: 760
    Bobvan said:
    You got your 1080 ti cool beans

    Yup, the Enchanted Ballroom and a fully-clothed V7 prompted that $780 decision.  GPU-Z kept reporting max VRAM usage with Iray.  Last week, I learned that Octane uses both VRAM and RAM, so I spent seven hours with the demo.  My test scene crashed the demo, twice, taking Studio with it.  Since Iray is built into Studio, and didn't crash, I bought a 1080 Ti instead of an Octane license.  What a no-brainer that was.  I also much prefer how DAZ integrated Iray's shader properties into Studio's Surfaces pane.

     

  • PA_ThePhilosopherPA_ThePhilosopher Posts: 1,039
    edited June 2017
    Sedor said:

    With active Iray at a  viewport: 21seconds

    @Sedor ... you can't leave us handing like that. What are your system specs? Three, four 1080 Ti's? 

    -P

    Post edited by PA_ThePhilosopher on
  • BobvanBobvan Posts: 2,652
    Nyghtfall said:
    Bobvan said:
    You got your 1080 ti cool beans

    Yup, the Enchanted Ballroom and a fully-clothed V7 prompted that $780 decision.  GPU-Z kept reporting max VRAM usage with Iray.  Last week, I learned that Octane uses both VRAM and RAM, so I spent seven hours with the demo.  My test scene crashed the demo, twice, taking Studio with it.  Since Iray is built into Studio, and didn't crash, I bought a 1080 Ti instead of an Octane license.  What a no-brainer that was.  I also much prefer how DAZ integrated Iray's shader properties into Studio's Surfaces pane.

    Good for you man!

     

  • Nyghtfall3DNyghtfall3D Posts: 760
    edited June 2017

    Thank you.  I... could not... be happier.  :)

    EDIT: Though I would certainlly love to throw another one in there at some point, just to even out my setup with x2 1080 Ti's.  ;)

    Post edited by Nyghtfall3D on
  • BobvanBobvan Posts: 2,652
    Nyghtfall said:

    Thank you.  I... could not... be happier.  :)

    I know the feeling..

  • Toonces said:

    Can't wait to see when you get the 3rd 1080 in there. Gotta put that water cooling system to good use!

    @Toonces

    I am really looking forward to putting a third and fourth 1080 Ti in my rig. If I ever was a believer in the 1080 Ti, I am even more now, having just installed two of them. It not only speeds up renders, but everything feels a bit faster. Scenes load faster, turning on Iray in the viewport is faster. I never had any reason to upgrade, prior to the release of the 1080 Ti. I was happy with my quad 780 Ti's. But these... its such an obvious upgrade now...

  • BobvanBobvan Posts: 2,652
    edited June 2017
    Toonces said:

    Can't wait to see when you get the 3rd 1080 in there. Gotta put that water cooling system to good use!

    @Toonces

    I am really looking forward to putting a third and fourth 1080 Ti in my rig. If I ever was a believer in the 1080 Ti, I am even more now, having just installed two of them. It not only speeds up renders, but everything feels a bit faster. Scenes load faster, turning on Iray in the viewport is faster. I never had any reason to upgrade, prior to the release of the 1080 Ti. I was happy with my quad 780 Ti's. But these... its such an obvious upgrade now...

    4 1080ti's insane

    Post edited by Bobvan on
  • GaryHGaryH Posts: 66
    edited June 2017

    How about some Titan Xp benches?

    Phase 1 of me maxing out my uber stable Z68 32 GB Intel 2600K (4GHz OC) from 2011 for Iray rendering is complete.  Phase 2 will involve externally watercooling the GPUs (using a MO-RA3) to dramatically lower the noise while rendering.

    I started out with 2 x GTX 970 Strix and now use just one for the display.  I then added a Titan X Pascal and just today, the Titan Xp.

    Some things to note about the benchmarks.  First the good news for those with older systems.  You don't need PCI-E 3.0 lanes to have good performance and you don't need to be running your render GPUs in x16, or even x8 mode.  The 2600K has only 16 PCI-E 2.0 lanes and 4 more are avaiable via the Z68 chip set.  So the benchmarks below were run with the Titan Xp in a 2.0 x8 slot and the Titan X Pascal in a 2.0 x4 slot running off the Z68 via DMI to the 2600K.  When I run that slot in x4 mode I loose my front panel USB 3.0 ports  I can also run that slot in 2.0 x1 mode and not loose my USB ports with minimal impact on rendering performance.  Optix was on and these were run to the default 95% convergence. 

    Power draw tops out at about 563 watts during these short - and also long Iray renders (the GTX 970 Strix is NOT rendering and is not contributing much to the load).

    Titan Xp (initial run):  1 minute 57.12 seconds

    2017-06-09 15:33:51.178 Total Rendering Time: 1 minutes 57.12 seconds
    2017-06-09 15:38:39.521 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Device statistics:
    2017-06-09 15:38:39.521 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 0 (TITAN Xp): 4872 iterations, 13.997s init, 101.771s render

     

    Titan Xp (pre-loaded):  1 minute 43.67 seconds

    2017-06-09 15:37:36.360 Total Rendering Time: 1 minutes 43.67 seconds
    2017-06-09 15:38:41.148 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Device statistics:
    2017-06-09 15:38:41.148 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 0 (TITAN Xp): 4855 iterations, 0.316s init, 102.293s render

     

    Now with the Titan X Pascal in the 2.0 x4 mode slot.

    Titan Xp and Titan X Pascal (initial run):  1 minute 10.06 seconds

    2017-06-09 13:53:33.334 Total Rendering Time: 1 minutes 10.6 seconds
    2017-06-09 13:59:55.424 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Device statistics:
    2017-06-09 13:59:55.425 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 0 (TITAN Xp): 2582 iterations, 0.371s init, 54.618s render
    2017-06-09 13:59:55.425 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 2 (TITAN X (Pascal)): 2319 iterations, 0.370s init, 55.588s render

     

    Titan Xp and Titan X Pascal (pre-loaded):  57.15 seconds

    017-06-09 13:58:43.280 Total Rendering Time: 57.15 seconds
    2017-06-09 13:59:54.300 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Device statistics:
    2017-06-09 13:59:54.300 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 0 (TITAN Xp): 2592 iterations, 14.351s init, 53.976s render
    2017-06-09 13:59:54.300 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 2 (TITAN X (Pascal)): 2309 iterations, 14.312s init, 54.208s render

     

    An even better benchmark is to run a larger, more complicated scene, especially when attempting to understand load times on x4, and even x1 mode slots.  I have some results for Jack Tomalin's "Dine on the Orient Express" that should fit the bill.

     

     

    Sickleyield Titan Xp benchmark (pre-loaded).png
    2560 x 1440 - 1M
    Post edited by Chohole on
  • junkjunk Posts: 1,229
    garyh.pub said:

    How about some Titan Xp benches?

     

    Heck yeah!  Very detailed and thorough benchmarking.  Thank you very much for this.

  • junkjunk Posts: 1,229
    edited June 2017

    Wait.  Just noticed you have only a 850 Watt power supply powering all of this.  I know the Kill-A-Watt is only showing 563 Watt usage but if you fired up all of the GPU's to 100% that seems dangerousely close to a full on melt down. Especially if you accidentally include the CPU in the mix.

    Post edited by junk on
  • AtiAti Posts: 9,081

     i7-4770K, 16G RAM, CPU only (6 threads): 53 minutes 10.75 seconds
    GeForce GTX 760, 2G, GPU only: 9 minutes 32.39 seconds
    CPU (6 threads) + GPU: 11 minutes 43.64 seconds

     

  • GaryHGaryH Posts: 66
    edited June 2017
    junk said:

    Wait.  Just noticed you have only a 850 Watt power supply powering all of this...

    It's more than adequate for the way I use my system.  The 970 Strix is just for the display, as I still want to be able to use my display if I have a background render going.  It's also been well establishied that adding the CPU to the mix is worthless.  In fact, in should be noted that CPU usage seems to increase fairly linearly as you add GPUs to the render.  My observations have been around 17% for one GPU, 34% for two, and around 60% for three.  This is for a four core, 8 thread CPU.

    Again, this is NOT how I would render, but out of curiosity I rendered Jack Tomalin's "Dine on the Orient Express" Antique Brass for 4000 iterations.  This is a complex model that takes just over a minute to load.  Lots of light sources and big textures.  It just fits on the 970.  Kill-A-Watt averaged around 660 watts, with 675 being the peak I observed during the 6 minute or so render.  The GPUs were maxed out (as best as can be done on air) and 60% on the overclocked CPU.  No "full on meltdown" here! wink

    Also note the Titan X Pascal was running in PCI-E 2.0 x1 mode but it did not negatively effect the init or render time.

    2017-06-12 21:56:46.771 Total Rendering Time: 5 minutes 48.88 seconds
    
    2017-06-12 22:00:53.011 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Device statistics:
    2017-06-12 22:00:53.011 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 1 (GeForce GTX 970): 298 iterations, 63.506s init, 276.583s render
    2017-06-12 22:00:53.011 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 0 (TITAN Xp): 1965 iterations, 63.470s init, 277.081s render
    2017-06-12 22:00:53.011 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 2 (TITAN X (Pascal)): 1737 iterations, 63.488s init, 276.813s render
    Dine on the Orient Express (970 Strix, Titan Xp, Titan X, 4000).png
    2560 x 1440 - 2M
    Post edited by GaryH on
  • GaryHGaryH Posts: 66
    edited June 2017

    - deleted -

    Post edited by GaryH on
  • junkjunk Posts: 1,229

    Here is more information on the headless GTX 1080 cards that will go for about $350.  Yes they are tuned for crypto-mining but I bet they would work beautifully for Daz3D rendering.  Would love to see these benchmarked for rendering when available.

    http://wccftech.com/nvidia-pascal-gpu-cryptocurrency-mining-price-specs-performance-detailed/

  • greymouser69greymouser69 Posts: 499

    Finally got a laptop with nvidia, only a 3gb gtx 1060 but still something, so here is my 1060 + cpu:

    2017-06-22 07:31:26.836 Total Rendering Time: 6 minutes 59.42 seconds

  • Single MSI GTX 1080ti GamingX 11G (no CPU): 2 minutes 1 Second.

    Is there any way to keep the history up after the render finishes or to show it again?

  • ZarconDeeGrissomZarconDeeGrissom Posts: 5,412
    edited July 2017

    Uh, well there is that log file. Copy and paste the text into another document with notes I guess.

    The log file did not show all the render stats at one time, I'm not sure how much is logged now. Tho the log does capture the important bits, render time. It will look something like this.

    (That was beta, and a very long time ago on CPU only I think, lol)

    Oh, garyh.pub   I don't care what Paul and Kyle sais about cable management and ketchup and mustard colors, That is a Beautiful triple-card setup.

    LogFileView_01001.png
    760 x 360 - 46K
    Info_Is_On_Log_001.png
    1163 x 334 - 28K
    Post edited by ZarconDeeGrissom on
  • System Specs:

    AMD Phenom II X4 955 Black Edition @ 4,6 Ghz
    Asus Crosshair V Formula-Z
    16GB DDR3-2133 GSkill RipJawsX
    2x Evga GTX780 Superclocked

    Windows 8.1 Professional 64


    GPU-Only (with OptiX Prime):
    2017-07-06 13:03:44.654 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Device statistics:
    2017-07-06 13:03:44.654 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 0 (GeForce GTX 780):      2440 iterations, 28.417s init, 143.314s render
    2017-07-06 13:03:44.654 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 1 (GeForce GTX 780):      2405 iterations, 28.424s init, 142.535s render
    2017-07-06 13:03:33.750 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Received update to 04799 iterations after 170.473s.
    2017-07-06 13:03:35.141 Finished Rendering
    2017-07-06 13:03:35.219 Total Rendering Time: 2 minutes 53.65 seconds


    GPU-Only (without OptiX Prime):
    2017-07-06 13:22:13.811 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Device statistics:
    2017-07-06 13:22:13.811 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 0 (GeForce GTX 780):      2425 iterations, 28.509s init, 287.087s render
    2017-07-06 13:22:13.811 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : CUDA device 1 (GeForce GTX 780):      2446 iterations, 28.512s init, 288.054s render
    2017-07-06 13:19:52.342 Iray INFO - module:category(IRAY:RENDER):   1.0   IRAY   rend info : Received update to 04831 iterations after 314.435s.
    2017-07-06 13:19:54.655 Finished Rendering
    2017-07-06 13:19:54.717 Total Rendering Time: 5 minutes 18.69 seconds


    I didn't render the scene at CPU-Only, for obvious reasons. As soon as I have the time, I'm going to test this with my main system (i7 4790k @ 4,9 Ghz, 32GB DDR3-1866, 2x RX480 Strix), obviously without GPU. ;)

  • ebergerlyebergerly Posts: 3,255
    edited July 2017

    This is pretty much old news, and matches almost exactly with what others have posted for the GTX 1070, but here goes:

    CPU: i7 6700 (4 core, 8 logical CPU's)

    Memory: 48GB DDR4 2133

    GPU: NVIDIA GTX 1070 8GB Founders Edition

    OptiX on 

    Total Rendering Time (GPU only): 3 minutes 14 seconds (first render after opening scene)

                                          (CPU only): I gave up after 8 minutes and 92%.....

    Max system power consumption during render: 180 watts (GPU render) 

                                                                                       110 watts (CPU render)

    Post edited by ebergerly on
  • eicmareicmar Posts: 4

    CPU: Intel i5-2500k

    Memory: 16 GB

    GPU: MSI nVidia GTX 970

    OptiX on

    Render Time: 5 min 41 sec

     

    I will get a new computer next week with a new GTX 1080 TI and I was thinking about to keep my GTX 970 for driving my monitor and small scenes. Do I have to keep something special in mind for this (i.e. PCI-e slot order, ...)?

  • GaryHGaryH Posts: 66

    I also have a 970 driving my monitor - it's not used for rendering.  I'd put the 970 in your fastest PCI-E slot (usually the top one) since it's driving your monitor.

  • Just a dumb question: how can I see how long the last render took? Can't find anywhere, I just installed the second 1070 and I want to give it a try

  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,126

    Just a dumb question: how can I see how long the last render took? Can't find anywhere, I just installed the second 1070 and I want to give it a try

    https://www.daz3d.com/forums/discussion/comment/2592811/#Comment_2592811 ? ;)

Sign In or Register to comment.