iRAY Animation?

2»

Comments

  • MEC4DMEC4D Posts: 5,249

    I agree on that , if you have low end card with small VRam , CPU will be your insurance just in case it does not fit on the card preventing iray from crash .

    Iray don't use full power of every device you use for rendering for that reason there is zero benefit of using CPU on top with good gtx cards , 1 titan X running  100% , 2 x Titans X 145% , 3 xTitan X 165% 

    4 x Titan X running only 185% faster than 1 x Titan X so the CPU only slow down everything even more , how more devices you use how less power will be used for the money you spend  ( different story for the real time iray viewport ) but for rendering nope  and when you render animation sequences every winning minute per frame is golden and you can save many hours of rendering by switching the CPU off if you can 

    also about the CPU cores , if you run 4 Titan X on CPU with 4 cores or 20 cores there is not difference in rendering time as iray will benefit only from the 4 cores in this case and ignore the other 16

    when I render complex scene only 2 cores are used and 4 threads for loading up the scene  the other 2 cores anx 4 threads are idle unless I use only CPU for rendering 

    for that reason I don''t see here any benefit of using CPU with high end gtx cards so unless you plan to use more than 4 GPU cards in your system your 4 core CPU processor will be the best choice 

     

    y

    MEC4D said:

    Yes it does if you render only with CPU, on the other hand CPU slow down rendering time when selected together with GPU  and that is a fact at last when in combination with GTX cards, faster CPU load the scene faster to the card  and that is but after that it get back to idle and there are no activity while rendering 

    overclocking is not recomended if you don't know your system limits so only for advanced users . 

    JCThomas said:
    MEC4D said:

     

    CPU does affect Render speeds in Iray. 

     

    And while Mec4D has managed to make overclocking work, it is still not recommended by Daz 3D as it has a very serious chance of making your CPU unstable and crashing software. 

     

    As a clarification, having multiple devices increases load time because the scene has to be loaded to each device. However having the CPU rendering in addition to GPU's decreses actual render time. Having multiple devices may increase your overall time, which includes both, but that definitely will depend on the scene you are rendering, what percentage of the total time each component is and your actual hardware.

    A slow small video card with limited number of CUDA cores may not be worth adding to a render, and the same applies to a slow limited core CPU, because the speed increase of the render may not offset the additional render speed of the device.  This can be especially true if you are not rendering photoreal to high convergence percentage. 

    Complex photoreal renders to a high convergence will generally benefit from having every device that can hold hte scene, rendering it, presuming that you are not using hte computer for other things while rendering and speed of render is your goal. :) 

     

  • FrankTheTankFrankTheTank Posts: 1,513

     

    I agree that using the CPU slows things down significantly when using multi-GPU, I have tested this extensively myself and found it to be true.

    But I don't understand this part:

    "Iray don't use full power of every device you use for rendering ...  1 titan X running  100% , 2 x Titans X 145% , 3 xTitan X 165% " then at the end you say, "for that reason I don''t see here any benefit of using CPU with high end gtx cards so unless you plan to use more than 4 GPU cards in your system your 4 core CPU processor will be the best choice "

    When I started iray animation a few months ago, I had only 1 GPU, a Geforce 980ti and I did testing with and without CPU enabled for render. With 1 GPU, I found the CPU gave a very slight decrease in render time (but was not worth it, nonetheless, due to other problems that causes, rendering CPU for 80 hours straight, stress on CPU, heat, tying up system resources, etc). But when I added the 2nd GPU, renders were significantly slowed down if the CPU was enabled for renders. So of course all my rendering is done with GPUs and Optix enabled only. I never use the CPU.

    Also, with 1 GPU, the load on the card is around 99% (not memory load), the GPU load.

    When I run both cards, the GPU utilization is about 95% on both cards, so they are both getting used to almost full capacity.

    So there's a very slight drop in utilzation when going from 1 GPU to 2, but not as significant as you say, at least not in my experience.And I don' see any reason why anyone with a decent GPU would ever want to enable CPU for rendering iray, especially, the more cards they have, the more significant the negative impact of having the CPU enabled would be. The CPU becomes a bottleneck very quickly, I think even someone with a lowend CUDA card would be better off not using CPU. 512 cores vs 4 cores, still no contest.

  • JoepingletonJoepingleton Posts: 746
    edited May 2016

    I have a question about using CPU to back up GPU rendering in Iray. I am running a i7-37700 CPU @ 3.5GHz with 32 GB of RAM with a NVIDIA GeForce 970 card.

     I had been using the default CPU method for all my renders up until a few weeks ago when my CPU cooling system broke. Since replacing the cooling system I started looking into how everything was operiating with DAZ Studio.  By default it seems to use the CPU first and then use the GPU (rarely).  Is there a way to have Iray use the GPU and then if needed the CPU? 

    Post edited by Joepingleton on
  • JCThomasJCThomas Posts: 254
    MEC4D said:

    I am using 4790K , I tried higher speed but it was overheating too quick  with 135W power consumption , with the 240mm radiator I can do that better , when rendering animations it load faster the frame updates to the card when overclocked  and does have the time to slightly cool off between frames . For one render not worthy but for a lot of renders the extra speed save a lot of time , the same with the GPU , if you set the Maxium Power under Nvidia Panel it will render faster the sequences since the clock will be in stand by mode all the time,  again not worthy for 1 render .

    JCThomas said:
    MEC4D said:

    i7 4Ghz overclocked 19% to 5Ghz water cooled, I use Corsair Hydro Series™ H80i V2 Water / Liquid CPU Cooler. with 140 mm extra thick radiator 2 fans  . CPU have nothing to do with rendering when using GPU , all it does only it load the scene to the GPU so faster CPU will only load faster the scene to your card but not improve rendering, so if you have 4 cores or 10 it does not matter for iray software , 4 cores i7 are optimal to run 4 x Titans X with iray .

    CPU Voltage max 1.344 per core max power mode on full load temp. 50-52 C ,  power draw from 16w to max 95 w depends of the mode , Auto , Silent or Performance etc since I use Asus motherboard software to control all cooling system in my rig depends of what I need ..   ,  temp for cards 35-37 C per each GPU on full load .

     CPU and each GPU have own closed water cooling system independent from each other  as normally I just use half of the rig power if not rendering so half of the radiator fans are off . so far that is my best rig ever with the lowest temp I ever had . But going to build external card case this year when the additional 2 x 1080 arrive so all my 4 cards going outside the main rig so I have more space on top for for my CPU new cooling Corsair Hydro Series H100i V2 Extreme Performance Water / Liquid CPU Cooler. 240mm , right now the place is occupied by GPU radiators 

    JCThomas said:

    Off topic, but what i7 CPU did you get to 5 Ghz and what kind of cooling are you using? I'd love to know your voltages, temps at load, and power draw, if you've got a chance and the info is handy. If not, don't sweat the extra stuff...just especially curious about the CPU and cooling. Thanks.

     

    Thanks for the detailed reply. Right, I had been aware that the CPU  doesn't affect the render times in IRay or octane. I just haven't been able to get my CPU stable at 4.7, although I've been unwilling to take my voltages that high. Are you using a 6700k or a 4790k?

     

    Was that 1.344v the result of an auto-profile in your motherboard, or did you manually input it? You lucked out on the silicon lottery, being able to hit 5 ghz, that's awesome. I ask about the auto-profile (like an auto overclock utility or something, like some ASUS boards have) because all of my research has indicated that 1.3v is the max recommended safe voltage for the 4790k. If you'd seen that somewhere before and were willing to risk it, cool, power to ya. If not, you might consider lowering the voltages and decreasing the clockspeed a bit.

    I hope I don't appear to be second-guessing you. I'm just saying that if you overclocked your CPU by typing "5.0" into a multiplier setting in your BIOS and then let your motherboard determine the required voltages, you might have ended up with higher voltages than you wanted / are safe.

  • MEC4DMEC4D Posts: 5,249

    It seems that both card run on full power because the scene is loaded on both and the clock is running but they are not used efficient in the same ratio by iray software while rendering with multiple GPUs for that reason Octane is faster than Iray and Blender Cycles slower than iray due to the rendering software no the cards or drivers 

    the results was from benchmarks for rendering with iray software . The same was before with Octane until the software got fixed , Blender Cycles use even less resource the worse speed of them all in unbiased GPU rendering 

    you may thinking 2 cards give you 200% power but that is not the case, only the first card is at 100% additional cards not so 4 x Titan X will give me only 186% not 400% rendering power plus all also depends of what you render and how big the scene is .  That why Games works best with  SLI so they can suck the full power of the GPU

    try this, Render simple cube with CPU, then switch just to one GPU and after to more GPUs if you have you will see that the rendering time is the same for all and more cards don't do nothing to the speed of this single scene , now trow in 2 genesis with sss shader and you starting to see the speed difference .. how more complex the scene how  better iray use the resources but never 100% for each additional GPU

    in the article it was explained how the iray software handle the data  , but I lost the link some how , I only remember the utilization of  the CPU by iray  , I was searching for info about best CPU to use with 4 Titan X for rendering with iray but 4 core CPU was the answer according to iray render benchmarks with various cards , well Quadro cards was the worse with iray speed rendering gtx performed better .

      

     

    I agree that using the CPU slows things down significantly when using multi-GPU, I have tested this extensively myself and found it to be true.

    But I don't understand this part:

    "Iray don't use full power of every device you use for rendering ...  1 titan X running  100% , 2 x Titans X 145% , 3 xTitan X 165% " then at the end you say, "for that reason I don''t see here any benefit of using CPU with high end gtx cards so unless you plan to use more than 4 GPU cards in your system your 4 core CPU processor will be the best choice "

    When I started iray animation a few months ago, I had only 1 GPU, a Geforce 980ti and I did testing with and without CPU enabled for render. With 1 GPU, I found the CPU gave a very slight decrease in render time (but was not worth it, nonetheless, due to other problems that causes, rendering CPU for 80 hours straight, stress on CPU, heat, tying up system resources, etc). But when I added the 2nd GPU, renders were significantly slowed down if the CPU was enabled for renders. So of course all my rendering is done with GPUs and Optix enabled only. I never use the CPU.

    Also, with 1 GPU, the load on the card is around 99% (not memory load), the GPU load.

    When I run both cards, the GPU utilization is about 95% on both cards, so they are both getting used to almost full capacity.

    So there's a very slight drop in utilzation when going from 1 GPU to 2, but not as significant as you say, at least not in my experience.And I don' see any reason why anyone with a decent GPU would ever want to enable CPU for rendering iray, especially, the more cards they have, the more significant the negative impact of having the CPU enabled would be. The CPU becomes a bottleneck very quickly, I think even someone with a lowend CUDA card would be better off not using CPU. 512 cores vs 4 cores, still no contest.

     

  • MEC4DMEC4D Posts: 5,249

    No way ..that why it slow down the GPU as the primary become CPU as it is slower  and need more utilization so it run at 100% and then the GPU is used less than 100% that why iray render slower, you should use just GPU if you sure your scene fit well on the card plus you need some memory left for the rendering  if you not sure use CPU +GPU just in case 

    I have a question about using CPU to back up GPU rendering in Iray. I am running a i7-37700 CPU @ 3.5GHz with 32 GB of RAM with a NVIDIA GeForce 970 card.

     I had been using the default CPU method for all my renders up until a few weeks ago when my CPU cooling system broke. Since replacing the cooling system I started looking into how everything was operiating with DAZ Studio.  By default it seems to use the CPU first and then use the GPU (rarely).  Is there a way to have Iray use the GPU and then if needed the CPU? 

     

  • JoepingletonJoepingleton Posts: 746

    Thanks for the info, I am a newb to this stuff and it helps alot.

  • MEC4DMEC4D Posts: 5,249

    I am using my Asus motherboard software to do that  not manually  , I left it to the software as it does it optimal settings for all my profiles, it adjust all the  fans powers in my system for best performance too , I did not lucked out on the silicon lottery I just got already tested one for the speed  when I build my new rig in December , next time check here https://siliconlottery.com/collections/1150 for references

    My offset voltage is : -0.020V

    OC Voltage is : +1.395 V

    = 1.375 CPU Core Voltage

    set to Adaptive so not constant 

    I can see definitely faster performance while working  when set on extreme performance and that is cool to have , but I would never go manually with this level of settings with untested CPU from the store ..  so I got the special one lol  and I paid for a little more

     

     

     

    JCThomas said:
    MEC4D said:

    I am using 4790K , I tried higher speed but it was overheating too quick  with 135W power consumption , with the 240mm radiator I can do that better , when rendering animations it load faster the frame updates to the card when overclocked  and does have the time to slightly cool off between frames . For one render not worthy but for a lot of renders the extra speed save a lot of time , the same with the GPU , if you set the Maxium Power under Nvidia Panel it will render faster the sequences since the clock will be in stand by mode all the time,  again not worthy for 1 render .

    JCThomas said:
    MEC4D said:

    i7 4Ghz overclocked 19% to 5Ghz water cooled, I use Corsair Hydro Series™ H80i V2 Water / Liquid CPU Cooler. with 140 mm extra thick radiator 2 fans  . CPU have nothing to do with rendering when using GPU , all it does only it load the scene to the GPU so faster CPU will only load faster the scene to your card but not improve rendering, so if you have 4 cores or 10 it does not matter for iray software , 4 cores i7 are optimal to run 4 x Titans X with iray .

    CPU Voltage max 1.344 per core max power mode on full load temp. 50-52 C ,  power draw from 16w to max 95 w depends of the mode , Auto , Silent or Performance etc since I use Asus motherboard software to control all cooling system in my rig depends of what I need ..   ,  temp for cards 35-37 C per each GPU on full load .

     CPU and each GPU have own closed water cooling system independent from each other  as normally I just use half of the rig power if not rendering so half of the radiator fans are off . so far that is my best rig ever with the lowest temp I ever had . But going to build external card case this year when the additional 2 x 1080 arrive so all my 4 cards going outside the main rig so I have more space on top for for my CPU new cooling Corsair Hydro Series H100i V2 Extreme Performance Water / Liquid CPU Cooler. 240mm , right now the place is occupied by GPU radiators 

    JCThomas said:

    Off topic, but what i7 CPU did you get to 5 Ghz and what kind of cooling are you using? I'd love to know your voltages, temps at load, and power draw, if you've got a chance and the info is handy. If not, don't sweat the extra stuff...just especially curious about the CPU and cooling. Thanks.

     

    Thanks for the detailed reply. Right, I had been aware that the CPU  doesn't affect the render times in IRay or octane. I just haven't been able to get my CPU stable at 4.7, although I've been unwilling to take my voltages that high. Are you using a 6700k or a 4790k?

     

    Was that 1.344v the result of an auto-profile in your motherboard, or did you manually input it? You lucked out on the silicon lottery, being able to hit 5 ghz, that's awesome. I ask about the auto-profile (like an auto overclock utility or something, like some ASUS boards have) because all of my research has indicated that 1.3v is the max recommended safe voltage for the 4790k. If you'd seen that somewhere before and were willing to risk it, cool, power to ya. If not, you might consider lowering the voltages and decreasing the clockspeed a bit.

    I hope I don't appear to be second-guessing you. I'm just saying that if you overclocked your CPU by typing "5.0" into a multiplier setting in your BIOS and then let your motherboard determine the required voltages, you might have ended up with higher voltages than you wanted / are safe.

     

  • FrankTheTankFrankTheTank Posts: 1,513

    @MEC4D, thanks for all the info. That makes more sense to me now. (Your shaders are awesome by the way, just bought PBS Vol 2 a few day ago)

  • MEC4DMEC4D Posts: 5,249

    Thanks , really appreciate the support! I hope you enjoy the pack !

    @MEC4D, thanks for all the info. That makes more sense to me now. (Your shaders are awesome by the way, just bought PBS Vol 2 a few day ago)

     

  • RuphussRuphuss Posts: 2,631
    edited May 2016

     

    nice to look at, Joe

    Post edited by Ruphuss on
  • MEC4DMEC4D Posts: 5,249

    I missed this one , looks so much better than the last one with better frame speed , the other made me disy , was too fast  lol 

    I have been finding that turning off the CPU render node greatly decreased my render times and I am having much better results using GPU only.
    This is a link to my latest test.

     

  • JoepingletonJoepingleton Posts: 746
    edited May 2016
  • FrankTheTankFrankTheTank Posts: 1,513
    edited May 2016

    That's 10 seconds a frame, thats really fast. How many iterations is that rendering? I find it hard to believe its only 5. Granted its very washed out but I'm curious about these "custom iray shaders". I normally render to 400 iterations and nver get anything useable below 100.

    Post edited by FrankTheTank on
  • MEC4DMEC4D Posts: 5,249

    Interactive mode is fast I render 1 sec HD frame with my metal shaders on a single robot in movement but that is direct light so some of the shader functions are not  used , no volume, refraction or sss or light bauncing from each object it is almost like old Open GL Game but with AO , if you go in photo real mode you will need more than 350 iteration to even begin with and for indoor scene a lot more .

     

    That's 10 seconds a frame, thats really fast. How many iterations is that rendering? I find it hard to believe its only 5. Granted its very washed out but I'm curious about these "custom iray shaders". I normally render to 400 iterations and nver get anything useable below 100.

     

  • FrankTheTankFrankTheTank Posts: 1,513

    OK, I never rendered in interative mode, normally only photoreal or occasionally OpenGL if I'm running a quick test

  • JoepingletonJoepingleton Posts: 746

    I have been experimenting with interactive mode to see how fast I can get a usable render.  In the latest test I basically turned the roughness down to get rid of the "fireflys".  I liked the water in the outside test because The noise made the water seem like it was shining.  But the noise in the shadows in that one were distracting.  So this time I tried to get the shadows to be solid. It may have washed things out to much. ;)

  • MEC4DMEC4D Posts: 5,249
    edited May 2016

    here is my very old animation from 2012 with Direct Light in Octane, I used batch processing to export all my DS scene frames and load it to Octane standalone automatic that was the first DS animated character (M4)  in Octane . The light is the same as  Interactive mode in Iray , you can do cool stuff with if  you don't need the extras  stuff like volume or sss , reflections works perfect but the shaders will be less physical .

    Sometimes people think that Octane is so many times faster than iray because they use Direct Light only ( Interactive mode ) , and it is faster but not that much when both render in photo real mode , Octane just use better the cudas than iray when more than 2 cards are in the system , with 1 card in the system almost the same speed

    Joe, you should add note that is Interactive iray mode or everyone will think it is photo real what will make iray looks not so good in general since it is just a draw mode and not full physical based rendering  

    OK, I never rendered in interative mode, normally only photoreal or occasionally OpenGL if I'm running a quick test

     

    Post edited by MEC4D on
  • JoepingletonJoepingleton Posts: 746

    Thanks for the suggestion Mec3D,

    I moved all my tests to their own thread to avoid any confusion.

  • MEC4DMEC4D Posts: 5,249

    I hope I did not offend you or something ..  it is not me that was confused but you know how it works  

    Thanks for the suggestion Mec3D,

    I moved all my tests to their own thread to avoid any confusion.

     

  • JoepingletonJoepingleton Posts: 746
    edited May 2016

    NP,  I agree that it's better that I move my tests to it's own thread to avoid any confusion and to give me a place to post my experiments without cluttering this thread:)

    Post edited by Joepingleton on
  • MEC4DMEC4D Posts: 5,249

    Yeah you are right it deserve  own place  

    NP,  I agree that it's better that I move my tests to it's own thread to avoid any confusion and to give me a place to post my experiments without cluttering this thread:)

     

  • MythmakerMythmaker Posts: 606

    Thanks guys and gals for the tips on having CPU ticked or not ticked.

    Bumped into this thread looking for interactive mode specific info. 

    @ MEC4D - the point you made about Octane and direct light unfair comparison is spot on! Other significant variables like sun/sky use vs Env map choice, also host-plugin optimized file loading etc affect speed too so a real scientific diagnosis is hard to come by indeed!

    Agree with your 'game style AO indirect' analogy for interactive mode too, now that I've done some actual tests and visited nvidia sites...

    @ Joepingleton - the bit outdoor "accidental" shimmering water is interesting insight! Will check out your anim focused thread soon once I'm done with the iray basics...

Sign In or Register to comment.