I can't help it, but I just prefer working on Mac - so a MBP M1 Max is my main machine, where I work in Photoshop on, After Effects as well as everything around (email, research, etc.). I have a Windows Workstation with three watercooled RTX 4090s, that I only use for C4D. The machine sits in another room and I work on it 100% remote with parsec. Is it possible to work with C4D on my Mac and use the power of the three 4090s in Live Viewer? That would be extremely awesome and my dream setup.
Hey and thank you very much for your comment and question. As far as I know linking Mac and Windows systems is not yet possible but according to Otoy will come in the future with Octane 2023.2 I hope it will as you are not the only one waiting for it. So fingers crossed that it comes rather sooner then later! Cheers, a great Sunday evening and a good start into next week to you!
Brilliant, thank you, this explains why when using the octane solution my laptop with 2060 can handle bigger than usual scenes when combined with 2080 on my PC at that time, this might be an idea for a future tutorial, maybe do a quick tip on batch rendering since this month's theme is rendering on networks it can play well in the context of setting up network rendering, then batch render your scenes after, I like that you specified the emoji's at the end, I wish you a great week ahead, cheers! 😂🖥🖥
Hey Zaidi, thank you very much for your comment ✨🖥️🖥️✨ Thank you for your suggestion. The batch rendering part is done best with C4D Teamrender. In my experience its the most stable solution. But as you will see next week, there are also some disadvantages to it. I am not a huge fan of batch rendering in the render Queue as it has it´s own set of limitations. So what I am trying to say is that there probably is no perfect way of batch rendering Octane scenes unfortunately. And we have to choose the one that fits best for us. Cheers and a great start of the week to you as well 🙌
your'e welcome, i'm looking forward to your next one, i tried the c4d render queue for the first time about 2 days ago haha, so I'm definitely a noob but I figured out that it's stable so far with a 9 render jobs queued overnight however, to render the octane exrs you'd have to have the c4d outputs too, so I let c4d save jpegs while I get my exrs sorted in a different folder, it's a weird hack but worked idk if you faced a similar issue with it, but yeah, I'm super hyped for next week and the weeks after haha, thank you for sharing your knowledge 🌠 @@SilverwingVFX
@@0zaidi You discovered some workarounds that are the same for C4D Render Queue and Teamrender... JPEG for C4D and Octane output elsewhere. It´s exactly that ha ha. So you are probably good and do not need to watch next weeks tut XD 😆
@@SilverwingVFX oh you uninvited me? I'm messing with you 😂, I enjoy your content, there's always some nice nuggets of knowledge here and there even if I'm familiar with the topic, sometimes it's simply affirming 🌟
Thanks Raphael. Awesome tutorial. Really made those old machines laying around useable again. Quick question: since we're effectively sending data over a network, is the slowdown (assuming there's any) for first pixel in the LV noticeable as opposed to rendering on one machine?
Hey there Kenan. Thank you very much for your comment ans your question! The most load of the network comes from initially distributing the scene. But your local machine starts rendering right away and the clients start once they have received the scene. You can even see this in the video in both cases: Rendering in the live viewer at 7:50 where the clients join over time. And then also in the final rendering bit 11:42 also you see one node join first and then shortly after the second one. I have 10Gbit networking from my Workstation to the switch (10Gbit switch as well of course) and then 5Gbit network on my two network nodes. Its more important to have a fast connection between the server and the switch because the server serves every node his own scene. So while every node only has to receive its own scene the server has to send multible of those (one to every node) The faster the connection the faster the nodes can start rendering and therefore also the faster the rendering will finish. As far as I know the rendering is never slower then rendering on the main machine only. Sorry for the long read😇 Hope this clears that up a bit ✨
@@SilverwingVFX Thanks Raphael for taking the time to explain. This is a brilliant answer. Happy to hear that the slaves join in later to alleviate any bottlenecks in sending data. Again, really thanks for the awesome tutorial and thorough explanation. Looking forward to next week’s one
Hey there Loh and thank you very much for your comment! This should be available for every octane plugin application yes. With the Octane Render Node you render frame by frame locally on your own machines. With ehe Octane Render Network you export your scene / animation as a whole and then upload it to the decentralized network. Single frames of your animation will then be distributed and rendered on "node operator" machines that are part of the render network. Hope this helps to understand. If you have more questions, feel free to ask ✨
You might have an older Enterprise Subscription that is limited to 2 GPUs and no network rendering. If you have the Studio + subscription, this should work. What is your setup looking like. You have 2 PCs and they are connected via 1G network?
Thanks for the video as always! Have you played around with the NVLink functionality to extend VRAM capability ? Im not sure if it is available on the 3090 series cards. But here with the 2080Ti's we run it is possible to combine the vram to 44GB, basically adding the vram of all cards.
Hey hey. Thank you for your comment. Appreciate it! I have not tried it my self because of three reasons: 1: It´s tricky to do with more then two GPUs (On the 4090 there is even no option at all) 2: It makes rendering slower. 3: I do not need it with the scenes I usually work with (6gb vram is mostly enough there 😇) So for me it would be extra work, probably less stable (simply because it´s rarely used and therefore the feedback is missing) and also slower because there is always an overhead in shared memory solutions. Not as much as going out of core in to the RAM. But slower then if your GPU had 48GB on it´s own. Hope this helps. A great start into this week to you! Raphael
"obviously if you can read, you already have the information" haha loved this. We often use 1 server with 4 or 5 added nodes, but does this mean our render would be faster if we only added 3?
Ha ha 😇 Thank you for your comment. If it´s faster or not is dependend on some factors. Scene size, frame render time, network speed etc. In general, what I probably should have added is that rendering animations with short frame times (20-50sec) is causing much more bottlenecking then rendering 5min per frame. I think you will only know if you leave render power on the table if you really meassure it. There are two ways. The more easy one is to render a frame again and again with different machines and then see if you get the expected speed bump with 3, 4 and 5 machines added. The other method is to render the a frame on your workstation, note the frame time, then run octane benchmarks on all the machines (including your workstation) and then do some math how fast that frame would be rendered if you add all the benchmarks together. Then find out if the actual speed is close or of it is way off. The Team Rendering approach next week is usually way better for animation rendering as each machine does it´s own thing and therefore renders at its full capacity. Though also has some disadvantages as well.
@@SilverwingVFX Thanks for the comprehensive reply. I never knew I could use the teamrender before this, but I did notice Octane bottlenecking during 50second scenes where it would look like it was up/downloading the frame to other nodes. I'll run some tests here to see how extensive the bottleneck is.
hey Raphael.. on the render node it asks me for the login. so i type it in.. press enter and then it asks for the password. it does not accept ANY password.. the only thing you can press is enter. then it complains that the l/p combination is not authenitic... and it exits.. WTF ?
Hey there Vladan. Thank you very much for your comment. It just looks like you are typing nothing. Its a security meassure so someone standing by can see how many letters your PW is. So just hit enter after entered you entered login email address and then enter the PW (no other keystrokes as all of them are registered) Then hit Enter again. That should work. Cheers, good success with the PW and a good start into the week to you and!
@@vladan.Poison Thank you very much for your feedback. Super glad it worked! I only knew about that because I used some versions of Linux that had the same "feature". Otherwise I would have been as lost as you.
I can't get this to work. I don't know what I'm doing wrong. I've tried everything. Do both computers need to have the same drivers? Do both computers need to have the same versions of Octane? The nodes don't show up. ugh
Hey there and thanks for your comment / question. Both computers have to have the exact same version of octane. So Client and the version the e.g. C4D plugin (Or what ever Plugin / Standalone you are using) is based upon need to match. The computers do not have to have the same drivers. Depending on your system sometimes Firewalls can block the service from reaching out into the network. So you could do a test run turning off firewalls on both machines. Of course it should be a given that both machines are using the same network and find each other. You can easily test that by "pinging" the IP address of the other machine via the Windows command line. "ping 192.168.1.10" for example (without quotation marks) Usually if it's not working, it's a firewall issue. Whenever I have problems with software communicating over the network I reset my windows firewall (I do not use any 3rd party solution) Fingers crossed you will get it to work. Usually it should show right up without problems 🥲
Hey Raphael, I've been looking all over the net, but couldn't find a satisfying answer. Let's just say, I've got a server with 16gb of vram and a node with two 8gb cards, will my server vram be capped to the 8gb like it would if I had all cards in one machine? Or do the node cards not count into the global vram max? Any hints would be greatly appreciated, thank you for doing what you do.
Hey Maximilian, thank you very much for your comment and your question! It is always the same logic here. If you have smaller cards anywhere in the chain, those have to go out of core on bigger scenes. Or, if out of core is not turned on, they are excluded from rendering. It would be exactly the same on one single system with different VRAM sized GPUs. If the scene is larger then the lowest of those, this card will go out of core and the more out of core it goes, the slower it will be. If you have OOC turned off, the card will be disabled from rendering after a couple of tries to load the scene there. It does not matter where the card is. As long as the render process is started the scene can either fit the card, it can go out of core (if enough Ram is available and OOC is turned on) or it can´t be rendered. Note that OOC Ram usage is accumulated. So if you have 4 GPUs and all of them have 2GB OOC, then the Ram usage is 8GB (4 x 2GB) This is because programs as well as Windows can hold different amounts of VRAM "hostage" on a GPU and therefore wary the amount of free VRAM per cards. And therefore the out of core can´t be unified for multiple cards. I hope this answers your question. If you have any more specific ones, please absolutely feel free to ask! Cheers and a great Start into the weekend to you! Raphael
Thank you so much for your detailed answer! That makes perfect sense and was what I feared! Guess I'll have to get way better at optimizing my scenes then or take the time hit from ooc. I find myself frequently running out on vram in even medium detailed shots, so I was definitely hoping that buying another card with twice the vram would fix all my problems.
@@kesslau Thank you very much for your feedback. Yeah Octane needs quite a bit of Vram these Days. Resource management would be one of the biggest critiques I have when it comes to Octane. It has gotten a bit out of hand. I think this is the downside of having all those nice features on the other hand! Optimizing is really scene depended so it really depends if you have a lot of meshes or a lot of textures. Or maybe even both. This might be a good idea for a tutorial in the future 🙌 Oooh, and I just saw you now are supporting me on my Patreon. Thank you so much. This is highly appreciated! Cheers and thank you very much again and an awesome weekend to you! Raphael
Hi Raphael, thanks again for a great video. Maybe I missed it, but Bepe from Otoy recommended a few years back to uncheck "minimize net traffic" (for whatever reason, normally I would say it sounds lik a good thing for netowrk rendeirng :D) and set the max tile samples to 8 for getting improved network rendering times. Do you have any experience on that? It worked on my end when struggeling with bad performances over the network.
Thank you for your comment on my other video as well. I was not mentioning that because mentioning all the extras would make the video to long. But as I did my Team Render one I thought it might have been better if I did. Well there is room for a network rendering gotchas video... ha ha. I usually turn off "minimize network traffic" as well as it will just render all of the samples of one tile on one node until max SPP are reached and then will send back the data. But I had instances where the node would error out and leave a 2h render with a noisy rectangle because the rendered data never came though. I am sure Otoy has improved that since. I just mostly render animations with the Teamrender method nowadays. But when I render with Octane Network, as I said above, I usually turn "minimize network traffic" off still... Old habbits die hard. Cheers and thank you for your help. Appreciate it! Raphael
I own just one PC... this still got exiting. I'm eventually gettin a Mac Studio (unimpressive renderwise) and I'll hook my RTX4090 to compensate. Cornelius mentioned once there's no delay with NetRendering and after this demo I'm impressed. 🖥🖥 Thanks Rau 🙏🏻
Hey Emmanuel, thank you very much for your comment. Right now network rendering between Mac and Windows is not working. But Otoy said it will in the future. So better wait a bit until you get some new hardware. Thank you for the ✨🖥🖥✨ means a lot!
Thanks a lot fot this great tutorial 🙏 Excited to know more about rendering with Octane and team render 🙂 On my side I have a weird behavior on one of my node who have 2 GPUs, one seems inactive while the other is helping rendering the image.
Hey there and thank you very much for your nice comment. About your problem. There are of course multible scenarios that could lead to such an outcome 🤔 I hope you don´t mind me asking some questions: What kind of GPUs do you have attached to your rig? When installing the Deamon, does it list both of the GPUs in the "Available GPUs" section? Video @ 5:37 If it´s listed I would say the most common cause for not participating in the render are Vram limitations. e.g. if one of your GPUs has more Video Ram then the other and the scene only fits the bigger one. Also in that regard, Windows takes up a chunk of Vram on the GPU the display is connected to. I use MSI Afterburners Hardware Monitor to monitor my GPUs so I have a better understanding what´s going on if one of them fails. Hope this was helpful. If the problem presists, feel free to write some more info so we can manage to get it solved!
@@SilverwingVFX Thanks a lot for your answer, I have on my own machine a RTX 4090 and on the distant node a PC with the 2 gpus : RTX 3080 + RTX 4090. It seems the one that is rendering is the 4090, so as you was suggesting it's maybe because of Vram limitations. But when installing the deamon, yes the 2 cards was listed and available. Also when I'm enabling the network rendering on my machine I can see that the distant node has 2 GPU. I changed the Parallel samples to 28 ahd Max tiles samples to 54 in the kernel setting, maybe it's too high for the 3080, I will try to reduce those numbers :) Thanks for you help
@@BRIAUXROMAIN Thank you very much for your answer. I really appreciate the information! The RTX 3080 seems to have quite plenty of Vram with 10GB / 12 GB depending on what model you have. Though of course it really depends on what scene you are trying to render there. I personally have not noticed a difference in upping the max tile samples. Usually I only go for max parallel and set this as high as possible. Though this is only my experience and very well might vary in different scenes. The best way to know if it indeed is a Vram thing is to install a monitoring software. Like I said MSI Afterburner for example. This way you can see if the 3080 indeed is ruining out of it. Cheers and all the best for you and your setup. Hopefully you manage to fix it. Raphael
I can't help it, but I just prefer working on Mac - so a MBP M1 Max is my main machine, where I work in Photoshop on, After Effects as well as everything around (email, research, etc.). I have a Windows Workstation with three watercooled RTX 4090s, that I only use for C4D. The machine sits in another room and I work on it 100% remote with parsec. Is it possible to work with C4D on my Mac and use the power of the three 4090s in Live Viewer? That would be extremely awesome and my dream setup.
Hey and thank you very much for your comment and question.
As far as I know linking Mac and Windows systems is not yet possible but according to Otoy will come in the future with Octane 2023.2
I hope it will as you are not the only one waiting for it. So fingers crossed that it comes rather sooner then later!
Cheers, a great Sunday evening and a good start into next week to you!
@@SilverwingVFX Thank you Raphael! You are the best - absolutely love your content!
Brilliant, thank you, this explains why when using the octane solution my laptop with 2060 can handle bigger than usual scenes when combined with 2080 on my PC at that time, this might be an idea for a future tutorial, maybe do a quick tip on batch rendering since this month's theme is rendering on networks it can play well in the context of setting up network rendering, then batch render your scenes after, I like that you specified the emoji's at the end, I wish you a great week ahead, cheers! 😂🖥🖥
Hey Zaidi, thank you very much for your comment ✨🖥️🖥️✨
Thank you for your suggestion. The batch rendering part is done best with C4D Teamrender. In my experience its the most stable solution. But as you will see next week, there are also some disadvantages to it.
I am not a huge fan of batch rendering in the render Queue as it has it´s own set of limitations. So what I am trying to say is that there probably is no perfect way of batch rendering Octane scenes unfortunately. And we have to choose the one that fits best for us.
Cheers and a great start of the week to you as well 🙌
your'e welcome, i'm looking forward to your next one, i tried the c4d render queue for the first time about 2 days ago haha, so I'm definitely a noob but I figured out that it's stable so far with a 9 render jobs queued overnight
however, to render the octane exrs you'd have to have the c4d outputs too, so I let c4d save jpegs while I get my exrs sorted in a different folder, it's a weird hack but worked
idk if you faced a similar issue with it, but yeah, I'm super hyped for next week and the weeks after haha, thank you for sharing your knowledge 🌠 @@SilverwingVFX
@@0zaidi You discovered some workarounds that are the same for C4D Render Queue and Teamrender... JPEG for C4D and Octane output elsewhere. It´s exactly that ha ha. So you are probably good and do not need to watch next weeks tut XD 😆
@@SilverwingVFX oh you uninvited me? I'm messing with you 😂, I enjoy your content, there's always some nice nuggets of knowledge here and there even if I'm familiar with the topic, sometimes it's simply affirming 🌟
@@0zaidi 😆 Ha ha ha. But great to hear that you can snatch a bit of knowledge here and there!
Thanks Raphael. Awesome tutorial. Really made those old machines laying around useable again. Quick question: since we're effectively sending data over a network, is the slowdown (assuming there's any) for first pixel in the LV noticeable as opposed to rendering on one machine?
Hey there Kenan.
Thank you very much for your comment ans your question!
The most load of the network comes from initially distributing the scene.
But your local machine starts rendering right away and the clients start once they have received the scene. You can even see this in the video in both cases: Rendering in the live viewer at 7:50 where the clients join over time. And then also in the final rendering bit 11:42 also you see one node join first and then shortly after the second one.
I have 10Gbit networking from my Workstation to the switch (10Gbit switch as well of course) and then 5Gbit network on my two network nodes. Its more important to have a fast connection between the server and the switch because the server serves every node his own scene. So while every node only has to receive its own scene the server has to send multible of those (one to every node)
The faster the connection the faster the nodes can start rendering and therefore also the faster the rendering will finish. As far as I know the rendering is never slower then rendering on the main machine only.
Sorry for the long read😇 Hope this clears that up a bit ✨
@@SilverwingVFX Thanks Raphael for taking the time to explain. This is a brilliant answer. Happy to hear that the slaves join in later to alleviate any bottlenecks in sending data. Again, really thanks for the awesome tutorial and thorough explanation. Looking forward to next week’s one
Is this also available for Houdini?
And what’s the difference between this plugin vs exporting the OBRX file and add them into RNDR network?
Hey there Loh and thank you very much for your comment!
This should be available for every octane plugin application yes.
With the Octane Render Node you render frame by frame locally on your own machines.
With ehe Octane Render Network you export your scene / animation as a whole and then upload it to the decentralized network. Single frames of your animation will then be distributed and rendered on "node operator" machines that are part of the render network.
Hope this helps to understand. If you have more questions, feel free to ask ✨
I followed the instructions, but I couldn't find my render node when rendering online
You might have an older Enterprise Subscription that is limited to 2 GPUs and no network rendering.
If you have the Studio + subscription, this should work.
What is your setup looking like. You have 2 PCs and they are connected via 1G network?
Thanks for the video as always!
Have you played around with the NVLink functionality to extend VRAM capability ? Im not sure if it is available on the 3090 series cards. But here with the 2080Ti's we run it is possible to combine the vram to 44GB, basically adding the vram of all cards.
Hey hey. Thank you for your comment. Appreciate it!
I have not tried it my self because of three reasons:
1: It´s tricky to do with more then two GPUs (On the 4090 there is even no option at all)
2: It makes rendering slower.
3: I do not need it with the scenes I usually work with (6gb vram is mostly enough there 😇)
So for me it would be extra work, probably less stable (simply because it´s rarely used and therefore the feedback is missing) and also slower because there is always an overhead in shared memory solutions.
Not as much as going out of core in to the RAM. But slower then if your GPU had 48GB on it´s own.
Hope this helps.
A great start into this week to you!
Raphael
and forgot the magic word.... love u 4090
Ha ha ha that´s a good one 🙌
"obviously if you can read, you already have the information" haha loved this.
We often use 1 server with 4 or 5 added nodes, but does this mean our render would be faster if we only added 3?
Ha ha 😇 Thank you for your comment.
If it´s faster or not is dependend on some factors.
Scene size, frame render time, network speed etc.
In general, what I probably should have added is that rendering animations with short frame times (20-50sec) is causing much more bottlenecking then rendering 5min per frame.
I think you will only know if you leave render power on the table if you really meassure it.
There are two ways. The more easy one is to render a frame again and again with different machines and then see if you get the expected speed bump with 3, 4 and 5 machines added.
The other method is to render the a frame on your workstation, note the frame time, then run octane benchmarks on all the machines (including your workstation) and then do some math how fast that frame would be rendered if you add all the benchmarks together. Then find out if the actual speed is close or of it is way off.
The Team Rendering approach next week is usually way better for animation rendering as each machine does it´s own thing and therefore renders at its full capacity.
Though also has some disadvantages as well.
@@SilverwingVFX Thanks for the comprehensive reply. I never knew I could use the teamrender before this, but I did notice Octane bottlenecking during 50second scenes where it would look like it was up/downloading the frame to other nodes.
I'll run some tests here to see how extensive the bottleneck is.
hey Raphael.. on the render node it asks me for the login. so i type it in.. press enter and then it asks for the password. it does not accept ANY password.. the only thing you can press is enter. then it complains that the l/p combination is not authenitic... and it exits.. WTF ?
Hey there Vladan. Thank you very much for your comment.
It just looks like you are typing nothing. Its a security meassure so someone standing by can see how many letters your PW is.
So just hit enter after entered you entered login email address and then enter the PW (no other keystrokes as all of them are registered) Then hit Enter again.
That should work.
Cheers, good success with the PW and a good start into the week to you and!
@@SilverwingVFX you're right once again. !! i would never cross mind this ultra nuclear secret measure.. totally useless but it works. thanks !
@@vladan.Poison Thank you very much for your feedback. Super glad it worked!
I only knew about that because I used some versions of Linux that had the same "feature".
Otherwise I would have been as lost as you.
For God's sake, it's not working hahaha Went through all the steps...but still the Active Network Daemons, remains empty =(
Hey hey Rubao. Thank you for your comment. Hopefully we can get this sorted together.
Could you explain what you mean with "remains empty" ?
hello.. Is it possible for u to share your 4gpu machine specs .. even as a text post.. i think that would save lots of people .thank u in advance.
Sure. No problem. Here you go:
Motherboard: ASUS WS X299 SAGE 10G (Bios Version 3701)
Processor: i7 10980XE, 18 Cores - 36 Threads, 3.0 GHz
Ram: 128 GB DDR4 3200 MHz CL18
Graphics: 1 X Nvidia RTX 4090, 3 X Nvidia RTX 3090 (Reference Design / Founders Edition)
two of those GPUs are in external "Razer Core" enclosures connected via thunderbolt 3.
PSU: 1500 W, BeQuiet
Cheers.
@@SilverwingVFX OMG !!! i would offer you beer but it is harmfull (hehehhe) therefore i am sending u a crate of KIWIs.. !!!
@@vladan.Poison Ha ha ha. Thank you. I am fine with kiwis 😊
🖥🖥
Yeeees ✨🖥🖥✨
🖥🤘🖥
I like your network symbol 🤘
💻🖥️🖥️
❤🙏
💻📨🖥
Ha ha also a workable solution. ✨🖥🖥✨ pushing the envelope.
I can't get this to work. I don't know what I'm doing wrong. I've tried everything. Do both computers need to have the same drivers? Do both computers need to have the same versions of Octane?
The nodes don't show up. ugh
Hey there and thanks for your comment / question.
Both computers have to have the exact same version of octane. So Client and the version the e.g. C4D plugin (Or what ever Plugin / Standalone you are using) is based upon need to match.
The computers do not have to have the same drivers.
Depending on your system sometimes Firewalls can block the service from reaching out into the network. So you could do a test run turning off firewalls on both machines.
Of course it should be a given that both machines are using the same network and find each other. You can easily test that by "pinging" the IP address of the other machine via the Windows command line. "ping 192.168.1.10" for example (without quotation marks)
Usually if it's not working, it's a firewall issue.
Whenever I have problems with software communicating over the network I reset my windows firewall (I do not use any 3rd party solution)
Fingers crossed you will get it to work. Usually it should show right up without problems 🥲
@@SilverwingVFX Thank you for your prompt response! I will give this a try!
Hey Raphael, I've been looking all over the net, but couldn't find a satisfying answer. Let's just say, I've got a server with 16gb of vram and a node with two 8gb cards, will my server vram be capped to the 8gb like it would if I had all cards in one machine? Or do the node cards not count into the global vram max? Any hints would be greatly appreciated, thank you for doing what you do.
Hey Maximilian,
thank you very much for your comment and your question!
It is always the same logic here. If you have smaller cards anywhere in the chain, those have to go out of core on bigger scenes.
Or, if out of core is not turned on, they are excluded from rendering. It would be exactly the same on one single system with different VRAM sized GPUs.
If the scene is larger then the lowest of those, this card will go out of core and the more out of core it goes, the slower it will be. If you have OOC turned off, the card will be disabled from rendering after a couple of tries to load the scene there.
It does not matter where the card is. As long as the render process is started the scene can either fit the card, it can go out of core (if enough Ram is available and OOC is turned on) or it can´t be rendered.
Note that OOC Ram usage is accumulated. So if you have 4 GPUs and all of them have 2GB OOC, then the Ram usage is 8GB (4 x 2GB)
This is because programs as well as Windows can hold different amounts of VRAM "hostage" on a GPU and therefore wary the amount of free VRAM per cards. And therefore the out of core can´t be unified for multiple cards.
I hope this answers your question.
If you have any more specific ones, please absolutely feel free to ask!
Cheers and a great Start into the weekend to you!
Raphael
Thank you so much for your detailed answer! That makes perfect sense and was what I feared! Guess I'll have to get way better at optimizing my scenes then or take the time hit from ooc. I find myself frequently running out on vram in even medium detailed shots, so I was definitely hoping that buying another card with twice the vram would fix all my problems.
@@kesslau Thank you very much for your feedback. Yeah Octane needs quite a bit of Vram these Days. Resource management would be one of the biggest critiques I have when it comes to Octane. It has gotten a bit out of hand. I think this is the downside of having all those nice features on the other hand!
Optimizing is really scene depended so it really depends if you have a lot of meshes or a lot of textures. Or maybe even both. This might be a good idea for a tutorial in the future 🙌
Oooh, and I just saw you now are supporting me on my Patreon. Thank you so much. This is highly appreciated!
Cheers and thank you very much again and an awesome weekend to you!
Raphael
Hi Raphael, thanks again for a great video. Maybe I missed it, but Bepe from Otoy recommended a few years back to uncheck "minimize net traffic" (for whatever reason, normally I would say it sounds lik a good thing for netowrk rendeirng :D) and set the max tile samples to 8 for getting improved network rendering times. Do you have any experience on that? It worked on my end when struggeling with bad performances over the network.
Thank you for your comment on my other video as well.
I was not mentioning that because mentioning all the extras would make the video to long. But as I did my Team Render one I thought it might have been better if I did.
Well there is room for a network rendering gotchas video... ha ha.
I usually turn off "minimize network traffic" as well as it will just render all of the samples of one tile on one node until max SPP are reached and then will send back the data. But I had instances where the node would error out and leave a 2h render with a noisy rectangle because the rendered data never came though.
I am sure Otoy has improved that since. I just mostly render animations with the Teamrender method nowadays. But when I render with Octane Network, as I said above, I usually turn "minimize network traffic" off still... Old habbits die hard.
Cheers and thank you for your help. Appreciate it!
Raphael
I own just one PC... this still got exiting. I'm eventually gettin a Mac Studio (unimpressive renderwise) and I'll hook my RTX4090 to compensate. Cornelius mentioned once there's no delay with NetRendering and after this demo I'm impressed. 🖥🖥 Thanks Rau 🙏🏻
Hey Emmanuel,
thank you very much for your comment.
Right now network rendering between Mac and Windows is not working. But Otoy said it will in the future.
So better wait a bit until you get some new hardware.
Thank you for the ✨🖥🖥✨ means a lot!
love u 3000
Ha ha ha. Amazing to hear that. I love you 3000 as well ❤
Thanks a lot fot this great tutorial 🙏 Excited to know more about rendering with Octane and team render 🙂
On my side I have a weird behavior on one of my node who have 2 GPUs, one seems inactive while the other is helping rendering the image.
Hey there and thank you very much for your nice comment.
About your problem. There are of course multible scenarios that could lead to such an outcome 🤔
I hope you don´t mind me asking some questions:
What kind of GPUs do you have attached to your rig?
When installing the Deamon, does it list both of the GPUs in the "Available GPUs" section? Video @ 5:37
If it´s listed I would say the most common cause for not participating in the render are Vram limitations.
e.g. if one of your GPUs has more Video Ram then the other and the scene only fits the bigger one.
Also in that regard, Windows takes up a chunk of Vram on the GPU the display is connected to.
I use MSI Afterburners Hardware Monitor to monitor my GPUs so I have a better understanding what´s going on if one of them fails.
Hope this was helpful. If the problem presists, feel free to write some more info so we can manage to get it solved!
@@SilverwingVFX Thanks a lot for your answer, I have on my own machine a RTX 4090 and on the distant node a PC with the 2 gpus : RTX 3080 + RTX 4090.
It seems the one that is rendering is the 4090, so as you was suggesting it's maybe because of Vram limitations.
But when installing the deamon, yes the 2 cards was listed and available. Also when I'm enabling the network rendering on my machine I can see that the distant node has 2 GPU.
I changed the Parallel samples to 28 ahd Max tiles samples to 54 in the kernel setting, maybe it's too high for the 3080, I will try to reduce those numbers :)
Thanks for you help
@@BRIAUXROMAIN Thank you very much for your answer. I really appreciate the information! The RTX 3080 seems to have quite plenty of Vram with 10GB / 12 GB depending on what model you have. Though of course it really depends on what scene you are trying to render there.
I personally have not noticed a difference in upping the max tile samples. Usually I only go for max parallel and set this as high as possible. Though this is only my experience and very well might vary in different scenes.
The best way to know if it indeed is a Vram thing is to install a monitoring software. Like I said MSI Afterburner for example. This way you can see if the 3080 indeed is ruining out of it.
Cheers and all the best for you and your setup. Hopefully you manage to fix it.
Raphael
yeaaaaah finally thanx❤
You are very welcome 🙌 I think next weeks video will be a great addition to this one!
Happy Diwali 🪔
Ohhh, thank you for broadening my horizon. I did not know about Diwali! Also ✨🥇✨
🖥🖥
Ohhh thanks so much Yassin! ✨🖥️🖥️✨