Jump to content

Alekseys Blog

Sign in to follow this  
  • entries
  • comments
  • views

GPU for GPU rendering Octane Thea Indigo etc..

Sign in to follow this  


Buying a GPU for GPU rendering? Read this first. I find myself answering the same questions over and over. So here is a summary of what you need to know when buying a GPU for GPU rendering. This information is valid as of writing this, I will try keep it updated as we go along.

Temporary disclaimer: The Nvidia 1080/1070 cards are currently not supported by any major GPU render engine. This will hopefully be fixed soon. (Blame nvidia)


Nvidia. All the gpu renderers are nvidia CUDA based. This might change in the future, but right now if you want a fast stable GPU renderer you are stuck with nvidia.


The onboard memory of GPU cards does not stack. ie, 2x 6gb cards, can handle a 6gb scene. 1x 3gb card and 1x 6gb card, can only handle a 3gb scene. So if you have a bigger scene, you will have to disable the smaller card. And for the most part everything needs to fit on the GPU. Octane has out of core textures, which means it keeps the textures in the system ram, which helps at  times, but apparently also slows down the render. Thea has bucket rendering mode, so when you are rendering the final image the buffer doesn’t take up all the memory on the card.

Also be aware of Dual GPU video cards. Some Nvidia cards including the titans. Have 2 gpu’s on one card. Here, the memory has to be halved. It will say it has 6gb of memory on it, but its actually only 3gb per GPU.  the GTX 790 is an example of a card like this.

PCI lanes

Some cpu have less pci lanes then others. This is mostly irrelevant to GPU renderers, as long as you have at least 1 or 2 lanes per card you will be fine. The pci lanes is more important for games.


once again SLI is irrelevant to GPU rendering. You should not install the sli bridge or enable this in the options.  Furthermore unlike in SLI ( which is only useful for games). GPU render engines don’t care if you have different cards working together. so you can have, 780’s and 980’s and titans all working together. Just keep in mind the memory issue i brought up earlier.


If you are planning on putting multiple GPU into your computer, make sure they have fans that push the hot air out of the case. So either reference coolers, coolers that look like reference coolers or liquid cooled/hybrid cards.

Screenshot 2016-07-13 12_56_49-Google Image Result for http___rog.asus.com_wp-content_uploads_2013_07_matrix-co

Because as you can see here, the traditional fans just expell  the hot air into any direction as long as its away from the card. Works fine if you have 1 gpu in your case, but if you have 4 they are just blowing it at each other.

So what you do want, is the cards with heatsinks like this to blow the hot air outside of the case.



Secondary Brands

There are lots of brands that manufacture Nvidia based cards. EVGA, MSI, ASUS and Gigabyte are the largest in the west. EVGA generally has the best reputation for their great warranties and customer support. Asus probably has the worst reputation for Customer support, but they are much loved for their prices and performance. ( I have asus, even though i got burnt with their laptop). Just you know, read the warranty terms, maybe do some googling. This stuff changes very quickly. Some brands become terrible, some get better.

Dedicated Viewport Card

Many people like to have a dedicated video card which isn’t engaged in rendering to handle the viewport and windows desktop. So they will have 3 titans for example to handle rendering and then a gtx 950 to handle the monitors. This stops your system from becoming unresponsive while rendering. While this works in most situations, you might not want to get the cheapest possible card, because cinema4D uses the card that the monitor is connected to to handle the viewport. And possible other programs do the same. C4D’s viewport isnt too stressful on the video card, most of the slow downs you will see will be cause by your CPU, but still something to keep in mind.


So there you go, now hopefully you will be able to make a more informed decision </div>

			<div class='ipsItemControls'>

	<div data-controller='core.front.core.reaction' class='ipsItemControls_right ipsClearfix '>	
		<div class='ipsReact ipsPos_right'>
				<div class='ipsReact_reactCountOnly ipsAreaBackground_light ipsType_blendLinks ipsHide' data-role='reactCount'>
					<span data-role='reactCountText'>0</span>


		<div class='ipsResponsive_showPhone ipsResponsive_block ipsSpacer_top'>

<div data-followApp='blog' data-followArea='entry' data-followID='42' data-controller='core.front.core.followButton'>
		<span class='ipsType_light ipsType_blendLinks ipsResponsive_hidePhone ipsResponsive_inline'><i class='fa fa-info-circle'></i> <a href='https://www.c4dcafe.com/ipb/login/' title='Go to the sign in page'>Sign in to follow this</a>  </span>

	<div class= Followers 0


Recommended Comments

Not sure about rendering but for simulations such as TurbulenceFD the MEM Bandwidth is the crucial factor..higher the better

Share this comment

Link to comment

Lets say I put 2x Titans. 1x card connected to the monitor and dedicated to the viewport, 1x card dedicated to rendering. 
Will the GPU renderer take also part of the power of the Titan dedicated to viewport, or it will only take part of the other Titan power ?

To make my question simple : Will the GPU use both card power for rendering even if one is connected and dedicated to viewport ?

Thanks ~

Share this comment

Link to comment
Add a comment...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Latest Topics

Latest Comments

  • Create New...