RTX Titan or 2 RTX 2080 ti?

[Boomba 5]

so let's say my company is building me a new machine for C4D/After Effects.

 

We are going with the 32 Core Threadripper AMD just announced and 64GB RAM, but undecided on the video card. 

 

What are your thoughts on a GPU for C4D?

I am coming from the old Mac Trashcan, so I never had the Nvidia option and everything will be an upgrade.

 

I see the RTX Titan has 24GB of RAM.

I was thinking of getting 2 RTX 2080 Ti (11GB RAM each)

 

I have not used Redshift yet, but now that I will be able to have a GPU renderer, I am excited about the idea, but don't know whether there are any speed advantages to splitting the workload between cards or not?

I hear SLI-linking the 2 cards is not worth it anymore?

 

Any help would be appreciated, I am a total PC newb, this will be my first.

 

Thanks

 

[Rectro 1,185]

Hi

 

I use Redshift and 1 single 20180ti so cant comment about having two cards, I guess you need to weigh up the cost and power consumption as well as horse power. Nvlink is the new way forward, SLI is of no use to you.  Redshift are doing things with Nvilink which will combine two cards memory in which you get most not all of the combined memory.  I would suggest you head over to Redshift forum and ask this question as your bound to get a more  answers as to that combo.  

 

As you can see in this artical the 2080ti is not that far behind the RTX Titan.

https://www.pugetsystems.com/labs/articles/Redshift-2-6-22-NVIDIA-Titan-RTX-Performance-1325/

 

The RTX Titan is over double the cost of the 2080ti, I cant at this point imagine how there could be any advantage in a RTX Titan over 2 2080ti other than maybe heat, and less power consumption, and double the ram, these are things your need to look at.  If you need 24gb now then there is no other option, but if you can wait to see how nvlink pans out, and what are the gotchas which Im sure there will be a few.

Here is a link I think may help answer some Nvlink questions.

https://www.redshift3d.com/forums/viewthread/28320/

 

Dan

 

 

[DasFrodo 240]

Titans are NOT worth it. They are basically the ultra uber highend versions of the respective generation. The added speed does not warrant their excessive prices, at all. According to UserBenchmark they are a measly 5% faster while being twice as expensive.

 

Unless you REALLY need the VRAM (most scenes don't) I would not bother with the Titan Cards, especially since we have OOC-Memory in Octane and Redshift this point.

 

Getting two GPUs yields the additional advantage of using one in full speed for preview rendering and the other handling your OS and C4D while partially being used for rendering. I have a similiar setup in my Workstation and it's great.

[Boomba 5]

thanks for the replies guys!

@DasFrodo is there a setting in c4d to split graphics cards (ie half of one does rendering and OS, the other 1 does full rendering?)

[Boomba 5]

@Rectro do you find the one 2080ti is enough to meet your needs?

I may start with one card and then add if needed

[DasFrodo 240]

12 minutes ago, Boomba said:

thanks for the replies guys!

@DasFrodo is there a setting in c4d to split graphics cards (ie half of one does rendering and OS, the other 1 does full rendering?)

 

Inside C4D, not that I know of. Unless you're using ProRender C4D doesn't use GPU to render anyways, it's all CPU. Octane at least supports a setting that allows GPU's to be set to a lower render priority on a per device level. This will always be the GPU that your screen is connected to, so basically the one that does all the interface and windows rendering as well.

This is what it looks like:

[Boomba 5]

so in your setup the 2080 is doing the interface?

I would think you'd do it the other way, I'd think the 1060 can handle the screen, no? and the 2080 would be a faster renderer?

or do you notice the 2080 is smoother when working?

 

[DasFrodo 240]

Just now, Boomba said:

so in your setup the 2080 is doing the interface?

I would think you'd do it the other way, I'd think the 1060 can handle the screen, no? and the 2080 would be a faster renderer?

or do you notice the 2080 is smoother when working?

 

 

No, thats actually exactly how I set it up. The interface is just a little bit confusing. Ticking the "use priority" just means it uses the "Render priority" setting above which I set to low. Every GPU that doesn't have this tick enabled is used with maximum priority.

[Rectro 1,185]

2 hours ago, Boomba said:

@Rectro do you find the one 2080ti is enough to meet your needs?

I may start with one card and then add if needed

Yes, I dont even get to close to using all the ram on my 2080ti but I do smaller projects.  Id imagine a larger production would take much more memory but I went with the 2080ti as even with my last card which had 6gb I didnt even use all that.  That would be a character portrait using SSS, 8 displacement maps, and lots of control maps such as spec, reflection, sss amount, hair density, over 25000 hairs and rendered at 5k res and didnt use all my last cards memory up.  Id imagine with 2x 2080ti and the Nvlink when its implemented into Redshift would be a sure winner.  Even one card is crazy fast, and that is a time saver for iterations.

 

I set my 2080ti to do the rendering, and the GTX 780 to handle the viewport and other things.

 

Dan