Moving from 3070 to 3080ti. Now how long should I put my scenes to render?

Redgamer369

I was a 3070 user for years and gotten used to rendering with my 3070 and got the scenes done in 20-25 minutes(Depending on the model count). But now since I got myself a 3080ti how long should I put my scenes to bake? below are my amateur calculations, what I want to hear is your idea and most of all your suggestion about this?

3070 = 5888 cuda cores/DaZ version 4.15, Nvidia version 461.40 and I can get renders done in 20-25 minutes(depending on the model count).

3080ti =10240 cuda cores/same DAZ version 4.15, Nvidia driver 471.68. Since 3080ti has almost double the cuda cores amount and faster GDDR6x memory should I expect double performance as well and get the scenes done in just 12-13 minutes. Or should I put my scenes to render for 15-16 minutes just to be precise.

I already did couple of days worth testing and I really couldn't tell the difference between 12 minute render and a 16 minute one using my new 3080ti, also other than iray preview mod being faster in dazstudio than my old 3070 I didn't notice any significant performance advantage using this same generation faster gpu. Is that normal?

I'll be grateful for any professional opinion.

 

Kitsumo

Here's one thing you could do: Do your normal render with the 3070 for 25 minutes or whatever it takes, then look at the log and see how many iterations the 3070 did. Then set up that same render on the 3080ti and set it for the same number of iterations the 3070 did, and see how long it takes.

Other than that, I'd say just stop the render when you're satisfied with how it looks.

Redgamer369

Okay, thanks.

 

crosswind

Suggest use the test scene in the thread to make a comparions if possible : https://www.daz3d.com/forums/discussion/341041/daz-studio-iray-rendering-hardware-benchmarking/p1

Based on the benchmark scene, comparing to a 3070 card, the only card that could appr. double the rendering performance or speed is 4090. (I remember I gave the caculation in your another thread of 4070 ) . So you may roughly calculate the time based on the number of Cuda cores. If a range of rendering time on 3070 is 20 - 25 mins, it will be appr. 13.8 - 17.2 mins on 3080Ti

Besides, if I were you, I might go for the way of setting Max Samples other than Max Time. For example, if you set 4500 iterations as Max Samples for a scene, it will take 20 mins to give you a rendering of pretty good quality with 3070. Then you use the same settings to render this scene with 3080Ti, you'll get the rendering with same even higher quality while it just appr. takes 14 mins.... so you may be satisfied with such a result.  I suggest you consider this way...

Redgamer369

That is a good idea really even though I never thought about that. I usually set max samples to 10400/ rendering quality to 1533/ rendering quality enabled/ Max time to 1500 seconds, but I'm thinking that your suggested way might save me a lot of time while giving my renders the same quality as before. I'll try your suggestion right away.

It's just that I can definitely notice some performance improvement with this new card but I guess I expected a little more because of that 10k or so cuda cores and GDDR6X memory. But again like you pointed out I guess the only thing that can gives me exactly double the speed would be a 4090 and that's out of my league, sadly.

 

 

crosswind

Don't be sad ~ Suggest not expect that much, then it would be easier to have Happy Rendering most of the time.

Redgamer369

Yeah, I totally agreed. You can't expect much from papa Ngreedia without selling a kidney first. Once again thank you for your max samples suggestion, crosswind.

crosswind

Haha ~~ papa Ngreedia   And you're welcome~