Video Card Discussion: 980 Ti vs. 1080 Ti

Hi there,
My computer is pretty decent and it seemed like the only thing that I could really upgrade that would make a lot of difference would be the video card. The 980 Ti is certainly no slouch but it had 6 gigs of memory and the 1080 Ti has 11.
Prior to receiving the 1080 Ti, I ran a render. Pretty simple one. Just a bust shot with Render Studio lights. With the 980 Ti, the render took 34 minutes.
With the new video card having nearly twice the VRAM, I expected to see a significant reduction in render time. But I didn't. Without Optix Acceleration, it finished up at 25 minutes.
With Optix Acceleration, it wrapped up at 22 minutes.
Any thoughts on what I can do to increase the speed? I honestly expected to shed more than 10 or 12 minutes by putting in this video card.
Thanks!
Comments
Quick Update: I turned off 'CPU' and allowed the video card to shoulder 100% of the load. That reduced the render time down to 19 minutes.
Dropping more than 10 mins on a render just by upgrading a card is a pretty good improvement. Mo RAM does not mean mo speed, it will just let you render bigger scenes (generally speaking). The speed increases come from an increase in the number and/or efficiency in CUDA cores (again, generally speaking),
Probably turning off the cpu would have been boosted the 980 as well. I wouldn't expect a big increase in performances since the two cards score almost the same on passmark (it's 13K vs 11K). As Jim says what makes the difference is more vram so you can handle bigger scenes.
EDIT: and I would strongly recommend to use the scene optimizer and a good denoiser filter in post anyway. Since using only brute force gpu stills a very slow workflow.
Thank you for the comments! I appreciate it very much.
Fantastic advice.
Passmark isn't a good predictor of Iray scores. The 1080ti has >20% more CUDA cores yet render times were reduced significantly more than 20%. The impact of CUDA core numbers shouldn't be discounted for Iray