The entire lava lamp effect is a shader, not just the bokeh. :)

I haven't noticed any drop in registrations, conversions, or any noticeable differences in traffic patterns after launching the redesign, so I'm not sure if this is actually happening. Though it's a valid concern and issue, and I do want to fix it. And I appreciate all the reports. I think it may be a retina resolution issue, but could be wrong.

On launch day, everyone loved the effect (it goes along with the new logo) and only a couple people said it performed badly, but they were on exotic devices.

I may have caused this a few days ago when I enabled rendering at retina resolution for tier 3 devices. What I wasn't expecting was so many tier 3 devices that aren't really what I'd consider "tier 3." I was expecting all tier 3 devices to be gaming-level GPUs.

I guess I should have read the source [^0] more closely and I could've avoid this.

[^0]: https://github.com/pmndrs/detect-gpu/

> What I wasn't expecting was so many tier 3 devices that aren't really what I'd consider "tier 3." I was expecting all tier 3 devices to be gaming-level GPUs.

Could be the detection of GPUs doesn't work correctly? My nVidia 1080Ti is detected as nVidia 980Ti, and a sibling comment mentions an integrated laptop GPU being detected as tier 3 with isMobile=false.

Another danger here is testing with the website being the only thing on the PC: some people use their computers like that but many others will have many different things open at once: just because it can run at an acceptable FPS on the detected hardware doesn't mean the resource use is acceptable on a system which is multitasking. I usually have >100 tabs in my browser and at least 20 of which are loaded: they cannot all demand what your site demands, even though the hardware is quite high-end.

How many tabs are simultaneously visible? An animation in a non-selected tab won't get much in the way of resources, so the number of loaded tabs shouldn't be a big factor.