The Apple TV’s hardware is wildly more powerful than that bundled in any smart TV. The current ATV 4K is running on the last gen flagship Apple SoC with a big passive heatsink attached while smart TVs use hardware comparable to that of a low-to-midrange Android phone from 2012-2014. Even the first gen ATV 4K from 2017 is several times more powerful than current smart TVs.
That difference in power is felt quite a lot in the user experience.
I am not sure if "powerful" matters in this context though. (That is, I expect the chipsets built into TVs to be plenty powerful for their intended purpose.)
Have you tried the average "smart" TV you find at an Airbnb? I have and let me tell you, it does matter.
We were staying at one just a few days ago that had a cheap Samsung TV. The UI latency was so horribly laggy that simply clicking an arrow on the remote to try to navigate to the next menu would take up to 10 seconds to finally register on screen. It was also variable, meaning some button presses only took 1-2 seconds to respond, but some took 10 seconds, and if you pressed more than once you'd end up with a whole bunch of your delayed button presses registering at once and taking you to a menu option you didn't want.
Sad to say, but state of the art in these Android menu systems is horrible latency, most likely because the UI devs are building in new javascript features that run horribly slow on older ARM processors and they just don't give any F's about the actual user experience or testing...
I love your reference to AirBnB. That is precisely when I get to experience what I assume the rest of the world is used to. Firing up a random TV at an AirBnB is simply painful. You're 100% right that CPU power matters. The delay on every menu is painful. The UX is just atrocious compared to my AppleTV. I cringe that people use this for their normal viewing.
In many cases the SoCs used in TVs are so underpowered that they can’t render menu screens without frame drops, or if they can they lose that ability after a software update or two because there’s so little margin.
Just how “underpowered” are we taking here? I’m having a really hard time imagining a chip which cannot render a menu in non-fractional fps. An 8088 can do this…
Like I mentioned in an earlier comment, their power is roughly on par with a 2012-2014 low-to-midrange smartphone, which sits somewhere between 5-15% a powerful as a modern midrange-and-up smartphone.
That would be fine if they were rendering to a 720p screen or had much more simplistic menus like the those found on most A/V receivers, but they’re usually running recent-ish Android or something similar, which has fancy graphics and animations all over the place designed for newer devices which make that hardware choke at the 4K resolution that the majority of TVs now ship with. Exacerbating this are the terrible lowest-bidder smart TV apps which are written terribly.
TV manufacturers will never ship an OS more suitable for the hardware though, because they’re concerned that it will make the TV look less modern than competing TVs. They also won’t ship better hardware because that’d cut $5 per unit off of their margins. As such, it’s best to just write off integrated “smarts” and plug in a streaming box that’s not so anemic.
The Netflix app ran fine at first but had outgrown my 2018 smart TV's IQ by 2022. Freezes for a good moment then crashes the TVOS. Hulu as well. Factory reset was a waste of time and fixed nothing. But TCL made a few bucks more going with the cheaper cpu and accelerated obsolescence.
One more reason to get a dumb display...
But on the other hand TV SoCs need to process the video signal at 4k 120hz 4:4:4 without dropping a single frame, although most if not all those tasks are most likely done by an ASIC embedded into the SoC. Would a modern but very cut down GPU be able to do this and also at a low enough power draw?
There's hardware transcoders on the chip that render your video stream. The menu system is just a terribly outdated Android SoC... the latency is entirely from the speed at which it can render the user interface and has nothing to do with how fast your 4K 120/240hz panel can draw a frame.
That difference in power is felt quite a lot in the user experience.