Oh yeh, I'm using the full Windows Horizon Client to keep things simple, no zero or thins
You left out a very important detail. What type of network connection are you testing this on? Are you applying any display protocol group policies or UEM smart policies other than the defaults?
Are you testing in full screen, or windowed mode? I've found 2vCPU's to be adequate for watching videos within the browser window, not maximized, and have had good performance with 720/1080 when the machine has 4vCPU. This was with no other apps running, just a base Win7 or Win10 optimized install. This is on a local, 1GB LAN connection with no display protocol GPO's in place (such as bandwidth limiting, quality settings, etc)
What I'm trying to get at, is what is expected with assumed perfect environmental conditions, fast networks etc, but default optimizations and policies in place.
Two screens enabled 1920 x 1200 on the pool (only one actually in use), with YouTube on Default View (showing comments and so on)
Windows 7 32-bit
No other applications open
Using Chrome - IE seems worse
ESXi hosts with very few other desktops running
10 GB datacenter LAN
Over the internet using UAG and a fast connection
Is this with or without HTML5 MMR feature? I would recommend comparing with and without it. See Configuring HTML5 Multimedia Redirection for more info on how to set it up. It requires the feature to be installed on both sides, a Chrome extension installed, and GPOs configured to enable it and to set the URLs in which it will be enabled.