[graysky]

I build the latest git of popcornmix/newclock5 tried a few 10-bit HEVC clips on different RPi hardware: RPi2, RPi3, and RPi3B+.  What I found is that playback was unwatchable on the RPi2.  Decent but not perfect on the RPi3, and good to excellent on the RPi3B+.  I'm thinking there is some minimum clockrate needed to get smooth playback but am wondering what others are experiencing.  Is that trend about right?

[HiassofT]

I build the latest git of popcornmix/newclock5 tried a few 10-bit HEVC clips on different RPi hardware: RPi2, RPi3, and RPi3B+.  What I found is that playback was unwatchable on the RPi2.  Decent but not perfect on the RPi3, and good to excellent on the RPi3B+.  I'm thinking there is some minimum clockrate needed to get smooth playback but am wondering what others are experiencing.  Is that trend about right?

 HEVC needs quite a bit of CPU and GPU power and SDRAM bandwidth, applying a bit of overclock can help - especially on the lower performance RPi2 and RPi3, but also on the 3B+.

I'm using these settings on my 3B+, without these German DVB-T (1080p50 HEVC) stutters

Code:

temp_soft_limit=70
gpu_freq=500
over_voltage=2
sdram_freq=580
over_voltage_sdram=5
sdram_schmoo=0x02000020

so long,

Hias

[smp1]

What I found is that playback was unwatchable on the RPi2.

 RPi2 has 2x less QPU units inside the GPU compared to RPi3/3B+. It will not be as fast as RPi3 even at identical clock rate.

[asavah]

I'd add to the above:
I would (and do) use active cooling while watching HEVC content to avoid CPU throttling.
I have a 5v fan + a transistor to control it via gpio with simple script when the temperature reaches the threshold which I set to 60C.

[popcornmix]

RPi2 has 2x less QPU units inside the GPU compared to RPi3/3B+. It will not be as fast as RPi3 even at identical clock rate. 

Not quite - GPU (including QPU) is identical on Pi1/Pi2/Pi3/Pi3+ (default clock frequency varies across products, but you can usually get the same behaviour with a a manual overclock).
The arm on Pi2 has half the Neon SIMD width compared to Pi3.
There are also architectural improvements and a higher clock frequency which gives about 50% more arm performance which helps somewhat
(not as much as you might imagine as with the QPU and VPU offload, the arm isn't always the bottleneck).