Handbrake Comparison - Alder Lake i5 vs M2 Pro

[1291030]

If the point is to compare how fast you can get something done (real world performance), then yeah I think it should be enabled.

just make sure you enable hardware encoding in what you compare it against as well though 😂

 

[sunnee]

just make sure you enable hardware encoding in what you compare it against as well though 😂

Hardware encoding can be used only for "save/export" file, not for production. The ratio of size to quality when using hardware encoding is very poor! Files with high quality have a very large size.

Software encoding (like x264/x265) is better for maintaining a small size and good quality. All streaming services (like Netflix, Apple TV) use software encoders or specialized professional hardware encoders (not built into CPU/GPU). Good software encoding requires significant CPU power, and the encoding time demonstrates the real power of the chip!

 

[1291030]

Hardware encoding can be used only for "save/export" file, not for production. The ratio of size to quality when using hardware encoding is very poor! Files with high quality have a very large size.

Software encoding (like x264/x265) is better for maintaining a small size and good quality. All streaming services (like Netflix, Apple TV) use software encoders or specialized professional hardware encoders (not built into CPU/GPU). Good software encoding requires significant CPU power, and the encoding time demonstrates the real power of the chip!

Ok. I wasn't commenting on whether or not he should, or shouldn't use hardware encoding though.

OP was comparing apple silicon versus intel silicon, and "discovered" hardware encoding. My comment was simply, to remember when comparing the two architectures not to mix and match.

🙂

 

[Populus]

I wonder if the M3 will improve substantially the performance of Handbrake, if the number of CPU cores remains the same…

 

[3166792]

@MuckSavage70

Thanks for doing these tests. Will be interesting to see how much time is saved on the M2 Pro using the hardware acceleration codec

 

[Populus]

@MuckSavage70

Thanks for doing these tests. Will be interesting to see how much time is saved on the M2 Pro using the hardware acceleration codec

Like others said in this thread, the hardware accelerated encoding gives usually poor results of file size and quality.

I hope Handbrake is even more optimized for Apple Silicon in the future.

 

[HobeSoundDarryl]

OP's original post reads like the clash of computing power vs. power per watt. The former generally cranks the computations faster. The latter uses less power to get the computing done a bit more slowly but more (power) efficiently.

I'm not really surprised by this outcome and I don't really think it's about unoptimized software. One platform is focused on computing brute force power and the other is focused on computing power-sipping efficiency.

Is there room for software optimization? Of course, there's always room for improvements. But Power and PPW are simply not aiming for the same target. If someone wants to compute as fast as possible, that's probably a powerful PC. If someone wants to compute as efficiently as possible, that's Silicon.

While Apple is easing up processing power and Intel is working on improving efficiency, I don't think there is any impending "meeting in the middle" somewhere. HB encodes needs a LOT of CPU processing power. I would guess PCs would generally win such computationally-intensive contests. On the other hand, if someone is not in a "fastest possible conclusion" hurry, Silicon Macs can do the same while using less power for the task.

 

[galad]

Actually there is a lot of room for optimisations. There are still many missing Neon SIMD optimisations in open source softwares. For example FFmpeg HEVC decoder is quite slow on Arm, and in the next version it will be at least 50% faster. SVT-AV1 is still not optimized at all for Arm.
Amazon is contributing some Neon and SVE code lately because they started settling Arm instances on AWS.