2x , 3x , 4x times improvement would mean results that are 100% , 200%, 300% faster than testing. 10x even higher. 30x even higher still. Apple also doesn't particularly say that 30x is relative too.
I think the Apple hype machine spun a little out of control in the Mini context on the HEVC encoding stuff. In other Macs with a T2, the built in camera is directly hooked to the T2 chips. That specific video stream compressed to HEVC extremely fast... that's probably true. Stuff that is on the disk and needs to be loaded into Memory and then back to the T2 ... I don't buy that. Especially at 30x ( some directed DMA transfer from disk by T2 to encoding by T2 and then back to the disk again perhaps. ).
[doublepost=1553200862][/doublepost]
This testing write up makes not mention of the T2 at all. In fact,
"...
EXECUTIVE SUMMARY
- The 6-core i7 Mac mini is significantly faster (about 40%) for video compression than the test 2017 iMac
- Almost all video compression is CPU-based.
...."
CPU based, not "out of the CPU based".
Other extremely salient point...
"... The i7 is multi-threaded, which yields faster results than an i3 or i5 Mac mini at the same clock speed. This Mac mini is also running the latest version of macOS. ...
... This is the 27″ iMac I used for testing. It is 4-core Intel i5. While it has a faster clock speed, the i5 is not multithreaded. This will prove to be significant in the tests. ..."
The i7 can runs more decode and encode processes concurrently than the specific iMac being tested here can. They are both . The 2017 iMac is running a Gen 9 iGPU with the tech to do HEVC. The 2018 Mac Mini has the same gen 9 baseline tech (CPU labeled "Coffee Lake" but GPU didn't have huge changes). ( perhaps clocked a bit faster but same fixed function logic for HEVC seconding at 8-bits).
If the i5 has to switch back and forth between decoding and encoding then while the i7 ran both at the same time you'd expect to see the i5 to hit around the 50% mark in relative performance.
And what was the general results of this testing????? Go back and look at the first quote. About 40% faster. The Symmetric Multi Threading (SMT) [ or as Intel marketing term goes Hyperthreading] isn't performance. There is some overhead when the latency reads don't open up a big enough window for the "other thread" to get work done while the its partner gets work done.
Just like the testing commentary says.... the SMT makes a difference here. If the T2 was 300-600% faster then you'd see at least one of these multitude of decode/encode tasks top the 100% mark.
None of them do. zero , zip, nada.
There is no evidence here for the T2 making a big impact here at all.