Here are the results of the Handbrake test you requested. I added VideoToolbox Encoder (VTE) performance for grins, including a no-audio run.
(For anyone new to this subject, VTE uses Intel Quicksync HW encoder built in to the processor to offload CPU.)
Some notes / takeaways:
- One of the tricky aspects of using VTE is setting a bitrate. In this case, I used the output of the CPU-only run as a guide. Changing output compression doesn't change transcoding speed.
- The Ave. Thread load is based on poking at iStatMenus while the task runs and noting its report.
- The audio workload is negligible for this task; a small fraction of a core
- About 1.5 cores never get utilized. Occasionally util gets close to 100%. Maybe RAM-bound? The task should not be I/O bound. Idk why it's limited.
- VTE is worth about 4-cores for this transcode (single-pass encode). If you count the time saved, it's more like 6 cores. — This assessment assumes the output quality of the codings are equivalent. I've read Intel's encoder is good, but idk.
View attachment 540542
HOLY SMOKES! THANK YOU!
Ok, so this is GREAT insight, mainly because it gives us real world results as to how fast Alder Lake CPUs perform compared to previous gen CPUs.
So to summarize the encode times:
4790k: 1,369 seconds
10900k (CPU): 507 seconds
10900k (VTE): 306 seconds
12700k (Mac OS 12.2.1): 399 seconds
12700k (Win 11): 392 Seconds
12900k: 341-394 seconds (depending on adjustments)
MacBook M1 Pro (10 core 8+2 efficiency): 554 seconds
Raw CPU performance in this scenario from the 10900k to the 12900k is anywhere between
28.6%-48.6% difference. That's significant.
However, when utilizing the iGPU with VTE, the 12900k falls behind anywhere from
11.4%-28.7%
That's actually pretty incredible seeing how the iGPU from 2 generations ago CAN outperform the 12900k depending on the case. I just wonder if going with the 10900k for the sake of the iGPU is reason enough over the 12900k.
***IMPORTANT UPDATE Feb 16***
Apple's Video Toolbox (VTE) in Handbrake can't be used as a comparison. While VTE does speed up encode times BY A LOT, it does not give the same result. Ie; Handbrake cannot use RF values with VTE and instead has to use a fixed constant bitrate. This potentially inefficient; a complex scene may require higher bitrate, but something a scene as simple as a solid blue sky won't need much of a bitrate at all.
Having said that, the 12900k is clearly the better CPU by a wide margin, and it's not even close. 28% on the low end and nearly 50% on the high end. I just put in my order for a Gigabyte z690 Aero G motherboard
***IMPORTANT UPDATE March 10***
Updated the summary encode times to include a 12700k. Got the Aero G working with EFI folder from
@CaseySJ (THANK YOU!!!) All P+E+HT enabled, speeds are stock, voltage set to -0.090 offset using a NH-D15 Chromax Black running on 3600mhz RAM (Kingston Fury Beast with XMP Profile 1) on an 2.5" HDD WD Blue (temporary, only using for testing purposes). Temperatures hover around 67-70c sometimes peaking at 72c.
Glad I made the decision to go with the 12700k, it's not far behind the 12900k but way cheaper (and much lower power draw).
So far in Cinebench R23, my scores are around 22,200 in OS X and around 22,800 in Windows.