Rendering, Encoding and Compression
Rendering
Loading the cores up with the heavily-threaded Corona benchmark exposes a tie between the manually overclocked and IPM configurations. The 4% increase over the stock configuration isn't incredibly impressive, but we are only dealing with an additional 300 MHz boost to all-core performance. We see a larger improvement of ~7% during the multi-threaded Cinebench test.
Our single-core Cinbench test exposed an odd errata with the IPM-imposed 5.0 GHz overclock. During this test, our Core i9-9900K fell below both its 5.0 GHz all-core overclock to 4.8 / 4.7 GHz, and the issue is repeatable. It's hard to pin the issue on the software. It could just as easily be an issue with the motherboard firmware, but we've followed up with Intel for more information. We're working to identify the issue and will update as necessary.
Encoding and Compression
The IPM configuration trailed the MCE and manually-overclocked configurations during the HandBrake test by 2.45%. Those configurations also outstripped the IPM system during the compression workload. The auto-overclocked system redeemed itself in the decompression and lightly-threaded LAME workload, but the delta fell within the expected standard run-to-run variation for this test.
Y-Cruncher
Y-cruncher, a single- and multi-threaded program that computes pi, is a great test to use for measuring the affect of AVX instructions. It also exposes the main advantage of manual tuning: The potential for higher frequencies during AVX-enabled workloads.
Our manually overclocked configuration stays at the same 5.0 GHz all-core frequency during the multi-threaded y-cruncher test while the IPM and MCE configurations downshift to 4.7 GHz via the AVX offset. That hands the manually overclocked system a 19.6% advantage during the test.
All three overclocked systems score within the expected variance during the single-thread y-cruncher test.
MORE: Best CPUs
MORE: Intel & AMD Processor Hierarchy
MORE: All CPUs Content