Rocket Lake... Which colors do you like the most?
Rocket Lake... Which colors do you like the most?
The video discusses identifying preferred cores in a Rocket Lake CPU. For your i9-11900KF, you can use Intel Extreme Tuning Utility to check core performance data. Reviewing the per-core ratios in the BIOS might show each of your eight cores as 54 multiplier, but this seems unrealistic. If all cores are stable at 53 volts, you can determine which ones perform best by analyzing the tuning results.
I usually perform frequency checks for each core under identical voltage conditions. This gives a full ranking from best to worst core in your chip. It’s the more thorough method, though if you’re new to frequency validation, it might not be the most efficient choice. There are alternatives that examine default per-core ratios in XTU, but they aren’t always precise. Per-core overclocking is generally seen as a pointless effort since it doesn’t significantly improve single-thread performance.
Check the XTU manual for the steps. Look for sections related to the procedure you're referring to, and confirm if any settings are defined there. If not, refer to the BIOS documentation for further guidance.
Configure the desired voltage, choose static frequency mode and turn off all power-saving options in BIOS. Enable per-core limits and restrict all but one core to a very low ratio (either 8 or 34 based on the board's minimum setting). Adjust the multiplier until it no longer triggers Windows detection. Launch Windows and run a tool to boost BCLK while inside (assuming an external clock source). Continue raising it until validation fails. If your board lacks a clock generator, perform the same process more broadly, increasing until booting stops. XTU requires a default BIOS setting; it won't work unless you're at that level.
Last night I observed faint stars on two cores in XTU. It seems plausible, but I doubt it's intentional since it matches the setup from the video. I also found some UI issues in XTU, so I checked each core using the voltages he suggested. None of my chip cores remained stable at 5.5GHz—only a single core reached 5.4GHz, and at least five cores could hit that frequency. It’s conceivable KF has a lower binning than K, possibly because the disabled iGPU might have been flawed during manufacturing. That makes sense if other instances aren’t performing perfectly either. I can accept that. Is there value in digging deeper? I’m impressed by how the board and chip performed naturally, though the gains were minimal on synthetic tests beyond that. I recall those old Socket 7 days when you could see significant improvements.
CPU-Z displays clock information with the last two cores highlighted as preferred. It seems questionable if these are truly optimal for an additional 100 MHz boost, especially since I haven't verified their performance. I usually keep the CPU at a steady 5000 MHz for comfort. Beyond this speed, stable operation demands higher voltage quickly. The 10850K model didn't meet expectations for a 10900K build. Raising voltage and frequency beyond 5000 MHz becomes risky. The real upper limit appears around 5200 MHz, after which excessive voltage and heat emerge. Note from Asus BIOS the last two cores perform better at 200 MHz than the first eight. CPU-Z confirms this trend, matching BIOS data. Core 0 can sustain up to 50.00 ratio, while others max at 50.00. Core 8 reaches 52.00 effectively. Core 9 can go up to 52.00 as well.
CPU-Z highlights 2 and 3 as favorites, matching XTU's list. The same applies to the person in the video—this seems unusual. My tests show core 0 and core 1 perform more consistently, staying stable around 5.4GHz, while cores 2 and 3 drop out, with core 1 holding up at 5.5GHz. Can I change which cores are marked as favorites? I lowered the multiplier to 54 but increased the base clock to 101MHz. Passmark results: https://www.passmark.com/baselines/V11/d...3800639340. Why does my mismatched RAM setup work so well? I haven’t enabled XMP since there’s no profile.
I don't agree with that. The essential data is pre-set into the CPU during manufacturing. Intel doesn't have the capacity to identify the top two cores. Even if they did, two solid cores are sufficient. It makes no surprise you discovered the best ones through testing. Intel simply lacks the resources to thoroughly evaluate every core in each chip they make. A 50 MHz difference on a 5000 MHz processor is just about 1%. It's not worth the effort. Setting BCLK to 100 MHz probably offers better long-term reliability without any real performance gain.
I understand the situation, I was checking if the main clock was causing the issue or if another component was involved. At 101MHz base clock I noticed a problem with the ring ratio, so I adjusted it down to 42. I’d like to see if matching base clocks and peripheral clocks would improve synchronization. NVMe performed poorly at 102MHz, which led me to isolate the peripheral clock. It seems this core feature was likely introduced when Intel planned to shift to asymmetric designs, possibly labeling them as "performance cores" for scheduling needs.