High performance build with 10850k and 4.8ghz all cores, but expect increased heat.
High performance build with 10850k and 4.8ghz all cores, but expect increased heat.
I wouldn't label them as rejects, but they don't act the same as my earlier CPUs. It makes sense that the i9 10850k maintains a steady 4.8ghz consistently, unlike my i9 10900k which frequently hits 4.9ghz. When I run the 10900k without restrictions, Cinebench R20 reaches around 70c. With AI overclocking it still stays at 4.9ghz but keeps temperatures in the low 80s. An all-core 5.3ghz boost still matches those temps, though it throttles when paired with TimesSpy and my 360mm AIO. If I stick to Cinebench 5.3, it should work fine, but it struggles with the thermal limits of my setup. A manual 5.2 or 5.3ghz overclock seems the only viable path since stock performance is solid at 5 or 5.1ghz across all cores in most titles. I suspect the issue lies with the rendering core hitting 5.2 and 5.3ghz, which other configurations can surpass. Before the i9 10900k era, my i7 8086k and i9 9900k handled 5ghz all-core boosts effortlessly. Now, with the stock setup, I’d opt for a manual overclock that outperforms it and move on.
The 10850K model emerged as Intel aimed to repurpose CPUs that didn’t meet the 10900K standards. They sought new opportunities for their lower-tier cores, opting to sell them at a reduced price rather than discarding them. They adjusted the rated speed to 100 MHz and increased the voltage, making these CPUs more stable. The 10850K is clearly a 10900K candidate that didn’t make the cut. Why does it throttle in Time Spy? Could it be overheating, reaching up to 100°C? HWiNFO64 provides detailed insights into why an Intel CPU throttles. Under proper configuration, a 10900K can sustain up to 5.3 GHz across all cores indefinitely, provided it stays cool. Using Asus AI for overclocking often leads to excessive voltage, which generates more heat. Manually setting the voltage is usually the most effective method. Most users choose to adjust their settings in the BIOS. If you can’t locate or enable SVID, applying a negative voltage offset can help lower CPU voltage. Here’s an example of a -75 mV offset on an Asus board. Monitoring the idle state of a CPU—especially the C7 core—helps gauge how idle it truly is. Many systems run background processes that produce heat without users noticing. I find it useful to track the C state before and after any changes.
CPU-Z displays 1.38V during Cinebench R20 execution. Some Gigabyte boards provide more than necessary voltage, which the CPU doesn’t require. I’m not comfortable with Gigabyte motherboards or their BIOS settings. I checked the manual for a Z490 Aorus Elite, which mentioned 13 voltage levels and modes that can be changed. That information isn’t very useful. No additional details, just a suggestion to search forums or Google for others’ experiences.
It seems my CPU fans weren’t reaching full speed initially, so I adjusted them. Still, it remained too warm in Cinebench R23 (93°C). After adding a -0.100 offset to the throttle stop and running Cinebench on multicore, all other BIOS settings stayed at default (Auto), including MCE. The temperature dropped to 84°C after ten minutes. I’ll do more tests to confirm stability across all uses. Thanks for your support—it means a lot. The CPU Vcore reached 1.344 during Cinebench, and the voltage stays at a maximum of 2.052. Is that within normal ranges? I’m not familiar with VIN voltages, but it looks okay.
The process of using ThrottleStop to manage your power consumption appears straightforward. In the Gigabyte BIOS, enabling DVID seems necessary. It seems you can easily lower your temperatures by at least 10°C. You’re likely running Cinebench at a high frequency and want to see if ThrottleStop maintains performance or reduces it. Check the Limit Reasons section for any warnings about overheating. At 5000 MHz, your voltage regulator is delivering a stable output, so you might still manage a reliable undervolt adjustment of 40-50 mV. Try reducing it further by another 25 mV and observe the results. HWMonitor may not display the correct offset voltage; however, the FIVR monitoring table provides accurate readings. After adjusting, perform another 16 Thread 960M TS benchmark test. This partial load test offers a more realistic scenario, reflecting how Windows distributes tasks across cores. It helps identify potential issues early, especially if voltage adjustments lead to errors. If you notice an error, halt the test immediately—insufficient voltage means you should stop and increase it. Continue testing until you confirm adequate power supply. Regarding your motherboard, I couldn’t locate a VIN reading. With older CPUs, the default input voltage was around 1.80V, though some users increased it to 1.90V. Your reading of 2.05V seems elevated. The main input voltage is crucial, as the core draws from it and then adjusts to the required level. Typically, the VIN should be about 0.4V higher than the VCore value.
@unclewebb cinebench operates at a steady 4.8 GHz. The single-core performance matches expectations at 5.2. I checked similar models and found it slightly above both single and multi-core benchmarks, suggesting it's functioning properly. I'm uncertain about adjusting the voltage further since I'm mostly using it for gaming and studies, though I might revisit this later. I'm not sure if I should change anything else in the BIOS regarding throttle stop, it just started working unexpectedly. I ran a timespy test and got results similar to others with comparable specs. The limit reasons displayed were consistent with cinebench, showing it as a black box. Could this be its intended behavior? Below are the VIN voltage readings I mentioned—what do you know about your 10850k's readings for those values?
On my Asus board, HWMonitor fails to display VIN voltages that your setup shows. When you mentioned your Cinebench voltage, I assumed you were running at 5000 MHz. Here’s the voltage requirement for my 10850K to handle Cinebench at 4800 MHz. That’s why I noted some Gigabyte boards offer more flexibility with voltage settings. You still have plenty of room to lower the voltage further if you plan to push performance even more. You could potentially reduce temperatures by another 5°C or 10°C. When using ThrottleStop, I set the voltage to adaptive default and then apply an offset afterward. This lets the CPU follow Intel’s built-in voltage curve. A black box in Limit Reasons is ideal—it shows no throttling since you started or after clearing flags. A red box means the CPU is being throttled. PL2 indicates the CPU is constrained by its power limit, while a yellow box records prior throttling and a red one signals ongoing throttling. The boxes under the CORE column are most relevant. You can temporarily lower power limits in the TPL window and run a TS Bench test for better insight. If you wish to lower voltage for improved speed, open the FIVR window, click Turbo Groups, and press 50 to enable a 50X multiplier across all cores. The Defaults button reverts to standard turbo settings. A setting like 52 would apply a 50X boost when up to four cores are active, dropping to 50 when between four and ten cores are running.
You can push it to the limit and it works, that's just how it is. Back then, Intel enthusiasts often reached maximum overclock settings right out of the box. For instance, my 3770K needed around 1.525v to hit 4900 lol. The heat was extreme, but it only affected four cores.