Having the game freeze for 0.2 seconds during a team fight and missing a crucial spell is enough to make anyone want to smash their keyboard. As powerful as the 9800X3D is, Dota 2's ancient engine was throwing some threads onto non-cache cores, causing latency to jump between 12-25ms. I tried lowering the resolution to ease the load, which gave me maybe 10 more FPS but made the game look like a blurry mess—a pathetic compromise for this hardware. Instead, I used a process scheduler to force the main game thread onto cores 0-7 and locked the SOC voltage at 1.2V. In AIDA64 latency tests, my memory latency dropped from 72ns to a tight 64-68ns, and the micro-stutters vanished. I did run into two BSODs when I first bound the cores, but loosening the memory timings fixed that. The CPU stays around 60-66℃ now. I exported the stable parameters through my analysis tool, and the fans are humming along steadily at 1400-1600RPM. Last updated onFebruary 28, 2026 6:30 PM.
My clock speeds were bouncing all over the place around 3.8GHz, and the game would hitch every few seconds. I was actually kind of excited to hunt down the cause of this lag. It turned out the mounting pressure on my PA120 SE was uneven, causing core temps to spike to 92-96℃ and hit the thermal wall instantly. I tried undervolting in the BIOS first, which dropped the temp by 4℃ but killed my FPS from 240 down to 190—not a trade-off I was willing to make. I ended up remounting the whole cooler with a high-end 13.5W/mK paste and set the fan curve to hit 100% once the CPU touched 70℃. In AIDA64, the peak temp dropped from 96℃ to a stable 72-78℃, and my clocks finally locked in at 4.5GHz. The fans sounded like a helicopter taking off at first, but I tweaked the start-up speed to 500 RPM to balance the noise. The CPU load is around 80% and the heat is gone. HWiNFO shows the heat soak is solved, with temps staying between 72-78℃. Last updated onMarch 5, 2026 6:33 PM.
My clock speeds were bouncing all over the place around 3.8GHz, and the game would hitch every few seconds. I was actually kind of excited to hunt down the cause of this lag. It turned out the mounting pressure on my PA120 SE was uneven, causing core temps to spike to 92-96℃ and hit the thermal wall instantly. I tried undervolting in the BIOS first, which dropped the temp by 4℃ but killed my FPS from 240 down to 190—not a trade-off I was willing to make. I ended up remounting the whole cooler with a high-end 13.5W/mK paste and set the fan curve to hit 100% once the CPU touched 70℃. In AIDA64, the peak temp dropped from 96℃ to a stable 72-78℃, and my clocks finally locked in at 4.5GHz. The fans sounded like a helicopter taking off at first, but I tweaked the start-up speed to 500 RPM to balance the noise. The CPU load is around 80% and the heat is gone. HWiNFO shows the heat soak is solved, with temps staying between 72-78℃. Last updated onMarch 5, 2026 6:33 PM.
Every time I hit a heavy firefight, my FPS would dive from 140 down to 60 without warning, accompanied by these micro-stutters that are just nerve-wracking. The P-Cores and E-Cores on the Ultra 9 285K were fighting over physics calculations, causing instructions to bounce between cores constantly. I tried the 'Ultimate Performance' power plan in Windows, but the CPU just hit 98℃ and the drops didn't stop—it was a total dead end. I eventually went into the BIOS and manually locked both PL1 and PL2 power limits to 250W and disabled C-States entirely. Checking RTSS, my frame times tightened up from a wild 10-45ms swing to a consistent 8-14ms. I actually messed up the voltage offset at first, which caused random reboots while idling, but after dialing it back to +0.02V, it's been rock solid. The CPU now sits between 75-82℃ with a balanced load. 3DMark CPU tests confirm the scheduling is finally sorted, and the input lag is gone—it just feels snappy. Last updated onFebruary 26, 2026 2:49 PM.
Every time I hit a heavy firefight, my FPS would dive from 140 down to 60 without warning, accompanied by these micro-stutters that are just nerve-wracking. The P-Cores and E-Cores on the Ultra 9 285K were fighting over physics calculations, causing instructions to bounce between cores constantly. I tried the 'Ultimate Performance' power plan in Windows, but the CPU just hit 98℃ and the drops didn't stop—it was a total dead end. I eventually went into the BIOS and manually locked both PL1 and PL2 power limits to 250W and disabled C-States entirely. Checking RTSS, my frame times tightened up from a wild 10-45ms swing to a consistent 8-14ms. I actually messed up the voltage offset at first, which caused random reboots while idling, but after dialing it back to +0.02V, it's been rock solid. The CPU now sits between 75-82℃ with a balanced load. 3DMark CPU tests confirm the scheduling is finally sorted, and the input lag is gone—it just feels snappy. Last updated onFebruary 26, 2026 2:49 PM.
