I got my RTX 4090 under 300W without tanking my 1% lows, and the power limit slider wasn’t enough

I got my RTX 4090 under 300W without tanking my 1% lows, and the power limit slider wasn’t enough

Published Nov 20, 2026, 7:00 AM EST Mainak is an ardent tech lover who grew up with computers. Since childhood, he has had the privilege of tinkering, taking things apart to fuel his curiosity. He is a computer science graduate specializing in programming, particularly memory management, as well as networking. Still, he has always had a soft corner for tech, which led him to develop a deep passion for tech journalism. Mainak loves explaining complex technical terms in simple layman's terms and lightening the conversation to set the mood right. Beyond tech, he is also a self-taught audio engineer and jazz guitarist who has mixed and mastered many audio projects. If you already know the basics of undervolting, I won't bore you with another generic tuning tutorial. What rarely gets talked about is the hilarious paradox of the flagship GPU silicon. At one point, former Nvidia CEO Jensen Huang's infamous ​​​​​​"The more you buy, the more you save" catchphrase becomes true, at least on your power bill. While my 2022 Zotac RTX 4090 Trinity OC carries a 450W power rating on the box, taming it with MSI Afterburner unlocked a much lower power draw and consistent performance once I got past the usual stability shenanigans that come bundled with every undervolt. With hardware this powerful, I no longer chase FPS. Even in 2026, a card like the RTX 4090 has more than enough horsepower to deliver high frame rates, provided the games are properly optimized. Between upscaling and frame generation, hitting big FPS numbers is rarely a problem anymore. At first, I took the easy route and dragged the power limit slider down, forcing the GPU to stay below 270W. However, I wasn't satisfied with the GPU's behavior and overall gameplay performance. Only then did I decide to lock the voltage curve instead. And through some experimentation, I discovered just how efficient the RTX 4090 can be. Why the no-frills power limit slider changes GPU clock behavior and tanks 1% lows Choking board power down to 60% drops clock speeds without delivering true efficiency My real concern is longevity, and I've been skeptical ever since I bought the card. My 12VHPWR connector hasn't melted yet, but watching the power consumption creep toward 400W under heavy synthetic loads has always made me uneasy. For most, the power slider is the go-to shortcut. On both AMD and Nvidia GPUs, you can lower the cap in your vendor's OEM software or Afterburner with a single click. However, as soon as you launch a demanding game at 4K with a 60% power slider, Nvidia's GPU Boost algorithm lowers the default clocks in real time based on available power, severely crippling my GPU's efficiency potential. In my case, the entire test session felt sluggish and, more importantly, the 1% lows were inconsistent and kept dipping between 45–50FPS; GPU clocks remained inconsistent between 2,350MHz and 2,400MHz, and it still pulled more power than my undervolt while running at lower clocks. At 1440p, with the 60% power slider on, my 4090 pulled 270W at just 2,300MHz, and at either resolution, 1440p or 4K, the gameplay had micro-stuttering, leaving me dissatisfied with the results. How my tuned RTX 4090 dropped below 250W while keeping clocks stable Testing real-world silicon efficiency in Assassin's Creed Black Flag Resynced at 4K and 1440p Undervolting takes more patience on the RTX 40 series because its boost behavior scales, so it took me a while to dial in a stable profile across multiple gaming sessions and synthetic tests such as FurMark and OCCT stability tests. Even at stock settings in Assassin's Creed Black Flag Resynced, my 4090 stays well shy of its 450W TDP, peaking at 365W at 1.050V and averaging around 320-350W at 4K. Temperatures were never a problem because Zotac's cooler never let the GPU core temperature cross 72℃, the memory above 80℃, and the hotspot above 90℃, whether the GPU is stock or undervolted. For my in-game settings, I ran Raytracing Mode set to Extended and BVH Quality on High, paired with 75% custom DLSS resolution scaling using DLSS 4.0 (310.6), and Preset K to balance image quality and frame rate. Frame Generation was disabled entirely, as shown in the frame-time graph in my screenshots. At 4K with that 75% custom scaling, my stock setup averaged 66FPS with 1% lows sitting at 58FPS. Once I applied the V/F curve shown in the screenshots above and locked the core to 2,550MHz at 0.93V, my 365W peak at 4K dropped to 250W-253W territory. Average performance dipped by roughly 5% to 63FPS, with 1% lows at 53FPS. Since the core clock remained pinned at 2,550MHz and didn't dip below 2,400MHz as it did with the slider, gameplay felt much smoother, and the 1% lows stopped dropping. Intrigued, I tested the same game settings at 1440p, where it showed an even wider efficiency gap: peak wattage sat at 210–220W and dropped as low as 186W while standing near the ocean, beating the 1440p 60% slider's 270W consumption while running 250MHz faster. Those 1440p power draw figures sit only slightly above an RTX 5060 or RTX 5060 Ti at stock settings, showing how much efficiency is hiding inside the flagship die when you stop feeding it excess voltage. For all of these runs, I stayed on the 610.88 WHQL driver, which has remained stable across my testing and avoids the washed-out HDR contrast bug that popped up in later driver releases. Shouldn't a flagship GPU always be pushed to its maximum limit? Trading a 5% frame difference for over 100W less heat and connector stress makes sense Many enthusiasts argue that when paying for an expensive flagship GPU, holding back its clock ceiling makes little sense, and they aren't wrong. Some go a step further by cross-flashing another vendor's 600W vBIOS to squeeze every last frame out of the die. Still, the raw performance of the stock RTX 4090, especially in well-optimized games, leaves plenty of headroom. In my case, I traded in more than 100W of heat for roughly a 5% frame difference, but I achieved my objective: lower power draw and stable 1% lows that translate to smoother gameplay. A smarter voltage curve can make your GPU last longer Ultimately, these test results are specific to my setup and limited to my GPU only. As with CPUs, your undervolting mileage will vary due to the silicon lottery and other factors. Even so, relying on factory voltage curves or simply dragging the power slider leaves real efficiency gains on the table. Spending a little time figuring out what works best for your GPU is well worth the effort, especially when it trims your power bill and keeps an expensive GPU and its power cable running cooler under load.

Original Source

Read the full article at Xda-developers →

KhanList aggregates and links to publicly available news content. We do not host full articles from third-party sources. Always verify important information with original sources.