Compare iGPU, dGPU, and which offers better battery performance.
Compare iGPU, dGPU, and which offers better battery performance.
Hey everyone, I've been trying to make my gaming laptop use less energy when it's not connected to a charger. With the 9th-gen i7-9750h and a 57Wh battery, the power consumption is really high. I've reduced the CPU voltage and switched to using the integrated graphics instead of the dedicated GPU. But I'm stuck because even when idle on battery, it's drawing about 12W, which is way too much. I'm using ThrottleStop, and while my cores were in C7 at 95% or higher, the integrated GPU stayed in C3 and didn't reach C7 like the dedicated one. I've looked online but haven't found any solution—no processes are showing up in Task Manager for excessive CPU use. It seems the issue is with the integrated graphics. I'm surprised it uses only 10W when idle, especially since I set it to "optimal power" in NVIDIA controls. Compared to the dedicated GPU, which uses way less power, I wonder if switching back would help. Can anyone have solved this problem or suggest a more efficient setup? Also, is there a way to make the iGPU more energy-friendly without sacrificing performance too much?
It's best not to rely solely on software for these measurements. The most effective approach is to test with both the iGPU and dGPU in a controlled workload, observing battery life. Simply lowering voltages won't cut power consumption—it only limits performance at the same speed. Combining undervolting with a power cap helps manage draw without sacrificing efficiency. Adjusting both settings appropriately can maintain the same output while using less energy.
I very very much doubt that. NVIDIA GPUs are well known to have quite high idle power consumption, which is the entire point of using an iGPU in a gaming laptop. I wouldn't be trusting any of these figures as being accurate. A lot of the power consumption in a PC is everything other than the CPU and iGPU. Its very hard to measure what is consuming what. The closest you can do is run the laptop off wall power with a power monitor plugged in between, to see what its REALLY drawing from the wall. Even then you might run into accuracy issues as those power monitors aren't usually accurate down to single watts. But ultimately shaving a few watts here and there wont necessarily make an obvious difference, as its the moment by moment workload that sucks the power.
I previously relied on the laptop's reported battery drain for similar measurements. The GPU readings seemed less reliable because they might differ from the actual power usage. With my AMD 5800H and 3070, I observed 18W when idle versus 20W while streaming, and 20W during idle compared to 23W while watching YouTube. Since the power savings didn’t seem significant enough, I turned off the APU to free up RAM and avoid any issues from the graphics card running.
If battery life matters, then choosing a Zen4 laptop makes sense. The GPUs are on par with the RX570/580 or GTX 1060, which is a solid investment. It might cost a bit more upfront, but the older gaming PC still offers good value. Plus, you can wait for prices to fall now that the Zen5 line is available.
So about purchasing a new laptop: The latest chips are significantly more energy-efficient, but right now my Legion Y540 still performs adequately for gaming and I’m out of school in a semester. From a budget perspective, it doesn’t seem worth the investment at the moment. Regarding the power usage reported by the software, I generally believe them. Before switching to an iGPU, I was running on battery with the discrete GPU set to Optimal Power, Windows’ best energy-saving mode, and no undervolting yet. I could comfortably take notes during classes and last about three hours without draining my battery. But I considered trying the iGPU to see if it would improve battery life—though Windows already reduced my usage by half. I don’t fully grasp why this works, but I’m open to understanding. Based on what you’ve explained (thanks for clarifying), I realize that saying the discrete GPU is more efficient than the integrated one seems unrealistic, which is why I’m relying on the data you provided. I’d like to know more about the C-states, because that’s what’s causing me concern. You mentioned ThrottleStop: when using the iGPU, the CPU-core C-states stayed above 95% in C7, and the package C-states remained above 40% in C3, never reaching C7 or even C6. On the other hand, with the discrete GPU, the CPU-core stayed over 95% in C7, with package C-states consistently above 60% in C7. It’s frustrating, but I’m not an expert. I’d appreciate any insight into why the iGPU behaves this way or whether there’s a way to prevent it from hitting those high C-state levels. Also, if there’s any chance the BIOS might have been corrupted (as you mentioned), that could explain the issue and suggest a fresh Windows installation might be necessary.
So the issue was resolved. ASPM wasn't activated on my NVMe storage (running Windows), which meant it wasn't reaching the lowest power state and using energy unnecessarily. I turned it on in the BIOS via advanced settings, unlocked on a Y540-15IRH with BIOS version BHCN32WW (required a flash), and used a software tool I discovered online.
Despite this, my GPU (1660Ti Mobile) with Optimal Power in battery mode performs much better than the integrated graphics. It can be turned off the iGPU, leaving only the dGPU to handle tasks—visually cleaner and more efficient.
It lasts over four hours on battery compared to just two hours with the iGPU (even when dGPU is off). I don’t need your explanation, but my i7-9750H isn’t operating at its best. If it helps, follow my approach—it works well with my Legion Y540-15IRH and charges only about 75-80% of its original capacity.
Good luck to everyone! Thanks for the previous replies!