GPU operation without a cable?
GPU operation without a cable?
I went to a friend's place to play some games and noticed his HDMI cable was connected to the motherboard. I told him it should be on the GPU, but he said everything was fine. I checked and it actually worked well. However, when I took the cable off the motherboard and plugged it into the GPU, performance dropped significantly—by about 5-10%. For testing, I tried resetting the GPU settings, but the game ran poorly afterward. This is strange since the GPU was still active and the temperature increased. I also saw the monitor showing the signal coming from the iGPU alone. My questions are: 1. How does a GPU affect performance when it isn't connected to anything? (I could see the GPU usage in Afterburner and noticed the temperature rising) 2. Why did performance drop when I connected it to the GPU instead of the motherboard? Specs: Ryzen 7 7700, Asus TUF B650-M, Wi-Fi Zotac 5070.
Fascinating. This behavior is usually not expected on a desktop setup, where the discrete GPU is typically managed by the integrated graphics unit. Did you restart the system and repeat the test? Otherwise, it might be due to another process running in the background at that moment. It doesn’t seem to enhance performance unless part of the GPU’s tasks was transferred to the iGPU. Lossless Scaling could also account for this output being visible through the integrated graphics.
With the NVIDIA control panel you can manage which app uses the GPU or the built-in graphics by choosing one setting. I think there’s also a setting in Windows Settings that lets you pick which GPU to use; it should default to "auto" and adjust based on what the app needs. Your integrated GPU handles basic screen tasks, while the real GPU manages more complex work. Performance drops are unlikely unless you’re using a very large display.
I evaluated performance across various platforms. Nvidia Optimus works on desktops, and it seems similar functionality exists for AMD systems.
It’s obvious there’s an important detail being missed. You won’t boost performance by sending traffic through the system and out the iGPU. I’ve never found Optimus set as a choice in NVCP or NVApp on desktops with iGPUs. Are you talking about Windows’ option to select a render GPU in Settings > Display > Graphics?
Didn't realize this was something you could do on a desktop. It seems it needs the discrete GPU's output to go through the iGPU, although it might be possible via the PCIe bus.
It seems like the device was likely processing on the GPU, yet its output ended up going through the integrated graphics unit. Since the motherboard’s monitor connections are only for the iGPU, any performance shift after changing ports might stem from something like upscaling being handled by the iGPU. I imagine using tools such as LS to run the game on one GPU while the other handles scaling could help clarify things. However, a more probable issue is an inaccurate measurement or background process influencing results. To confirm, repeat the test several times and consider restarting after changing connections. If the average across three trials stays within 5-10%, it’s likely a genuine problem; otherwise, it probably just happened by chance.