Can someone explain how in the actual f**k youtube uses this much GPU at times? 4080 super
Can someone explain how in the actual f**k youtube uses this much GPU at times? 4080 super
30 Comments
do you have the resolution upscaling active in the Nvidia control panel, under adjust video settings there is RTX Video Enhancement, could be set to 4 and trying to upscale Youtube to as high as it can.
`if browser=not_chrome`
`then fry_gpu=enabled`
Bros watching 13 videos at once
Fr tho that definitely shouldn’t be happening, make sure your drivers are up to date and try starting a video on another browser and see if Firefox is the problem. You could also try a browser with a gpu limiter like opera gx
At that bit rate I’m assuming you’re watching 4K video, or something even more complex. Your GPU is performing hardware decoding on a very complex stream.
Turn off GPU decoding and use your CPU instead if you don’t like GPU usage.
you must be watching video games gameplay on youtube
Did a quick test on edge and firefox, watched a 4k60 video and both spiked to about 55% gpu usage on my 3080. They spent most of the time at around 30% or so though.
Firefox took on chrome policy, and wants all the rams.
Check the performance tab and look at what, exactly, is being reported. Most likely the video decode or video processing hardware is reporting “33% usage” (because why build dedicated hardware for decoding thirteen 4k videos in your gaming GPU) and that number is getting pushed to the processes tab because it’s the highest one, but the vast majority of your card is sitting idle (the “3D” tab which represents the floating point units).
Usage percentages are kinda hard to rely on as a measurement of power.
All we see is 33%. But it could be clocked to run at 500MHz since it can handle the load. So it’s not 33% of its peak but 33% of its current capacity.
Also, GPUs have different pipelines so it may be reporting that the Video Decode is at 33% whilst the 3D (Cores) are low.
Go check the actual hardware in task manager, it should give you a breakdown of the options for your GPU.
Same problem with me, but with brave till yesterday it was taking 300 mb for 1 tab open of memory but then for simply opening YouTube it takes 800 mb, also YouTube shows ads!!! With adblock enabled had to download the extension separately. Any solution? I have updated it to latest version
Do you use Adblock?
You would be surprised at how much usage mundane task produce.
If you get MSI afterburner, set your graphs to on screen display and then use the detach setting you can monitor your GPU usage in real time.
scroll your mouse across the screen without any additional apps open and you will be surprise at how much GPU usage there is.
GPU-z and other tools can show you power consumption in watts, which is a far better way to see how hard it’s working
check it in afterburner or hwinfo, youre probably not running max clock speeds so it shows dumb percentages
Hardware acceleration it probably won’t go down to 0 if you stop watching?
Didn’t they also come out with automatic upscaling for some gpus? I dont use it so I cant recall where the settings are for my NVIDIA but I am sure somebody smarter than I could pinpoint those settings to check.
Youtube upscales your 720p videos to 1440p.
For me it was ambient mode on videos. On dark theme, there is an option to make area around video not so dark. Useless thing, but uses GPU like in OP screenshot.
Most likely automatic AI upscaling is enabled in NVIDIA settings.
Get hwinfo (free) for more detailed stats if you haven’t got it. It includes power use, as well as real time clock speeds. Not only for you but anything your other hardware reports stats on.
Do you have a 1440p monitor or above? Take a look at Nvidia panel and Nvidia super resolution. Rtx cards can upscale yt videos but it’s really demanding
either your machine is old or YouTube has gone off the deep end and become a virus looking for ad blockers
Hardware Acceleration. The browser is able to offload part of the decoding process of the video stream to your GPU to handle, which is more efficient at the task than your CPU. This also improves playback quality as you can hit higher resolutions and playback speeds without rendering lag (as opposed to network buffer).
That said, 33% does seem high, but modern GPUs run at a variable clock speed, so theres a good chance that the GPU is underclocking itself as its not under a high demand
What’s the resolution and fps of your display and the video? Are you in windowed mode or full screen in the video?
Super long explanation of encoding and decoding on a GPU and why it might show up as high usage when just watching a video:
The GPU is just as responsible for showing a video as it is a complex 3D application, yet it has separate hardware to do exactly that (decoding, or recording with an encoder). The decoding hardware within a consumer GPU is typically intended not to do a lot compared to the rendering cores for example- at most, you’d convert a video with a codec which uses the encoder/decoder (one of the remaining uses of multi GPU setups today actually come from being able to transcode multiple videos at once). YouTube taking up 33% of your GPU is indicative that the decoding process is utilizing about a third of the GPUs total encoding/decoding capacity, but you can get a more accurate readout of this in particular by going to the graphs view for the GPU.
Specialized industry GPUs like quadro for example would have more encoding/decoding capability in order to support larger displays like a billboard at a stadium, or an array of 9 4k monitors, or encoding a video stream to 10 different formats to be sent to different servers for various purposes. In contrast, laptops with iGPUs often lack dedicated hardware, resulting in watching a video being far more intensive on the GPU or even CPU as either component would have to perform operations it can do, just not optimally (CPU transcoding falls under the x264 codec, and does actually run well on modern high end multi core CPUs). Try watching a 4k video on YouTube on a raspberry pi 3 or lower, or an old laptop- the poor CPU will struggle from this.
Last, encoding and decoding is the process of (decoding) taking a stream of bytes representing a video (often in a format like AV1 or H264/5), and interpreting it into an array of colors that represents the video (encoding is the opposite process, taking that array of colors and converting it into such a format). It is possible to stream directly all the colors (this is called raw, at least in photography) but it would take up a lot of data to store, whether it’s on a drive or streamed over the internet. The encoding formats effectively compress video to reduce the data size, by doing tricks from jpeg-like lossy compression (simplifying the image based on patterns), or lossless compression with key frame and delta info (rather than storing every frame, store only the pixels that change, in general a great compression method but at the risk of massive corruption from one error, hence key frames help establish a full image every n frames). This compression and extraction operation is often not like a common GPU operation (linear algebra and parallelized tasks in a nutshell) or a typical CPU operation (basic arithmetic, loading and storing single values). Rather, such operations would probably want to maintain fast memory reads and writes with a lot of data in chunks, something vram on a GPU is good at, as well as some parallelized tasks involving compressing the image with some algorithms or basic lookups (far less vs. rendering a complete scene every frame, and more specific algorithms can warrant specialized hardware beyond the CPU).
Task manager has never been totally accurate
Considering how shitty Google is I wouldn’t be surprised if they are making browser slower if you are using an adblocker
Probably have super scaling turned to max. Dial it back a notch
It’s down clocked but reads at the same relative loads from a power standpoint
Gotta buffer those adds and create the lag for what you’re watching.