• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Microsoft's New Windows Update Allows GPU Selection According to Workload

Raevenlord

News Editor
Joined
Aug 12, 2016
Messages
3,755 (1.18/day)
Location
Portugal
System Name The Ryzening
Processor AMD Ryzen 9 5900X
Motherboard MSI X570 MAG TOMAHAWK
Cooling Lian Li Galahad 360mm AIO
Memory 32 GB G.Skill Trident Z F4-3733 (4x 8 GB)
Video Card(s) Gigabyte RTX 3070 Ti
Storage Boot: Transcend MTE220S 2TB, Kintson A2000 1TB, Seagate Firewolf Pro 14 TB
Display(s) Acer Nitro VG270UP (1440p 144 Hz IPS)
Case Lian Li O11DX Dynamic White
Audio Device(s) iFi Audio Zen DAC
Power Supply Seasonic Focus+ 750 W
Mouse Cooler Master Masterkeys Lite L
Keyboard Cooler Master Masterkeys Lite L
Software Windows 10 x64
Microsoft's future update to Windows 10 will add a GPU-aware selector that allows both the OS and the user to adaptively select the best GPU for each usage scenario. The preview release of Windows 10 build 20190 features this in two ways. First is an OS-level layer that automagically selects the best GPU for the task at hand between installed options (let's assume, an Intel iGPU and your discrete GPU). For web browsing or productivity it's expected the OS will switch to the less power-hungry option, whilst for gaming and its all-cylinders philosophy, it would launch the discrete option.

However, if you're not much into ceding that kind of control to the OS itself, you can override which specific GPU is activated for a specific application. This change is made via the Settings panel with a drop down menu in Graphics Settings. This feature should be a particular boon for laptops that don't feature a power-saving technology that enables this kind of behavior, but there are some other usages for power users that might come in handy with this OS-level integration.



View at TechPowerUp Main Site
 
Wonder how this would work with streaming? Have one GPU dedicated to the game and another GPU dedicated to obs.
 
This is a nice feature, even for a desktop where many do not care about power consumption.
 
With so many GPU models out there I do no see large compatibility.
Most know commands trickery, this is sending sleep commands at GPU cores, but what if you have just one ? :)
 
Interesting, I like the sound of this.
 
Would be cool if you could use this set one GPU to run a game and assign another for reshade post process injection, but doubtful that works.
 
Laptops with NVIDIA dGPUs and Intel iGPUs have had this option in the NVIDIA Control Panel forever. Not sure why MS is moving it into Windows directly.

With so many GPU models out there I do no see large compatibility.
Most know commands trickery, this is sending sleep commands at GPU cores, but what if you have just one ? :)

Please educate yourself as to how graphics drivers work.
 
Laptops with NVIDIA dGPUs and Intel iGPUs have had this option in the NVIDIA Control Panel forever. Not sure why MS is moving it into Windows directly.



Please educate yourself as to how graphics drivers work.
That's similar but only use one Or the other, realistically a use case could be game on dgpu while outputting the tube to a second monitor via igp.

Seems like something you wouldn't use too often with a bit of thought.
 
Laptops with NVIDIA dGPUs and Intel iGPUs have had this option in the NVIDIA Control Panel forever. Not sure why MS is moving it into Windows directly.

Based on how GPU drivers work. I would say this would be better handled by the OS. Especially for notebooks where the Graphics Cards are of seperate vendors. This leaves a lot of the Compatibility issues out and lets the OS handle the grunt work. This way instead of the GPU vendors needing to work together to make this setup work they just need to make calls to the OS for this functionality to work. Honestly this should have been in the OS a long time ago.
 
Laptops with NVIDIA dGPUs and Intel iGPUs have had this option in the NVIDIA Control Panel forever. Not sure why MS is moving it into Windows directly.

Not exactly the same. The previous version (before the April 2020 refresh) of NVIDIA Optimus would load the IGP frame buffer first, check if the software being run needs more GPU power, then pass it on to the discrete GPU.

This implementation allows you to choose which GPU you want to run which software.

My curious question about this is how would it work if I run something on my RX 5700 XT screen, but the software's fullscreen/window view is on my 2080 Super. If I do that as it is right now, the drivers will attempt to pass on the software from one GPU to another. This works fine with 2D and lightweight 3D software as I can see the load "transferring" over using GPU-Z with the software flickering a bit. Software that places a huge GPU load however will either freeze up or crash, like I tried with Doom Eternal.
 
I thought this was already in ver. 2004? Gave me one helluva headache when I bought two Probooks with discrete graphics, which I couldn't use until manually assigned GPUs to power profiles (even if GPU is selected for an app, it was still defaulting to Intel iGPU).
 
I'm not seeing anything new here.
Graphics switching based on usage has been around a while.
 
I wanna know if i can use this for the Gsync/freesync hack - have a ryzen APU connected to my display with freesync on, but tell windows to render on nvidia?

I also thought this has been around for a while, unless its changed somehow
 
I've got a Ryzen 2500u laptop. Will this improve performance in gaming, in any measurable way?
 
Does it work on desktop, where you have separate video outputs on motherboard and dedicated video card?

I've got a Ryzen 2500u laptop. Will this improve performance in gaming, in any measurable way?

No.

Also, if your laptop has dedicated video card, it probably could do that already.
 
Does it work on desktop, where you have separate video outputs on motherboard and dedicated video card?

that should be its entire purpose yes - if they're able to pass one GPU's renderings to another GPU's output this could be a big boon for a lot of us (using a weakass NV GPU with freesync to display, but a 5700xt for rendering, etc etc)
 
Laptops with NVIDIA dGPUs and Intel iGPUs have had this option in the NVIDIA Control Panel forever. Not sure why MS is moving it into Windows directly.

Please educate yourself as to how graphics drivers work.

This is a bit difficult over patented technology.
ATI - AMD, this was using power saving by detecting 2D mode and GPU&Mem clocks those it is automatically lowered.
 
It's supported for quite a long time - I have one monitor connected to Intel iGPU and another to dedicated Radeon. Old Edge was able to move video decoding between GPUs on the fly when I moved the browser window. Most other browsers let you choose rendering GPU and when I move the window to other monitor the rendering gets transferred via PCIe (which btw hinders its bandwidth). It's the same with 3D apps, though not all APIs may equally like it (DX/OGL/Vulkan).
 
It would be nice to expand support to be able to choose between two different dedicated graphics cards in the system. If a game or application works better with AMD rather than NVIDIA, for example.
 
It is indeed useless "update" because Windows has always worked in this way, anyways. Light load is for light applications, heavy load and you get your discrete GPU up and running.

It is an option in Radeon Settings.
 
It is indeed useless "update" because Windows has always worked in this way, anyways. Light load is for light applications, heavy load and you get your discrete GPU up and running.

It is an option in Radeon Settings.

Radeon Settings != Windows.

Also since when Radeon Settings can allow you to switch between loading dedicated video card or IGP, especially non-AMD ones? :O
 
Back
Top