Sunday, February 5th 2023
Running Discord Lowers NVIDIA GPU Memory Clocks by 200 MHz, Company Posts Workaround
The Windows app of Discord, the popular social-networking software, apparently trims the graphics card memory clock of NVIDIA GPUs by an innocuous 200 MHz, or so observe gamers. NVIDIA GeForce GPUs dynamically adjust memory clock speeds in response to load, as part of their power-management. Ideally, with gaming workloads, the GPU is supposed to hit its maximum rated memory frequency, but some keen-eyed gamers with monitoring tools noticed that with the Discord app running in the background, the memory clock tops out at T-minus 200 MHz (i.e. if it was supposed to be 7000 MHz, it tops out at 6800 MHz). Even under the infernal stress of Furmark, something that's designed to push memory clocks to the maximum rated speeds until the graphics card runs into thermal limits; the memory clock is seen falling 200 MHz short.
NVIDIA took note of this issue, and assured that a fix is on the way in a future GeForce driver update. In the meantime, it posted a DIY workaround to the problem that involves downloading the GeForce 3D Profile Manager utility, making the utility "export SLI profiles" (applicable even to single-GPU machines), editing the exported SLI profiles file as a plaintext document, and importing the profile back. This basically alters the way the driver behaves with the Discord app running. The NVIDIA 3D Profile Manager utility can be downloaded from here, and step-by-step instructions on using it to fix this issue, here.
Update Feb 6th: NVIDIA released a GeForce driver application profile that automatically downloads to your driver, which should fix this issue. You don't need GeForce Experience to receive the update.
Source:
LinusTechTips (forums)
NVIDIA took note of this issue, and assured that a fix is on the way in a future GeForce driver update. In the meantime, it posted a DIY workaround to the problem that involves downloading the GeForce 3D Profile Manager utility, making the utility "export SLI profiles" (applicable even to single-GPU machines), editing the exported SLI profiles file as a plaintext document, and importing the profile back. This basically alters the way the driver behaves with the Discord app running. The NVIDIA 3D Profile Manager utility can be downloaded from here, and step-by-step instructions on using it to fix this issue, here.
36 Comments on Running Discord Lowers NVIDIA GPU Memory Clocks by 200 MHz, Company Posts Workaround
I do and my memoryspeed doesn't go down.
Quote:
"Hardware-accelerated GPU scheduling: It allows the video card to directly manage its video memory, which in turn significantly improves the performance of the minimum and average FPS, and thereby reducing latency. It works regardless of the API used for games and applications such as DirectX / Vulkan / OpenGL."
Just a thought...
İmo it is still %100 Nvidia problem, if you want software companies to rely on your products, you should take care of that.
I would understand that content creation software ike Da Vinci resolve ,Photoshop, etc could use Hardware acceleration but Why would Discord, or (max, or blender claymode/gray mode viewport) would use 400w or 450w to do something that even an embedded gpu is capable of doing ?
Afaiu, Nvidia determined the load from Discord can be ignored, so they updated profiles to push the VRAM disregarding Discord. Here's hoping future Discord versions won't tax the GPU even more.