Nvidia GeForce Driver Promises Doubled Stable Diffusion Performance

Nvidia today announced a new GeForce Game Ready Driver update that’s sure to catch the eye of anyone dabbling in local Stable Diffusion installations.up to date GeForce Game Ready Driver Release 532.03It will be released later today and is packed with learnings from. Microsoft’s Olive Toolchainis a hardware-aware model optimization tool that aims to fully couple diffusion model processing to graphics card capabilities.
According to Microsoft, Olive has the ability to change its base model depending on the available hardware capabilities, whether it’s a local installation, the cloud or the edge. Olive is a way to abstract all the different machine learning (ML) acceleration tool chains distributed by different market players. So part of its mission is to reduce market fragmentation around ML acceleration technologies.
Nvidia says the new Olive integration into GeForce will enable improved performance for AI operations that leverage the fixed-function RTX Tensor Cores present in Nvidia GPUs. With the Olive-optimized version of the Stable Diffusion text-to-image generator (used in conjunction with the popular Automatic1111 distribution), the performance is more than double his.
said Pavan Davuluri, corporate vice president of Windows Silicon and Systems Integration at Microsoft. “He works with Nvidia on hardware and software optimizations to give developers an innovative, high-performance, and easy-to-deploy experience.”
There shouldn’t be any reason AMD can’t implement Microsoft’s Olive toolset into their graphics drivers, but it’s unclear at this time how much these optimizations actually depend on Tensor’s performance (although if AMD has an equivalent hardware (Remember, I don’t own any hardware) For now, Nvidia users in particular seem poised to get even better performance than they can get from AMD’s GPUs (at least for now).
Interestingly, Nvidia reveals that the Tensor Cores of the RTX architecture are primarily responsible for the card’s performance in ML acceleration, while the graphics cards listed as supported in version 532.03 have includes cards from the GTX 700 family and newer. It will be interesting to see if these older architecture cards without Tensor Cores also see similar his 2x performance gains in ML acceleration.
In addition to Microsoft’s Olive integration, the latest GeForce Game Ready v532.03 driver also brings support for the NVIDIA RTX 4060 Ti while fixing known bugs that are still present. Age of wonders 4 and bus simulator 31.