Today, Nvidia released a video on their Youtube channel in which they explained how Nvidia Max-Q works. In the video, Rouslan Dimitrov, the Distinguished Engineer, and Notebook Platform Architect at Nvidia explained how Nvidia is making the 2022 RTX laptops thinner and faster. In laptops, power dissipation is shared between the major components the biggest two being the GPU and the CPU. So it is very important to get maximal efficiency from both of them.
That being said, Nvidia developed the CPU optimizer a key new four-gen MaxQ technology that makes the CPU more efficient when gaming or creating. Rouslan stated that the extra power saved can now be shifted between the CPU and the GPU wherever it is needed most.
Nvidia did the following six things to make laptop CPUs even more efficient:
1. Nvidia worked with CPU vendors to develop a new low-level framework. This enables the GPU software to optimize the performance temperature and power of the latest CPUs.
2. The CPU which are uncoordinated with the GPU will be for higher frequencies and this results in the consumption of more power. For games and creative apps, this is not efficient. In order to fix this issue, Max-Q technology causes the CPU at peak efficiency instead of peak frequency. This results in saving power which can be transferred to the GPU to boost overall performance.
3. Nvidia profiles the major graphics APIs for CPU inefficiencies and streamlines the entire GPU drivers stack. The driver runs multiple threads across the CPU cores and Nvidia optimized how they synchronize. They re-architected the internal data structures, so the threads don’t need to block and wait on each other. The individual function definitions were also tweaked to reduce co-overhead. With everything combined, the CPU efficiency has gone up by as much as 2x in directed API.
4. Another key technique for optimizing system power is to move work from the CPU to a dedicated command processor on the GPU. Nvidia has used it to offload the CPU from low-level tasks. An example is command validation, which performs pointer verification and balance checking. This routine task now happens on the command processor enabling both the GPU and CPU to focus on other computations.
5. Nvidia has also optimized how the CPU sends works to the GPU and collects the results in a creator app. The driver now puts the CPU into deep sleep while the GPU does the work. After the GPU completes it immediately wakes up the CPU improving responsiveness and efficiency.
6. The company has also optimized Nvidia reflex. One of the core reflex techniques is to make the game read the mouse and keyboard states is late as possible before a frame is displayed. This makes the image your eyes see, reflect the most recent possible input reducing latency. This has been done by using the same approach we’re using creator apps by putting the CPU into deeper low power states until it is time to sample the inputs.
The combination of all the four-generation Max-Q technologies dramatically reduces CPU power but as much as 55 percent. The whole suit of Max-Q technologies is adaptive, which adjusts to workload changes in real-time. It only kicks in when it makes sense to always ensure peak efficiency and peak performance.
