It might not be long before the built-in AI makes your graphics card less of a power hog
PowerColor is bringing an NPU on board its graphics cards, or that’s the plan with AMD Radeon boards, it seems.
As noted by Harukaze5719, one of the regular leaks we follow on
PowerColor integrated NPU into Radeon GPU. (Still in development) Through AI they claim that energy consumption has been reduced by 22%. More savings than AMD Power Saving. pic.twitter.com/9I8iEikotDJune 7, 2024
NPUs are currently included with the very latest CPUs and are present on the chip to speed up the AI ββworkload and take some of the load off the processor itself, but with PowerColor’s graphics cards that won’t be the plan.
Instead, the board maker is introducing an NPU to intelligently adjust the power consumption of the Radeon graphics card while gaming, meaning less power is used when you enjoy your favorite PC games.
According to the first teaser for the new hardware-based AI tuning, we can expect 22% less energy consumption (as tested in a few games: Cyberpunk 2077 and Final Fantasy XV).
The catch? Well, if VideoCardzwho noted this points out that the power savings mean a drop in frame rates (although PowerColor doesn’t admit this, or at least skirts the issue by saying the graphics card maintains “high performance” with the technology).
Based on a test with Final Fantasy Fan speeds also ran a lot slower β good news on the noise front, of course, although that did mean a bit of a temperature increase for the GPU.
Analysis: A careful and difficult balancing act
The idea is to strike a balance between not letting the temperature rise too much and dialing back the fan speed, requiring a small drop in performance for a big energy savings. In theory it’s a good idea β especially as an option β although there is already software (from AMD and elsewhere) that can perform a similar task.
The point is that PowerColor’s new NPU does a better job, as you might imagine, not only in terms of the sheer energy savings achieved, but also that it is constantly being monitored in the background and adjusts variables intelligently. way, depending on which game (or app) you are using at the time. (Which also sets this apart from manually ‘undervolting’ your GPU, especially considering that this kind of tinkering is beyond less tech-savvy gamers anyway).
Additionally, this is still in its early stages of development, so if this technology is further promoted we may see even better results (and perhaps less impact on performance).
It’s great to see hardware makers thinking about other uses for NPUs, and we didn’t see this coming. Perhaps built-in AI via a dedicated NPU has a place in the best graphics cards of the future, and we’ll see other card makers introduce a similar take on taming a GPU’s power consumption.
That could be especially useful for heavy gaming GPUs that draw a lot of power, although in next-gen terms that’s probably an issue that Nvidia suffers from more than AMD (or Intel, for that matter).