limitations on laptop, power and a consumption of 1 W

limitations on laptop, power and a consumption of 1 w

Tom Henry

limitations on laptop, power and a consumption of 1 W

In the words of Lisa Su, AMD is not late for the appointment when it comes to AI, since it is a growing market that will reach five times its current size. However, the brand’s presence in terms of Artificial Intelligence was at least limited outside of its Instinct GPUs for HPC. Well, AMD has given some additional information about Ryzen AI in their 7040 for laptops.

The purchase of Xilinx first and Pensando later are key to the future of the brand, just like the acquisition of ATI Technologies in the 2000s and of NexGen in the 1990s. The goal? Expand AMD’s portfolio with technologies that, if they had to be developed internally, would have taken several years to be implemented in the brand’s CPUs. The first sample of them we have in the implementation of XDNA and Ryzen AI in their latest processors for laptops.

Ryzen AI in 7040 APUs for laptops, but what does this technology bring?

Not much at the moment, since Ryzen AI in AMD’s 740 is nothing more than a very low-powered and fairly simple implementation. This is still an NPU with the ability to process up to 4 neural networks at the same time simultaneously, but with a very limited power, since it does not exceed 10 TOPS of power and for the moment its access is at least limited. not only at the user level, but also for application programmers.

And it is that for the moment its only utility, available only in version 2H22 of Windows 11, is to be able to do a series of tricks with the webcam at the operating system level, which Microsoft and AMD have baptized as “Windows Studio Effectsand consist of things like auto framing, eye contact and fund manipulation effects. All of them can be done without problems from the integrated GPU of the Ryzen 7040, but the great advantage of Ryzen AI is that it does this work by freeing up the graphics processor and consuming less than 1 watt.

Unfortunately, not all laptop manufacturers will use Ryzen AI in their processors, so it will be necessary to check which models have this feature and which do not, if we are interested in making use of it. Although for the moment its implementation is limited and it is still a technical demo.

AMD will release the SDK at the end of the year

AMD Ryzen AI SDK route map

In another order of things, AMD will release the development kit so that application creators can implement their neural networks and Deep Learning algorithms to run under Ryzen AI. Although due to the low power of its implementation in the Ryzen 7040, these will be low-powered, it is expected that various applications will use them to speed up certain functions. However, we will have to wait a few months for this, specifically at the fourth quarter of this year.

The Ryzen AI SDK will support to all functions of the XDNA unitjust like him support for models created in PyTorch, ONYX and TensorFlow. So we may see a flood of apps developed for mobile on PC. And it is also that, at the same time and with the launch of Intel’s Meteor Lake, they will finally become possible on PC without their performance being affected by not having a unit of this type.

Finally, this is a very low power unit and we know that AMD will release higher capacity versions to handle much more complex models and tasks that require much more power. At the moment the implementation of Ryzen AI in the 7040 is more of a statement of intent, which is rather timid.

Leave a Comment