2023-05-17 09:56:15
Yes, we are talking about Artificial Intelligence, and its derived disciplines such as Deep Learning or Machine Learning, the company that will come to mind at the first opportunity is NVIDIA. While its rival in graphics cards, AMD, has been a bit behind in this in recent years. However, the company led by Lisa Su does not want to be left behind in this field. Your plan? Bring Ryzen AI to all your CPUs and GPUs?
At CES, AMD announced the first fruits of the Xilinx acquisition, the implementation of the so-called XDNA Engine in its Ryzen 7040 mobile processors under the brand name Ryzen AI. In the field of hardware, it is a support processor or coprocessor capable of accelerating the algorithms and, therefore, the programs that use them. artificial intelligence algorithms. The concept behind it all? The same as any unit of this type, offloading said work to the CPU and even to the GPU in order to derive its power to other tasks.
What is AMD’s plan to implement Ryzen AI in the rest of its CPUs and GPUs?
We have to start from the fact that the XDNA units used in Ryzen AI to move from desktop systems would be integrated into the IOD chip where the integrated memory controller and peripheral interface are located. So it would share space with the integrated GPU, audio and video decoders and other elements.
Actually, much of the usefulness of this technology seems to be more focused on the graphic section, specifically on computer vision and with a focus on streaming via videoconference, with utilities such as: image self-centering
, eye contactor correction it is included automated background changes.
So Ryzen AI would not have enough power to compete in spaces where the power required for Deep Learning and Machine Learning are much larger. In the words of AMD’s David McAffe in an interview with PC World:
Possibly Ryzen AI will become more of a testing tool rather than something that drives the daily value of the device.
To this, McAffe has added that we are likely to see a Threadripper to train an algorithmbut what I wouldn’t use Ryzen AI. The reason? Said CPU has much more power than what said CPU can contribute. So while we may see some Ryzen desktops with such drives for task counterparts to those seen in laptops, if we’re already pulling workstations and servers, the answer is no.
All AMD roads lead to AI
The way in which the veteran company can implement solutions for Deep Learning and Machine Learning depends on many things. Especially the limitations assigned in consumption and cost for each of the ranges of processors and graphics cards.
For example, while Ryzen AI will not be implemented in the same way as in laptops, rumors and patents point to one or more additional dedicated hardware chiplets in the future AMD EPYC. What is expected? A version with more cores and, therefore, much more powerful of the XDNA technology that we have seen in the Ryzen 7040, but that will continue to be sold under the Ryzen AI trademark.
Our opinion? We believe in renewing the IOD chip for future EPYC and Threadripper. In any case, it does not appear that AMD has any plans to adapt the Intel’s AMX units, At least for the moment. What’s more, we can’t rule out the pairing between its CPUs and a future GPU, especially now that RDNA 3 has support for WMMA. In any case, we still have to see if this bet will be integrated into all models or there will be a special branch of products. What we do know for sure is that they will all use Xilinx technology.
1684326465
#AMD #bring #Artificial #Intelligence #CPUs #GPUs #Ryzen