If we are talking about artificial intelligence, the first name that comes to mind Cat OPENAI, Co -pilot from Microsoft and Gemini From Google. For a few days, he joined In depthA new open source AI that any user can ask locally without using an internet connection.
To be able to use it, it is necessary to have a state graphics card to obtain the highest performance. AMD published a performance test a few hours ago when it indicates that the RX 7900 XT It offers much higher performance at Nvidia RTX 4090.
If we do not have a dedicated graphic, something quite common in the portable teams, with a Neuronal processing unit It's more than enough.
If we are talking about Neuronal processing unit (NPU), we are talking about a processor designed to execute Artificial intelligence tasks Local without depending on a dedicated graph or an internet connection.
Since the middle of -2024, with the presentation of the first generation of processors with Arm architecture Of Qualcomm, Microsoft invented the term Co -pilot + PCa term that applies only to portable equipment that They include an NPU with performances equal to or superior to 45 tops.
The in -depth arrival noted a significant agitation of the companies dedicated to the AI, because it was formed with a much lower economic cost than the Openai models used by the Nvidia GPUs, wondering if we are in a bubble of IA and causing a spectacular drop in the decline in the decrease in the company's share of the company carried out by Jensen Huang.
Qualcomm Snapdragon X processors with Depsek support
Microsoft, one of the companies which, from the start, bet very hard for the artificial intelligence of Copilot, available in Windows 11 and in all desktop applications, has just announced that it already has an optimized version Ready for the NPU of the Qualcomm Deepseek-R1 processors.
In this way, any equipment compatible with CO -Philot + PC can take advantage of this NPU in the most effective way possible. According to the company, this optimized version for Deepseek NPU will offer time rates to the first token and very competitive performance and its functioning will barely affect the battery consumption
It should be remembered that the neuronal processing unit was specifically designed to perform local AI tasks without having dedicated graphics and has a very small battery consumption to perform the same tasks as we can do on an office PC or laptop using a graphic.
This initial version will have Deepseek-R1-Distill-qwen-1.bModel analyzed by a research team on AI of the University of Berkeley and which has a high percentage of correct responses. In the coming weeks, supporting models with 7,000 and 14,000 million parameters will also be introduced.
First Qualcomm, then the others
The fact that Qualcomm processors are the first to receive the support of Depseek should not surprise because the relationship between the two companies has been very good in recent years. Another test of this excellent relationship was observed in the middle of -2024 when Microsoft and Qualcomm presented the new Snapdragon X processors, teams which, as we mentioned above, created COPILOT + PC certification.
So many intex with the Core Ultra 200 Like AMD with AI Ryzen, they also have this certification, because the two integrate an NPU with a minimum yield of 45 peaks.
