- The CPU chief of the AMD client says he is looking for dedicated NPU accelerators
- These would be the equivalent of a discreet GPU, but for the tasks of AI
- Such boards would decrease the demand of the high -end GPUs, since they would no longer be bought for the work of AI, as they are in some cases
AMD is looking in the future in which not only produces independent graphics cards for desktop PC, but similar tables that would be the equivalent of an AI accelerator, a discreet NPU, in other words.
CRN reports (through WCCFTECH) that Rahul Tikoo de AMD, head of his CPU client business, said Team Red is “talking to customers” about “use cases” and “potential opportunities” for such a dedicated NPU accelerator card.
CRN points out that there are already movements in this line, such as a laptop Dell Pro Max Plus incoming, which is established to have a pair of inference cards of 100 PC Qualcomm AI. Those are two discreet NPU boards with 16 AI and 32 GB cores each, for 32 cores of AI and 64 GB of RAM in total.
To put that in perspective, the current integrated NPUs (in chip), such as those of Intel’s Lunar Lake CPU, or the Ryzen AI AI chips, offer around 50 tops, ideal for co -pilot+ PCs, while we are seeing up to 400 tops with the users of AI 100 of Qualcomm AI 100.
Tikoo observed: “It is a very new set of use cases, so we are seeing that space carefully, but we have solutions if you want to enter that space, we can do it.”
The AMD executive would not be attracted to provide a track of a time frame in which AMD could be planning to perform such discreet ambitions from NPU, but said that “it is not difficult to imagine that we can arrive quite quickly” given the “amplitude” of the technologies of the network team. “
Analysis: potentially eliminate the pressure from high -end GPU demand
So, does this mean that it won’t be long before you looking to buy your desktop PC and reflect with a discreet NPU along with a GPU? Well, not really, this is not yet consumer territory as such, as noted, it is more AI Power users, but it will have an important impact on everyday PCs, at least for enthusiasts.
These independent NPU cards will only be necessary by people who work in heavier tasks with their PC. They will offer benefits to execute large artificial intelligence models or complex work loads locally instead of the cloud, with much more receptive performance (dodge the delay factor that inevitably leads to the mixture when working online, to the cloud).
There are obvious privacy benefits of maintaining the work on the device, instead of addressing the cloud, and these discreet NPUs will be designed to be more efficient than the GPUs that assume this type of workloads, so there will be energy savings.
And it is here that we reach the crux of the matter for consumers, at least the enthusiastic PC players seeking to buy more expensive graphics cards. As we have seen in the past, sometimes people who work with AI buy high -end GPU, such as RTX 5090 or 5080, for their platforms. When dedicated NPUs leave AMD (and others), they will offer a better option than a high -end GPU, which will remove market pressure for graphics cards.
So, especially when a new range of GPU comes out, and there is an inevitable race to buy, there will be less general demand for high -end models, which is good news for supply and prices, for players who want a graphics card to, well, play PC games and not accelerate the workloads of AI.
It continues in the development of these independent NPUs, so: in the end it has to be something good for players. Another thought for the future much farther is that eventually, these NPUs may be necessary for the routines of the games, when complex NPCs driven by AI occur. We have already taken some steps along this path, in terms of the cloud, although that is something good or not a matter of opinion.
