Key Takeaways
- Intel has launched three new Xeon 6 server processors with enhanced turbo speed capabilities.
- Nvidia will use one of these, the Xeon 6 6776P, in its upcoming DGX B300 AI system.
- A new “Priority Core Turbo” feature significantly boosts the clock speed of select CPU cores for demanding tasks.
- These specialized chips are designed to better handle the intensive workloads of modern AI applications.
Intel has just pulled the wraps off three new server processors, part of its Xeon 6 family, also known as Granite Rapids. These chips are built for serious performance and come with some clever new ways to boost their speed.
One of these fresh processors, the Xeon 6 6776P, is already making headlines. Tech giant Nvidia has selected it to power its new DGX B300 AI server, according to a report from heise.de.
Nvidia’s DGX B300 system is designed as an AI workhorse. It will feature two of these Intel Xeon 6 6776P processors working alongside eight of Nvidia’s own B300 “Blackwell Ultra” AI accelerators, which are their fastest to date.
To achieve these performance boosts, Intel is introducing new turbo functions. A key one is “Priority Core Turbo” (PCT). This allows the system to dynamically push certain high-priority cores to much higher clock speeds when needed.
For instance, PCT can take eight CPU cores from a typical turbo speed of 3.9 GHz all the way up to 4.6 GHz. To keep power consumption and heat in check, the other CPU cores then run at their standard base clock speed.
Intel notes that AI systems particularly benefit from these higher CPU clock speeds. This is because AI tasks often involve sending many computing jobs sequentially to the powerful GPU accelerators, which limits how much the work can be spread across many CPU cores at once.
The three new models introduced are the 6962P, the 6776P (Nvidia’s choice), and the 6774P. The 6776P offers 64 performance cores and robust connectivity suited for AI. The 6774P is designed for single-processor systems, offering more lanes for expansion cards.
The most powerful of the trio, the 72-core Xeon 6 6962P, uses a larger physical platform that supports more memory channels for extremely data-intensive tasks. These releases show Intel’s commitment to providing specialized hardware for the rapidly growing field of artificial intelligence.
