Intel begins shipping its Xe-based Server GPU

[Karlston]

Intel begins shipping its Xe-based Server GPU

 

Today, Intel announced that it's now shipping its first discrete graphics card aimed at data centers, which it's simply calling the Intel Server GPU. If you've been following along, Intel gave us a thorough look at its plans for the Intel Xe graphics architecture back in August, and it included a server GPU based on it.

 

More specifically, the Intel Server GPU is based on the Xe-LP (low power) microarchitecture, which is the same one used in the integrated graphics in 11th-generation Tiger Lake processors and the recently-released Iris Xe MAX discrete graphics. As the name suggests, it's meant to be power efficient, and it has a 128-bit pipeline and 8GB of onboard memory. The Intel Server GPU is touted as a "high-density, low-latency" solution for Android cloud gaming experiences, while also being more affordable than competing solutions from Nvidia, based on Intel's estimates. The GPU can be combined with Intel's scalable Xeon processors, allowing customers to scale the graphics capabilities without increasing the server count itself.

Intel also revealed the H3C XG310 graphics card, which packs four Intel Server GPUs, and the company claims a two-card system could handle more than 100 Android cloud gaming users at a time, with up to 160 users depending on the system configuration and game. Tencent's Arena of Valor and Honor of Kings managed over 100 users per system on the company's own streaming service in China.

 

Aside from that, Intel also revealed today that its new oneAPI will see a Gold release in December, with toolkits available for free locally and on Intel DevCloud, along with a paid version that includes international support from Intel engineers. Since oneAPI is meant to be a unified programming model, Parallel Studio XE and System Studio will immediately transition to oneAPI products. Intel also said that it plans to add support for its datacenter products in Intel Graphics Software, as it moves towards offering a scalable architecture with Xe. Currently, its software already supports integrated Iris Xe graphics in Tiger Lake processors and the Iris Xe MAX GPU.

 

 

Intel begins shipping its Xe-based Server GPU