NVIDIA may release a standalone Arm server processor without a

NVIDIA may release a standalone Arm server processor without a GPU accelerator

NVIDIA CFO Colette Kress managed to attend two thematic conferences for investors this week, namely at Wells Fargo event She was asked if it made sense to share the Grace CPU with a server segment that does not have a companion GPU. An NVIDIA representative answered this question in the affirmative.

    Image source: NVIDIA

Image source: NVIDIA

The speech of the company’s financial director at conferences was preceded by the publication of a new version of the presentation for investors, which was published last Monday in the profile section of the NVIDIA website. The schedule for the release of new architectures, previously established in similar documents, is also reflected in one of the slides of the recent presentation to investors. Let us recall that the company is committed to releasing new computer architectures in the server segment every year. At the same time, Colette Kress, in a conversation with the host of the event, admitted that new products could actually appear even more often, even before the next architecture is released.

The Arm-compatible architecture of Grace CPUs recently made its debut in the server segment, NVIDIA partners will begin welcoming their representatives as part of it Hybrid module GH200, combines an accelerator with a layout familiar to graphics processors and a central processor. NVIDIA’s roadmap lists its successors in the “CPU + GPU” category.

This gave rise to a logical question for the host of the Wells Fargo event: will NVIDIA CPUs appear on the server market without the mandatory GPU add-on? Colette Kress answered this question more emotionally than usual and confirmed twice in a row: “The possibility of an exit is easy Grace is”. In their opinion, new scenarios for the use of NVIDIA products in data centers will imply the emergence of standalone central processors like Grace. Every year they are successively combined in tandem with representatives of new graphics architectures: for the first time next year GB200 is displayed (Grace Blackwell) and in 2025 the GX200 will be introduced.

By the way, the expansion of the offering in the server segment and the transition to more frequent architectural updates forced the company to increase by 40% the amount of advance payments to partners involved in the production process of chips developed by NVIDIA. This is quite natural, especially given the growing demand for the earliest representatives of this product category.

About the author

Dylan Harris

Dylan Harris is fascinated by tests and reviews of computer hardware.

Add Comment

Click here to post a comment