An unknown graphics card from Nvidia has appeared in online databases: codenamed PG506, this server GPU is a next-generation NVLink-enabled graphics module that appears to be the successor to the Volta V100, the so-called "Ampere" This is the best clue to date regarding the release of the so-called "Ampere" architecture.

All that is available as a specification is the model name "PG506," registered today, April 7, 2020, and a brief description: "server graphics module." However, this code speaks to the purpose of the card, as it follows on from the PG502 Volta accelerator, which was announced at the May 2017 GTC and released later that year.

The PG500 was a PCIe-based version of the Tesla V100 and was the first one that actually looked like a traditional graphics card before the Titan V was released in December.

PG504 was the latest release in the same series of server GPU modules as the 32GB version of the V100, which was announced again at GTC in March 2018. [PG504 was registered in July and released five months later in December. PG504 was registered in July and released five months later in December.

Thus, the reason today's PG506 listing is so important is that it anticipates the broad release of new server GPUs this year.

Such a card would be abbreviated "A100" if one follows the expected Ampere codename. This is because we ourselves are not entirely certain that the Ampere codename is for Nvidia's next-generation architecture. The codename Ampere, named after physicist André Marie Ampere, is the real deal for the time being.

The next-generation server architecture was strongly rumored to be first unveiled at Nvidia's Graphics Technology Conference in San Jose this year. However, GTC was moved online, Nvidia's major announcement was postponed, and we were left in the dark for a bit longer.

The original plan, as widely suspected, was to release a new server card at GTC 2020. This card would presumably utilize high-bandwidth NVLink connectivity through the server baseboard, and as such is also found today at RRA under the codename "P4597."

These codes do not necessarily support the architectural changes made by PG506. But does all of this portend the release of GeForce graphics cards? Volta and Turing are a year apart, and today's announcement only confirms the NVLink compatibility part of the server, not the PCIe GPU needed for a gaming PC.

Nvidia forked GeForce and server parts with Volta and instead chose to carry the gaming torch by leaning on the Turing architecture with the RTX 20 series after Pascal's popularity and power waned. release of GeForce preceded by Nvidia's (now a server-first company) datacenter announcement, it has been suspected that the same thing would happen this time around. While it is suspected that Nvidia will realize rumors of a new generation of GeForce graphics cards to fight off AMD's RDNA 2.0 graphics cards towards the end of the year,

so this is not necessarily the best details of imminent changes.