Ampere Computing is leveraging TSMC's 3nm process node for its AmpereOne-3 CPU, offering up to 256 cores in the coming year.

The AmpereOne-3 CPU leverages TSMC 3nm and Chiplet technology with up to 256 cores for data centers.

Ampere has been well entrenched in computing markets, particularly the cloud segment, as the firm gained traction in markets with its lineup of cloud-focused processors, known as the most “load” out there. Is. Base numbers for the ship are included in the company's release. AmpereOne processor choice, which comes with 192 physical cores, a TDP of 350W and eight DDR5 memory channels, competing with the likes of Intel's Xeon Scalable and AMD's EPYC. However, the firm has now revealed future plans, instead revealing new SKUs with “surprising” specifications.

The next platform Ampere is reportedly preparing to launch its second generation of AmpereOne processors. Speaking to the media outlet, the firm's chief product officer, Jeff Wittich, revealed that the new AmpereOne-2 is expected to drop sometime later this year, with an upgraded “A2” core architecture, 12 To guarantee the best with memory channels. performance

Combining such upgrades, AmpereOne-2 believes that AmpereOne-2's overall performance will increase with a 33% increase in DDR5 memory controllers on board, as well as 50% more memory bandwidth.

Apart from the AmpereOne-2, the company also plans to release the third generation of the lineup by 2025, and based on what has been revealed, the specifications are something we haven't seen much of in the markets. AmpereOne-3, or the third generation, will debut with a total core count of 256 and feature TSMC's 3-nanometer process in a chiplet configuration. The architecture will reportedly include PCIe-Express 6.0 and a dozen DDR5 memory controllers. With this, Ampere plans to compete with some of the top contenders in the industry including NVIDIA.

We are moving very fast towards computing. This design has a lot of other cloud features – things about performance management to get the most out of all those cores. In each chip release, we're going to make what would normally be considered generational changes in CPU cores.

We're also taking the chiplet approach with this 256-core design, which is another step forward. Chaplets are a huge part of our overall strategy.

– Jeff Wittich via NextPlatform

We cannot comment on whether Ampere Computing can manage to gain the same attention that NVIDIA did in the AI ​​segments. Yet, based on the firm's plans, they are sure to play a relatively dominant role going forward.

News Source: The next platform