AMPERE launches six new CPUs with up to 192 nuclei and 12 DDR5 memory channels

AMPERE has updated its official website this week with six processor models in a new Ampreone M line. The “silent” launch brings products with a count of 96 to 192 nuclei and support for 12 DDR5 memory channels. As the specifications indicate, we are talking about CPUs for professional use.

It is curious that the company has not made a big announcement or fanfare for its new products, but Ampreone M seem to position themselves like a “bridge” between AMPERE’s current products and their future generation. They maintain the same 5NM Lithography as TSMC for their ARMV8.6+nuclei, and use a cache of up to 64MB.

The biggest difference in new line specifications is in its 12 RAM channels, which can support systems with up to 3TB of DDR5-5600 memory. Thus, they can offer an interesting upgrade to specially demanding workloads, but they do not advance much in processing.

Related News:

Tom’s hardware estimates that the release is the most position as a preparation for the MX ampereone, which arrive only later. This new line will keep the 12 DDR5 channels, which make their debut now, while offering CPUs up to 256 nuclei, manufactured in TSMC’s N3 (3NM) process.

More details of the new AMPERE CPUS

While waiting for the MX Ampereone Generation, AMPERE customers can now opt for six new SKUS with 12 DDR5 channels. Below, the options with your counts of nuclei, clocks and TDP:

• AmpereOne A192-32M: 192 núcleos | 3,2GHz | 348W
• AmpereOne A192-26M: 192 núcleos | 2,6GHz | 278W
• AmpereOne A160-28M: 160 núcleos | 2,8GHz | 262W
• AmpereOne A144-33M: 144 núcleos | 3,3GHz | 334W
• AmpereOne A144-26M: 144 núcleos | 2,6GHz | 239W
• AmpereOne A96-26M: 96 núcleos | 3,6GHz | 331W

The Amperone M line is compatible with 96 PCIE 5.0 lanes and can divide into connections up to X4. There are 24 device controllers for compatibility with a variety of system components, including common products such as SSDs, network cards, etc.

Finally, products do not support multithread, so this count is similar to nuclei numbers in each model.

Source: AMPERRE