Nvidia's Vera CPU targets extremely high single-core performance — not for PCs yet

Nvidia’s Vera CPU and its Olympus cores drew little attention amid the DLSS 5 drama at GTC, but the chip looks substantial. It is an Arm-based, custom design with an 88-core structure, and Nvidia says it is intended for AI servers. Nvidia describes a different approach to simultaneous multithreading called “Spatial Multithreading,” which avoids time-sliced context switching.

The company says an Olympus core is 50% faster than any single x86 core for compilation, scripting, and compression inside an “agentic sandbox container,” while being 90% more efficient. Vera pairs LPDDR5X memory with a very wide memory bus to deliver 1.2 TB/s of bandwidth, offers a 16-lane PCIe Gen 6 interface, and the die includes support for 176 threads and 162MB of L3 cache.

Nvidia has framed Vera for AI factories and agentic AI and has not positioned it for PCs. Rumors suggest Olympus could appear in a second-generation PC processor codenamed N2, after the first-gen N1/N1X which uses the GB10 Superchip and off-the-shelf Arm Cortex cores.

nvidia vera, olympus cores, arm, spatial multithreading, ai servers, lpddr5x, pcie gen, memory bandwidth, single-core, agentic ai