The GCC and LLVM Clang open-source compilers have landed support for the NVIDIA Olympus cores for NVIDIA’s Vera CPU that is part of their next-gen Rubin microarchitecture succeeding Blackwell.
Last week at NVIDIA’s GTC conference when talking about Vera they noted they are using “custom” Olympus cores that are twice as fast as the Grace CPU making use of Neoverse-V2 cores. With Vera there will be 88 of these custom Arm cores named Olympus.
Now with the patches for LLVM/Clang and GCC, we have a bit more insight into NVIDIA Olympus. The GCC patch is merged for the upcoming GCC 15.1 stable release and there is the LLVM patch for the LLVM/Clang 21 release later in the year. Both patches carry a similar message:
“This patch adds support for the NVIDIA Olympus core.
This does not add any special tuning decisions, and those may come later.”
So we lack any special tuning insight but the compiler patches note that Olympus with NVIDIA “N” cores is making use of the Armv9.2-A ISA and is akin to the Neoverse-V3.
Additional Arm ISA capabilities on top of Armv9.2-A include SVE2_BITPERM, RNG, LS64, MEMTAG, PROFILE, FAMINMAX, FP8DOT2, LUT, SVE2_AES, SVE2_SHA3, SVE2_SM4.
With the LLVM patch for now at least its making use of the Neoverse-V2 scheduling model. Olympus is using two distinct CPU part numbers of 0x10 and 0x010.
These compiler patches confirm the Armv9.2-A ISA for NVIDIA Vera CPUs and some of the extensions supported but not much more insight with seemingly quite similar to the Neoverse-V3 that is Armv9.2-A based as well without these patches doing much to shed light on these “custom” Olympus cores.
