PantheonSim / GPUs / NVIDIA H200 NVL

Simulated NVIDIA H200 NVL: a GPU profile for CUDA and ROCm testing

PantheonSim, the GPU simulator for CUDA and ROCm, simulates the NVIDIA H200 NVL (Hopper, 9.0) as the profile nvidia/h200-nvl. The PCIe NVL version of the H200, with 141 GB of HBM3e. Its profile is built from public documents and not yet measured against a physical card.

The profile at a glance

These are the numbers a program can ask this simulated GPU, and the values nvidia-smi show for it. They come from the profile file in the simulator's repository, pinned to one commit of it.

Property	Value
Profile id	`nvidia/h200-nvl`
Vendor and architecture	NVIDIA, Hopper
Compute capability	`9.0`
Device memory	141.0 GiB, HBM
Streaming multiprocessors	132
Warp size	32
Max threads per multiprocessor	2048
Shared memory per block (opt-in maximum)	227 KiB
L2 cache	50 MiB
Memory bus width	6144-bit
Power limit (profile)	600 W
Maximum clocks (profile)	1755 MHz shader, 3201 MHz memory
ECC memory	yes
PCIe link reported	Gen5 x16
Tensor or matrix cores	yes
Verified (measured on a physical card)	not yet measured

What this GPU is for

The PCIe NVL version of the H200, with 141 GB of HBM3e.

Hopper is compute capability 9.0. The simulator implements Hopper's warpgroup matrix instructions (wgmma), the tensor memory accelerator (TMA), thread-block clusters with their barriers, and distributed shared memory, and checks them against CUTLASS and CuTe's own Hopper unit tests.

Among the 37 NVIDIA profiles it ranks 5th by device memory (141.0 GiB) and 12th by SMs (132), counting the largest as first.

Is this profile verified?

No. This profile is marked not verified: it is built from public documents only, and nothing in it was measured on a physical card. Some figures are derived or copied from the nearest profile where the vendor publishes none, so treat them as approximate, and memory sizes as upper bounds. Programs still run on it, and it still answers every call as the architecture's rules require, but a value such as a clock or a limit may differ from a real card's.

Simulate this GPU

Build the simulator (see the docs) or open the playground, then:

nvcc -cudart shared -arch=compute_90 my_app.cu -o my_app
vgpu run --gpu nvidia/h200-nvl ./my_app
vgpu shell -y --gpu nvidia/h200-nvl --count 4
vgpu info --gpu nvidia/h200-nvl

This profile is compute capability 9.0. A kernel built for compute_75 runs on it and on every other NVIDIA profile; a kernel built for a newer architecture is refused by a profile that is older than it. Link the CUDA runtime shared (-cudart shared) so the simulator can stand in for it.

To use it every time, put gpu: nvidia/h200-nvl in a pantheonsim.yaml settings file, or in a GitHub Actions job give the PantheonSim action gpu: nvidia/h200-nvl (see CUDA and HIP tests in GitHub Actions). The simulator checks behaviour and has no timing model, so a run on a physical GPU before a release still counts. The CUDA guide explains the rest.

Where the numbers come from

Each profile file begins with a note on its sources. This is the note for NVIDIA H200 NVL, as written in the file:

Source note for nvidia/h200-nvl

Public values; verified: false, nothing here has been measured on a card. Where each value comes from: - compute capability 9.0 and every per-SM limit: read from physical H100s (nvidia/h100, nvidia/h100-pcie); the H200 NVL is the same GH100 as the H200 (nvidia/h200 argues this at length), with the 132 SMs of the SXM part. - memory 141 GB, 4.8 TB/s, max TDP up to 600 W, FP64 30 TFLOPS, PCIe Gen5, 2- or 4-way NVLink bridge: NVIDIA's H200 page (NVL column). The page lists no SM count; the SXM part's 132 are used, and the FP64 rate (30 TFLOPS against the SXM's 34) is consistent with 132 SMs at a lower clock. This is derived, not published. NVIDIA's "GB" for graphics memory is binary (see rtx2080ti.yaml), so vram_bytes is 144384 MiB (141 GiB), an upper bound. - PCI device 0x233B, "NVIDIA H200 NVL", from the name table in NVIDIA's open GPU kernel modules (g_nv_name_released.h, main at ca2d03e, 2026-10-07).

The GPU index lists every profile, and this one is not yet measured.

Other Hopper profiles: