PantheonSim / GPUs / NVIDIA A30

Simulated NVIDIA A30: a GPU profile for CUDA and ROCm testing

PantheonSim, the GPU simulator for CUDA and ROCm, simulates the NVIDIA A30 (Ampere, 8.0) as the profile nvidia/a30. A 24 GB Ampere data-center GPU with HBM2, aimed at mainstream AI inference and training. 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/a30`
Vendor and architecture	NVIDIA, Ampere
Compute capability	`8.0`
Device memory	24.0 GiB, HBM
Streaming multiprocessors	56
Warp size	32
Max threads per multiprocessor	2048
Shared memory per block (opt-in maximum)	163 KiB
Memory bus width	3072-bit
Power limit (profile)	165 W
Maximum clocks (profile)	1410 MHz shader, 1215 MHz memory
ECC memory	yes
PCIe link reported	Gen4 x16
Tensor or matrix cores	yes
Verified (measured on a physical card)	not yet measured

What this GPU is for

A 24 GB Ampere data-center GPU with HBM2, aimed at mainstream AI inference and training.

Ampere covers two compute capabilities in these profiles: 8.0 for the GA100 data-center parts and 8.6 for the GA10x parts. The simulator runs the SASS a binary carries for the architecture and falls back to PTX when it carries none.

Among the 37 NVIDIA profiles it ranks 25th by device memory (24.0 GiB) and 32nd by SMs (56), 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_80 my_app.cu -o my_app
vgpu run --gpu nvidia/a30 ./my_app
vgpu shell -y --gpu nvidia/a30 --count 4
vgpu info --gpu nvidia/a30

This profile is compute capability 8.0. 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/a30 in a pantheonsim.yaml settings file, or in a GitHub Actions job give the PantheonSim action gpu: nvidia/a30 (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 A30, as written in the file:

Source note for nvidia/a30

Public values; verified: false, nothing here has been measured on a card. Where each value comes from: - compute capability 8.0: NVIDIA's CUDA GPUs table (developer.nvidia.com/cuda-gpus). - memory 24 GB HBM2, 933 GB/s, 165 W TDP, PCIe Gen4 (64 GB/s, which is x16): the A30 data sheet. NVIDIA's "GB" for HBM is binary (an A100 40 GB reports 40960 MiB, nvidia/a100-sxm4-40gb), so vram_bytes is 24576 MiB; a real card's nvidia-smi total is a little smaller, so this is an upper bound. - SMs 56: NVIDIA publishes no A30 SM count in the documents read here (the data sheet, the MIG user guide, which gives 1/4 of the SMs to its 1g.6gb profile, and the developer blogs). The figure is DERIVED: Lenovo's and PNY's product pages give 224 third-generation Tensor Cores, and NVIDIA's A100 whitepaper gives GA100 four Tensor Cores and 64 FP32 cores an SM, so 224 / 4 = 56 (3584 CUDA cores, the figure the retail listings quote). Unverified; the one number here that no NVIDIA page states. - Per-SM limits are compute capability 8.0's, the same as nvidia/a100, which was read from a physical card (A100 whitepaper: 64K registers, 2048 threads, 32 blocks, 164 KB shared memory an SM). - ECC: HBM2 with ECC on, as on the A100 (the data sheet does not list it). - SM clock 1410 MHz and memory clock 1215 MHz are the A100 40 GB's (NVIDIA publishes neither for the A30) until measured. - PCI device 0x20B7, "NVIDIA A30", from the name table in NVIDIA's open GPU kernel modules (g_nv_name_released.h). - The optional NVLink bridge (200 GB/s, two GPUs) is not recorded, as for nvidia/h100-pcie. - CUDA attributes (the `cuda:` section): the 3072-bit bus is a retailer figure (see the cuda section); the L2 is unpublished and left unset.

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

Other Ampere profiles: