PantheonSim / GPUs / AMD Instinct MI455X
Simulated AMD Instinct MI455X: a GPU profile for CUDA and ROCm testing
PantheonSim, the GPU simulator for CUDA and ROCm, simulates the AMD Instinct MI455X (CDNA 5, gfx1250) as the profile amd/mi455x.
The CDNA 5 Instinct MI455X (gfx1250) with 432 GiB of HBM4, which is wave32 only. No physical card exists to check the profile against, and AMD has not published all of its figures. 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 rocm-smi and amd-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 | amd/mi455x |
| Vendor and architecture | AMD, CDNA 5 |
| GPU target | gfx1250 |
| Device memory | 432.0 GiB, HBM |
| Multiprocessors (as HIP reports them) | 256 |
| Wavefront size | 32 |
| Max threads per multiprocessor | 2048 |
| Shared memory per block (opt-in maximum) | 320 KiB |
| L2 cache | 4 MiB |
| Power limit (profile) | 1800 W |
| Maximum clocks (profile) | 2400 MHz shader, 3200 MHz memory |
| ECC memory | yes |
| PCIe link reported | Gen6 x16 |
| Tensor or matrix cores | yes |
| Verified (measured on a physical card) | not yet measured |
What this GPU is for
The CDNA 5 Instinct MI455X (gfx1250) with 432 GiB of HBM4, which is wave32 only. No physical card exists to check the profile against, and AMD has not published all of its figures.
CDNA 5 is the architecture of the Instinct MI455X, built for gfx1250. It is wave32 only, and no physical card exists to check the profile against.
Among the 10 AMD profiles it ranks 1st by device memory (432.0 GiB) and 3rd by multiprocessors as HIP counts them (256), 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. The profile file itself calls some of its values placeholders. 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:
hipcc --offload-arch=gfx1250 my_app.hip -o my_app vgpu run --gpu amd/mi455x ./my_app vgpu shell -y --gpu amd/mi455x --count 4 --rocm 7.1.0 vgpu info --gpu amd/mi455x
Build for this card's own gfx target, gfx1250. A program built for another target is refused by name, as the real runtime refuses it, and hipErrorNoBinaryForGpu means exactly this.
To use it every time, put gpu: amd/mi455x in a pantheonsim.yaml settings file, or in a GitHub Actions job give the PantheonSim action gpu: amd/mi455x (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 ROCm and HIP guide explains the rest.
Where the numbers come from
Each profile file begins with a note on its sources. This is the note for AMD Instinct MI455X, as written in the file:
Source note for amd/mi455x
Values from public AMD announcements and LLVM. verified: false means nothing here has been confirmed against a physical card, none exists to check against, and AMD has not published the rest (compute-unit count in HIP's terms, clocks, board power, PCI ids). Those are marked "placeholder" below and chosen, not read from anywhere. What is sourced: 432 GB of HBM4 and 19.6 TB/s (AMD, Helios announcement); 256 workgroup processors, 32 to each of 8 compute dies; wave32 only; 192 MB of last-level cache. The 320 KB of LDS a work-group is LLVM's limit for gfx1250 (llc -mcpu=gfx1250 refuses a kernel that asks for more).
The GPU index lists every profile, and this one is not yet measured.
Related GPUs
Other AMD profiles: