Torch Hammer benchmark report

Sunspot — Sunspot XPU 1-GPU 8-benches stress — full_x1921c1s2b0n0_20261111_130031.json
source: full_x1921c1s2b0n0_20261111_130031.json GPU: Intel(R) Data Center GPU Max 1550 nodes: 1 GPUs: 1 benchmarks: 8
nodes
1
GPUs
1
benchmarks
8
outliers
0

(a) Batched GEMM (float32) [TFLOP/s]

fleet mean: 18.9 TFLOP/s
Fleet distribution of GPU mean performance (TFLOP/s)
0 0 0 0 1 GPU count 18.9–19.9 TFLOP/s: 1 GPUs 19.4 TFLOP/s mean
node GPU 0 node mean vs fleet thrt
x1921c1s2b0n01919+0.0%

(b) Convolution (float32) [img/s]

fleet mean: 1,640,940.0 img/s
Fleet distribution of GPU mean performance (img/s)
0 0 0 0 1 GPU count 1.6M–1.6M img/s: 1 GPUs 1.6M img/s mean
node GPU 0 node mean vs fleet thrt
x1921c1s2b0n01,640,9401,640,940+0.0%

(c) Einsum Attention (float32) [TFLOP/s]

fleet mean: 14.2 TFLOP/s
Fleet distribution of GPU mean performance (TFLOP/s)
0 0 0 0 1 GPU count 14.2–15.2 TFLOP/s: 1 GPUs 14.7 TFLOP/s mean
node GPU 0 node mean vs fleet thrt
x1921c1s2b0n01414+0.0%

(d) Memory Traffic (float32) [GB/s]

fleet mean: 27.7 GB/s
Fleet distribution of GPU mean performance (GB/s)
0 0 0 0 1 GPU count 27.7–28.7 GB/s: 1 GPUs 28.2 GB/s mean
node GPU 0 node mean vs fleet thrt
x1921c1s2b0n02828+0.0%

(e) Heat Equation (float32) [MLUPS]

fleet mean: 10,613.0 MLUPS
Fleet distribution of GPU mean performance (MLUPS)
0 0 0 0 1 GPU count 11k–11k MLUPS: 1 GPUs 11k MLUPS mean
node GPU 0 node mean vs fleet thrt
x1921c1s2b0n010,61310,613+0.0%

(f) Schrödinger Equation (float32) [iter/s]

fleet mean: 43.2 iter/s
Fleet distribution of GPU mean performance (iter/s)
0 0 0 0 1 GPU count 43.2–44.2 iter/s: 1 GPUs 43.7 iter/s mean
node GPU 0 node mean vs fleet thrt
x1921c1s2b0n04343+0.0%

(g) Atomic Contention (float32) [Gops/s]

fleet mean: 1.7 Gops/s
Fleet distribution of GPU mean performance (Gops/s)
0 0 0 0 1 GPU count 1.7–2.7 Gops/s: 1 GPUs 2.2 Gops/s +15% mean
node GPU 0 node mean vs fleet thrt
x1921c1s2b0n022+0.0%

(h) Sparse MM (float32) [TFLOP/s]

fleet mean: 2.0 TFLOP/s
Fleet distribution of GPU mean performance (TFLOP/s)
0 0 0 0 1 GPU count 2.0–3.0 TFLOP/s: 1 GPUs 2.5 TFLOP/s +15% mean
node GPU 0 node mean vs fleet thrt
x1921c1s2b0n022+0.0%

Outliers

None detected