Skip to content
KernelIndex
Search⌘K
Sign in

GEMM n34816 k5120

129 eligible runs
gemm

General matrix multiply (GEMM) C = A @ B.T. Captured from Qwen3 14B gate_up_proj (combined gate+up, intermediate=17408, N=17408*2=34816, hidden=5120).

Source baseline · unbeaten
72.4µsmean
flashinfer / wrapper8028beFlashInfer-Bench baselines · Apache-2.0 · python

Reported evidence · last observed 2026-03-23. The source's designated baseline implementation. Reported by source; not independently reproduced.

No package · vendor the source →View source →Run detail →

Current records

Workloadm = 4 · fp1643 cases
m
dtype
1fp162fp164fp167fp168fp1615fp1616fp1624fp1632fp1635fp1640fp1648fp1656fp1664fp1670fp1672fp1680fp1688fp1696fp16104fp16112fp16120fp16128fp16136fp16144fp16152fp16160fp16168fp16176fp16184fp16192fp16200fp16208fp16216fp16224fp16232fp16240fp16248fp16256fp16972fp162053fp162379fp168192fp16
scrolls · 43 cases total

Not measured on H100 for this workload. Challenges →

Source-native comparison · GPU NVIDIA B200 · Workload m = 4 · fp16 · CUDA 12.8 · Framework pytorch 2.9.1+cu128 · Protocol flashinfer-bench · mean · 3 results · last observed 2026-03-24Record history →
Estimated floor 44.6 µs · record 1.62× above itestimate, not evidence ›
DRAM 44.6 µs · compute 634 ns · bandwidth-bound on B200
every declared tensor crosses HBM exactly once (8,000 GB/s, B200 datasheet)
2·M·N·K with M=4, N=34816, K=5120 at the dense bf16 peak (2,250 TFLOP/s)
headroom-v1: a lower bound from declared tensors and datasheet peaks. A kernel can sit well above it for good reasons.
#
Implementation
Latency
vs #1
Trust
Observed
1
flashinfer / wrapper8028bebaselineFlashInfer-Bench baselines
72.4µs
1.00×
Reported · Apache-2.0 · source
2026-03-23stale

Measured exactly what you asked. The source's designated baseline implementation. Reported by source; not independently reproduced.

source mirroredApache-2.0no install recipeView source →Run detail →
2
flashinfer / wrapper8028bebaselineFlashInfer-Bench baselines
72.5µs
1.00×
Reported · Apache-2.0 · source
2026-03-24stale

Measured exactly what you asked. The source's designated baseline implementation. Reported by source; not independently reproduced.

source mirroredApache-2.0no install recipeView source →Run detail →
3
flashinfer / wrapper8028bebaselineFlashInfer-Bench baselines
72.8µs
1.01×
Reported · Apache-2.0 · source
2026-03-24stale

Measured exactly what you asked. The source's designated baseline implementation. Reported by source; not independently reproduced.

source mirroredApache-2.0no install recipeView source →Run detail →

Scaling by m

100.0 µs1.00 ms14248825620538km →72.7 µs73.5 µs72.4 µs73.1 µs72.8 µs73.7 µs72.7 µs73.7 µs73.0 µs74.8 µs74.1 µs73.9 µs74.8 µs74.9 µs74.8 µs75.7 µs74.9 µs75.5 µs75.1 µs75.8 µs75.5 µs76.5 µs76.1 µs77.1 µs76.7 µs77.7 µs76.9 µs76.5 µs76.7 µs77.5 µs77.4 µs78.5 µs78.6 µs79.6 µs79.0 µs80.2 µs80.4 µs95.2 µs94.8 µs227.3 µs494.6 µs582.1 µs1.86 msflashinfer / wrapper 8028be
flashinfer / wrapper 8028bebest per workload · NVIDIA B200 · CUDA 12.8 · pytorch 2.9.1+cu128 · flashinfer-bench held constant · log scale

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
flashinfer / wrapper8028beFlashInfer-Bench baselines
python
72.4µs
1.00×
Reported
Apache-2.0 · source

Semantics

Inputs and outputs
afp16 [m, k]
bfp16 [n, k]
cfp16 [m, n]
Axes and behavior
kconstant = 5120
mvariable
nconstant = 34816
determinismunspecified
constraintsNo mutation or aliasing
Identity
aliasgemm_n34816_k5120modelqwen3-14bsha256724ce684b1f0…
Sources: FlashInfer-Bench (2026-03-24) · Apache-2.0last observed 2026-03-24How records are decidedJSON