Skip to content
KernelIndex
Search⌘K

VAE conv3x3 GroupNorm SiLU residual fused

50 eligible runs
rmsnorm

Complete fused residual block combining two sequential Conv3x3->GroupNorm->SiLU operations with residual addition. This is the fundamental building block of Sana's VAE encoder/decoder: input -> Conv3x3 -> GroupNorm -> SiLU -> Conv3x3 -> GroupNorm -> SiLU -> add(input). Processes feature maps with 256 channels and 32 groups for GroupNorm.

Fastest reported
291.0µsmean · 72% of SOL
jonathanc.netLicense unknown · unknown

Reported evidence · last observed 2026-07-05. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 20 cases · mean latencychannels = 256 · num_groups = 32 · kernel_size = 320 cases

Not measured on H100 for this workload. Challenges →

Source-native comparison · GPU NVIDIA B200 · Workload suite of 20 cases · mean latency · Protocol SOL-ExecBench evaluation stack v1.0 · mean · 28 results · last observed 2026-08-24Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
291.0µs72% SOL
1.00×
Reported · license unknown · no source
2026-07-05

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
2
292.8µs72% SOL
1.01×
Reported · license unknown · no source
2026-07-01

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
3
295.6µs72% SOL
1.02×
Reported · license unknown · no source
2026-06-28

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
4
312.7µs70% SOL
1.07×
Reported · license unknown · no source
2026-06-28

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
5
333.4µs68% SOL
1.15×
Reported · license unknown · no source
2026-07-03

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
6
342.3µs67% SOL
1.18×
Reported · license unknown · no source
2026-06-07

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
7
345.6µs67% SOL
1.19×
Reported · license unknown · no source
2026-07-02

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
8
366.4µs65% SOL
1.26×
Reported · license unknown · no source
2026-07-07

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
9
458.3µs58% SOL
1.58×
Reported · license unknown · no source
2026-07-05

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
10
493.3µs56% SOL
1.70×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
11
495.8µs55% SOL
1.70×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
12
518.2µs54% SOL
1.78×
Reported · license unknown · no source
2026-04-15

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
13
546.5µs52% SOL
1.88×
Reported · license unknown · no source
2026-05-20

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
14
565.4µs51% SOL
1.94×
Reported · license unknown · no source
2026-04-28

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
15
589.1µs50% SOL
2.02×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
16
634.9µs48% SOL
2.18×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
17
681.4µs46% SOL
2.34×
Reported · license unknown · no source
2026-04-19

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
18
683.5µs45% SOL
2.35×
Reported · license unknown · no source
2026-05-28

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
19
751.5µs43% SOL
2.58×
Reported · license unknown · no source
2026-05-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
20
757.2µs42% SOL
2.60×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
21
765.7µs42% SOL
2.63×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
22
869.8µs39% SOL
2.99×
Reported · license unknown · no source
2026-04-15

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
23
1.05ms35% SOL
3.61×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
24
1.16ms32% SOL
3.98×
Reported · license unknown · no source
2026-07-03

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
25
1.17ms32% SOL
4.04×
Reported · license unknown · no source
2026-07-03

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
26
1.26ms30% SOL
4.34×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
27
1.28ms30% SOL
4.40×
Reported · license unknown · no source
2026-05-28

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
28
1.28ms30% SOL
4.41×
Reported · license unknown · no source
2026-04-14

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
unknown
291.0µs
1.07×
Reported
license unknown · no source
unknown
292.8µs
1.07×
Reported
license unknown · no source
unknown
295.6µs
1.08×
Reported
license unknown · no source
unknown
307.9µs
1.13×
Reported
license unknown · no source
unknown
312.7µs
1.15×
Reported
license unknown · no source
unknown
321.0µs
1.18×
Reported
license unknown · no source
unknown
328.9µs
1.20×
Reported
license unknown · no source
unknown
342.3µs
1.25×
Reported
license unknown · no source
Show all 50 implementations ›
unknown
360.8µs
1.32×
Reported
license unknown · no source
unknown
430.5µs
1.58×
Reported
license unknown · no source
unknown
458.3µs
1.68×
Reported
license unknown · no source
unknown
493.3µs
1.81×
Reported
license unknown · no source
unknown
589.1µs
2.16×
Reported
license unknown · no source
unknown
634.9µs
2.33×
Reported
license unknown · no source
unknown
683.5µs
2.50×
Reported
license unknown · no source
unknown
751.5µs
2.75×
Reported
license unknown · no source
unknown
757.2µs
2.77×
Reported
license unknown · no source
unknown
765.7µs
2.80×
Reported
license unknown · no source
unknown
869.8µs
3.19×
Reported
license unknown · no source
unknown
1.16ms
4.24×
Reported
license unknown · no source
unknown
1.28ms
4.69×
Reported
license unknown · no source
unknown
299.7µs
1.10×
Reported
license unknown · no source
unknown
303.5µs
1.11×
Reported
license unknown · no source
unknown
310.1µs
1.14×
Reported
license unknown · no source
unknown
311.8µs
1.14×
Reported
license unknown · no source
unknown
324.8µs
1.19×
Reported
license unknown · no source
unknown
325.3µs
1.19×
Reported
license unknown · no source
unknown
333.6µs
1.22×
Reported
license unknown · no source
unknown
333.4µs
1.22×
Reported
license unknown · no source
unknown
345.6µs
1.27×
Reported
license unknown · no source
unknown
357.2µs
1.31×
Reported
license unknown · no source
unknown
357.5µs
1.31×
Reported
license unknown · no source
unknown
364.5µs
1.33×
Reported
license unknown · no source
unknown
366.4µs
1.34×
Reported
license unknown · no source
unknown
379.0µs
1.39×
Reported
license unknown · no source
unknown
388.4µs
1.42×
Reported
license unknown · no source
unknown
390.9µs
1.43×
Reported
license unknown · no source
unknown
495.8µs
1.82×
Reported
license unknown · no source
unknown
518.2µs
1.90×
Reported
license unknown · no source
unknown
546.5µs
2.00×
Reported
license unknown · no source
unknown
565.4µs
2.07×
Reported
license unknown · no source
unknown
681.4µs
2.50×
Reported
license unknown · no source
unknown
793.1µs
2.90×
Reported
license unknown · no source
unknown
1.05ms
3.85×
Reported
license unknown · no source
unknown
1.17ms
4.30×
Reported
license unknown · no source
unknown
1.26ms
4.63×
Reported
license unknown · no source
unknown
1.28ms
4.70×
Reported
license unknown · no source
unknown
273.1µs
1.00×
Reported
license unknown · no source
unknown
284.3µs
1.04×
Reported
license unknown · no source
unknown
300.9µs
1.10×
Reported
license unknown · no source

Semantics

Inputs and outputs
xfp32 [batch_size, channels, height, width]
conv1_weightfp32 [channels, channels, kernel_size, kernel_size]
norm1_weightfp32 [channels]
norm1_biasfp32 [channels]
conv2_weightfp32 [channels, channels, kernel_size, kernel_size]
norm2_weightfp32 [channels]
norm2_biasfp32 [channels]
epsfp32 scalar
outputfp32 [batch_size, channels, height, width]
Axes and behavior
widthvariable
heightvariable
channelsconstant = 256
batch_sizevariable
num_groupsconstant = 32
kernel_sizeconstant = 3
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias002_vae_conv3x3_groupnorm_silu_residual_fusedmodelsana-sprint-1-6b-1024pxsha2569a26dc844f8c…
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24How records are decidedJSON