LTX 2.3 image benchmarks

Test hardware and platform limits.

Leaf control: Windows and native Linux use Diffusers leaf offloading with use_stream=True. WSL uses use_stream=False because streamed leaf offloading was incompatible with the tested WSL environment; this is not a limitation established for every WSL setup. This setting concerns weight offloading, not video VAE decode streaming.

Distilled BF16, 1280×704, 8 steps, seed 43, real prompt, CFG 1/rescale 0.7, no PAG. Shared official leaf text encoder and resident untiled VAE in the primary comparison.

Presets resolve to the executed transformer runtimes; the VAE route is independent. Scenario names link directly to measurement details. Times are seconds; GPU is sampled absolute driver usage, RAM is process RSS, and memory units are GiB. Average is denoise wall divided by steps, including warmup, not completed warm GPU timing. Tracked total excludes command startup and downloads. Before reading the tables, see the labels and memory glossary.

In comparable overview tables, official Diffusers leaf is the first row and the reference for that table: 🟢 ↓ is lower, 🔴 ↑ is higher. Equal displayed values have no arrow. Arrows describe measurements, not an overall preset ranking; higher memory can accompany faster execution. Phase tables and separate workloads without a leaf control have no arrows. Rows with different driver/cache sessions or a reused non-side-by-side observation also have no arrows.

Windows — BF16

Phase-table rows identify the complete pipeline by its transformer route. The shared leaf encoder and resident VAE keep their own implementations; a row labeled vmm_fast does not imply native VMM execution in every phase.

Phase measurements

The encoder phase includes load/encode/cleanup; connectors are inside the generate aggregate, which also includes transformer setup and denoise. Do not add this aggregate to the denoise column above.

Text encoder phase (load/encode/cleanup)

Generate aggregate (connectors/setup/denoise)

VAE load

VAE decode

Pipeline total

Measured tracked totals and global peaks for the same complete runs above. These are not sums of the phase tables. The encoder and VAE routes remain as stated for this workload; the runtime column identifies the transformer.

Linux — BF16

The NVMe observation uses a different driver/cache session from the route trio; it is not a controlled allocator-only A/B.

Phase measurements

The encoder phase includes load/encode/cleanup; connectors are inside the generate aggregate, which also includes transformer setup and denoise. Do not add this aggregate to the denoise column above.

Text encoder phase (load/encode/cleanup)

Generate aggregate (connectors/setup/denoise)

VAE load

VAE decode

Pipeline total

Measured tracked totals and global peaks for the same complete runs above. These are not sums of the phase tables. The encoder and VAE routes remain as stated for this workload; the runtime column identifies the transformer.

WSL — BF16

Leaf controls are non-streamed; linear uses an explicit 3 GiB override and native uses bounded registration. These are not Windows/Linux parity claims.

Phase measurements

The encoder phase includes load/encode/cleanup; connectors are inside the generate aggregate, which also includes transformer setup and denoise. Do not add this aggregate to the denoise column above.

Text encoder phase (load/encode/cleanup)

Generate aggregate (connectors/setup/denoise)

VAE load

VAE decode

Pipeline total

Measured tracked totals and global peaks for the same complete runs above. These are not sums of the phase tables. The encoder and VAE routes remain as stated for this workload; the runtime column identifies the transformer.

VAE strategies — separate native-encoder workload

1280×704, 8 steps, seed 43. These pairs use a native text encoder, unlike the shared-leaf-encoder comparisons above.

Phase measurements

Completed CUDA timing — native-encoder VAE pairs only

These completed event durations are not interchangeable with host wall averages.

Windows SDNQ INT4

Saved INT4 transformer, 1280×704, 8 steps, seed 43; resident VAE. Compare within this pair, not as an isolated BF16-versus-INT4 speedup.

Phase measurements

Windows warm_process — two generations in one process

1280×704 BF16, 8 steps per generation, initial seed 43; shared leaf encoder, resident VAE. This is a repeated workload, not a single-image comparison.

Recommendations

The local rc3 installed-wheel image checks on Linux and WSL are recorded separately. They confirm exact output for their fixed platform recipes; they do not replace the benchmark rankings or establish a performance improvement.

low_ram_safe did not have the lowest global RSS in the Windows comparison. No selected row certifies a low-system-RAM machine.

Single observations, not statistical or arbitrary-model guarantees. Missing measurements are unavailable, not zero; phase peaks must not be summed. See presets, reproduction and measurement details before comparing different recipes.