← Back to Compatibility Matrix

PRIMARY LOCAL TESTBENCH DOSSIER

NVIDIA GeForce RTX 5080 (16GB) Evidence

Authoritative local hardware verification results. 37 workflows have current successful evidence on the RTX 5080. Historical runs remain unclassified because STANDARD versus OPTIMIZED settings were not recorded. See each workflow for its complete execution history.

Test Environment Specifications

GPU & Memory
NVIDIA RTX 5080 (16GB GDDR7)
Architecture / Generation
Blackwell (sm_120) · RTX 50-Series
Host System RAM
61.64 GiB reported usable RAM (historical ComfyUI observation)
Software Stack
PyTorch 2.11 (Nightly) + CUDA 12.8
ComfyUI Version
ComfyUI v0.33.0 (historical win32 record)
Execution Coverage
37 / 50 workflows with current success

Hardware Limitations & Scope Caveats

  • We do NOT claim the RTX 5080 runs 100% of workflows: 13 catalog entries have not been executed. 5 are blocked by frontend subgraphs requiring canvas expansion, and 8 remain on hold due to model weight footprints (>15–30 GB).
  • No Lower-VRAM Guarantees: Success on this 16GB setup does not prove that an 8GB or 12GB GPU will succeed. Lower VRAM configurations must be evaluated conservatively.
  • PyTorch / CUDA Requirements: Running on Blackwell requires modern PyTorch builds with CUDA 12.8+ support (`cudaMallocAsync`).

Verified Execution Records by Model Family (37 Workflows)

Records below retain their saved execution artifacts. Workflow hashes match this build; semantic output review is not implied by execution success. Click any entry to inspect settings and download JSON.

SD 1.5 (12)

EXECUTION TESTED
IDWorkflow TitleResolution / FramesSteps / SamplerRuntimeOutput
15AnimateDiff Walking Character512×512 (8f)12 steps (euler)8.03sView Artifact ↗
16AnimateDiff Looping Particle Preset512×512 (8f)12 steps (euler)6.04sView Artifact ↗
17AnimateDiff Landscape Timelapse512×512 (8f)12 steps (euler)6.04sView Artifact ↗
18AnimateDiff Product Rotation512×512 (8f)12 steps (euler)6.03sView Artifact ↗
19AnimateDiff Slow Zoom512×512 (8f)12 steps (euler)6.03sView Artifact ↗
24SD 1.5 Style Transfer512×512— 4.49sView Artifact ↗
27SD 1.5 Object Removal (Inpaint)512×51220 steps (dpmpp_2m)6.03sView Artifact ↗
33ControlNet OpenPose — SD 1.5512×51225 steps (dpmpp_2m)6.03sView Artifact ↗
34ControlNet Lineart — SD 1.5512×51225 steps (dpmpp_2m)2.30sView Artifact ↗
35ControlNet Tile — SD 1.5512×51220 steps (dpmpp_2m)2.09sView Artifact ↗
42SD 1.5 Pixel Art Generator1024×102420 steps (euler)4.01sView Artifact ↗
04Stable Diffusion 1.5 Default512×51220 steps (dpmpp_2m)6.26sView Artifact ↗

Real-ESRGAN (2)

EXECUTION TESTED
IDWorkflow TitleResolution / FramesSteps / SamplerRuntimeOutput
214x ESRGAN Upscale2048×2048— 2.78sView Artifact ↗
224x Model Upscale — Real-ESRGANDefault20 steps (euler)2.00sView Artifact ↗

SDXL (17)

EXECUTION TESTED
IDWorkflow TitleResolution / FramesSteps / SamplerRuntimeOutput
23SDXL Image-to-Image1024×102420 steps (dpmpp_2m)5.85sView Artifact ↗
25SDXL Sketch to Image1024×102420 steps (euler)8.07sView Artifact ↗
26SDXL Inpainting1024×102420 steps (euler)4.03sView Artifact ↗
28SDXL Product Photography1024×102420 steps (dpmpp_2m)9.37sView Artifact ↗
29SDXL Architectural Visualization1024×102420 steps (euler)4.02sView Artifact ↗
31ControlNet Canny — SDXL1024×102425 steps (dpmpp_2m)18.10sView Artifact ↗
32ControlNet Depth — SDXL1024×102425 steps (dpmpp_2m)10.04sView Artifact ↗
36SDXL Batch ×4 Variations1024×1024— 17.78sView Artifact ↗
37SDXL Batch ×8 Variations1024×102420 steps (euler)14.05sView Artifact ↗
38SDXL Logo & Brand Design1024×102420 steps (euler)4.01sView Artifact ↗
39SDXL Concept Art Sheet1024×102420 steps (euler)4.03sView Artifact ↗
41SDXL Interior Design1024×1024— 4.43sView Artifact ↗
45SDXL Sci-Fi Environment1024×1024— 5.30sView Artifact ↗
46SDXL Abstract Art Generator1024×102420 steps (euler)10.05sView Artifact ↗
47SDXL Surrealist Dreamscapes1024×102420 steps (euler)4.03sView Artifact ↗
48SDXL Fantasy Creatures1024×102420 steps (euler)4.00sView Artifact ↗
49SDXL Cyberpunk City1024×1024— 8.03sView Artifact ↗

Flux (5)

EXECUTION TESTED
IDWorkflow TitleResolution / FramesSteps / SamplerRuntimeOutput
30FLUX Portrait v2 — Editorial512×51220 steps (euler)16.19sView Artifact ↗
40FLUX Realistic Person512×51220 steps (euler)16.06sView Artifact ↗
44FLUX Food Photography512×51220 steps (euler)16.07sView Artifact ↗
50FLUX Cinematic Portrait v2512×51220 steps (euler)18.06sView Artifact ↗
01Flux.1 Dev — Text to Image512×51220 steps (euler)10.05sView Artifact ↗

ACE Step (1)

EXECUTION TESTED
IDWorkflow TitleResolution / FramesSteps / SamplerRuntimeOutput
43ACE Step 1.5 — Text to SongDefault8 steps (euler)214.70sView Artifact ↗

Remaining 13 Unexecuted Workflows

These workflows have not completed execution on our testbench. Factual reasons are detailed below to distinguish testbed availability constraints from GPU compatibility limitations.

IDWorkflowClassificationFactual Reason
10AuraFlow Text to ImageMODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).
11LTX Video 2.3 — Basic T2VSUBGRAPH BLOCKERWorkflow template contains ComfyUI frontend composite subgraphs requiring visual canvas expansion. Headless execution blocked.
12LTX 2.3 — Portrait VideoSUBGRAPH BLOCKERWorkflow template contains ComfyUI frontend composite subgraphs requiring visual canvas expansion. Headless execution blocked.
13LTX Video — Action ChaseSUBGRAPH BLOCKERWorkflow template contains ComfyUI frontend composite subgraphs requiring visual canvas expansion. Headless execution blocked.
14LTX 2.3 — Short Video DraftSUBGRAPH BLOCKERWorkflow template contains ComfyUI frontend composite subgraphs requiring visual canvas expansion. Headless execution blocked.
20Wan 2.1 Video — 480p T2VMODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).
02Flux.1 Schnell — 4-Step FastMODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).
03SDXL Base + RefinerMODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).
05HiDream — 17B Text to ImageMODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).
06Hunyuan Image 2.1MODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).
07Lumina Image 2.0MODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).
08Qwen Image Edit (Multimodal)MODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).
09Chroma — Stylized No-NegativeMODEL HOLDModel weights download deferred under Phase 5/6 Large Model Hold policy (>15–30 GB footprint).