OpenVINO
Image generation on CPU and iGPU of Intel Core i7 1355U (power limits 35/55W) with 96GB DDR5 RAM using SD.Next + openvino
| CPU | GPU | |
|---|---|---|
juggernautXL [33e58e8668] 512x512 |
Prompt: VibrantlySharp style, 3pic_vist4, animeniji, in a soft-focus painterly concept art style Upper body profile view, autumn dryad elf adorned in translucent red foliage and bark-textured silk, wind animating her hair and dress into motion, background: ancient twilight forest, distant amber sunlight diffused through fog and drifting leaves. Parameters: Steps: 20| Size: 512x512| Seed: 3638080379| CFG scale: 6| Model: juggernautXL_juggXIByRundiffusion| Model hash: 33e58e8668| App: SD.Next| Version: 72eb013| Operations: txt2img| Pipeline: StableDiffusionXLPipeline Time: 4m 50.93s | pipeline 268.66 decode 19.12 move 2.62 prompt 2.61 prepare 0.52 preview 0.51 | RAM 38.08 GB 41% |
Prompt: VibrantlySharp style, 3pic_vist4, animeniji, in a soft-focus painterly concept art style Upper body profile view, autumn dryad elf adorned in translucent red foliage and bark-textured silk, wind animating her hair and dress into motion, background: ancient twilight forest, distant amber sunlight diffused through fog and drifting leaves. Parameters: Steps: 20| Size: 512x512| Seed: 3638080379| CFG scale: 6| Model: juggernautXL_juggXIByRundiffusion| Model hash: 33e58e8668| App: SD.Next| Version: 72eb013| Operations: txt2img| Pipeline: StableDiffusionXLPipeline Time: 1m 55.95s | pipeline 97.78 decode 15.76 move 2.39 prompt 2.38 | RAM 21.07 GB 22% |
1024x1024 | Prompt: VibrantlySharp style, 3pic_vist4, animeniji, in a soft-focus painterly concept art style Upper body profile view, autumn dryad elf adorned in translucent red foliage and bark-textured silk, wind animating her hair and dress into motion, background: ancient twilight forest, distant amber sunlight diffused through fog and drifting leaves. Parameters: Steps: 20| Size: 1024x1024| Seed: 2850092752| CFG scale: 6| Model: juggernautXL_juggXIByRundiffusion| Model hash: 33e58e8668| App: SD.Next| Version: 72eb013| Operations: txt2img| Pipeline: StableDiffusionXLPipeline Time: 19m 16.57s | pipeline 1079.19 decode 77.33 preview 0.94 | RAM 59.97 GB 64% |
Prompt: VibrantlySharp style, 3pic_vist4, animeniji, in a soft-focus painterly concept art style Upper body profile view, autumn dryad elf adorned in translucent red foliage and bark-textured silk, wind animating her hair and dress into motion, background: ancient twilight forest, distant amber sunlight diffused through fog and drifting leaves. Parameters: Steps: 20| Size: 1024x1024| Seed: 2850092752| CFG scale: 6| Model: juggernautXL_juggXIByRundiffusion| Model hash: 33e58e8668| App: SD.Next| Version: 72eb013| Operations: txt2img| Pipeline: StableDiffusionXLPipeline Time: 5m 0.59s | pipeline 252.54 decode 48.01 | RAM 21.81 GB 23% |
2048x2048 96GB RAM + 16GB SWAP | after 2 hours and 48 minutes of image generation (Step 20/20 1h:49m:18s) [12486.386048] Out of memory: Killed process 1753 (python) total-vm:135643124kB, anon-rss:97618012kB, file-rss:3008kB, shmem-rss:0kB, UID:1000 pgtables:240656kB oom_score_adj:0 | infinite running [16098.406688] INFO: task python3:17735 blocked for more than 122 seconds. |
juggernaut_reborn [338b85bc4f] 1024x1024 |
Prompt: Featured here is the Asian swallowtail, beautifully captured on a red spider lily. Found across Northeast Asia and even parts of Hawaii, this striking butterfly is known for its vivid colouration and powerful flight, making it a frequent visitor to both urban gardens and wild habitats. While it is common and not currently threatened, its role as a pollinator is no less important. The Asian swallowtail uses sophisticated colour vision and colour constancy to identify flowers, especially those in the citrus family, even under changing light conditions. This remarkable ability allows it to remain an effective pollinator across a wide range of environments. Delicate yet determined, the Asian swallowtail reminds us that even the smallest pollinators play a big role in keeping nature balanced and beautiful. Parameters: Steps: 20| Size: 1024x1024| Seed: 215801161| CFG scale: 6| Model: juggernaut_reborn| Model hash: 338b85bc4f| App: SD.Next| Version: 72eb013| Operations: txt2img| Pipeline: StableDiffusionPipeline Time: 17m 1.30s | pipeline 945.21 decode 75.50 preview 1.60 move 0.55 prompt 0.54 | RAM 19.36 GB 21% | RuntimeError: Exception from src/inference/src/cpp/core.cpp:109: Exception from src/inference/src/dev/plugin.cpp:53: Check '!exceed_allocatable_mem_size' failed at src/plugins/intel_gpu/src/runtime/ocl/ocl_engine.cpp:142: [GPU] Exceeded max size of memory object allocation: requested 8589934592 bytes, but max alloc size supported by device is 4294959104 bytes.Please try to reduce batch size or use lower precision. |
768x768 |
Prompt: Featured here is the Asian swallowtail, beautifully captured on a red spider lily. Found across Northeast Asia and even parts of Hawaii, this striking butterfly is known for its vivid colouration and powerful flight, making it a frequent visitor to both urban gardens and wild habitats. While it is common and not currently threatened, its role as a pollinator is no less important. The Asian swallowtail uses sophisticated colour vision and colour constancy to identify flowers, especially those in the citrus family, even under changing light conditions. This remarkable ability allows it to remain an effective pollinator across a wide range of environments. Delicate yet determined, the Asian swallowtail reminds us that even the smallest pollinators play a big role in keeping nature balanced and beautiful. Parameters: Steps: 20| Size: 768x768| Seed: 3003228961| CFG scale: 6| Model: juggernaut_reborn| Model hash: 338b85bc4f| App: SD.Next| Version: 72eb013| Operations: txt2img| Pipeline: StableDiffusionPipeline Time: 7m 5.52s | pipeline 384.68 decode 40.36 preview 1.26 move 0.46 prompt 0.46 | RAM 14.74 GB 16% |
Prompt: Featured here is the Asian swallowtail, beautifully captured on a red spider lily. Found across Northeast Asia and even parts of Hawaii, this striking butterfly is known for its vivid colouration and powerful flight, making it a frequent visitor to both urban gardens and wild habitats. While it is common and not currently threatened, its role as a pollinator is no less important. The Asian swallowtail uses sophisticated colour vision and colour constancy to identify flowers, especially those in the citrus family, even under changing light conditions. This remarkable ability allows it to remain an effective pollinator across a wide range of environments. Delicate yet determined, the Asian swallowtail reminds us that even the smallest pollinators play a big role in keeping nature balanced and beautiful. Parameters: Steps: 20| Size: 768x768| Seed: 3003228961| CFG scale: 6| Model: juggernaut_reborn| Model hash: 338b85bc4f| App: SD.Next| Version: 72eb013| Operations: txt2img| Pipeline: StableDiffusionPipeline Time: 2m 10.11s | pipeline 102.45 decode 27.63 preview 0.48 | RAM 5.93 GB 6% |
stable-cascade 1024x1024 | list index out of range | list index out of range |
512x512 | list index out of range | list index out of range |
shuttleai/shuttle-3.1-aesthetic 512x512 | Shape inference input shapes {[1],[4]} Requested output shape [1,24,4608,128] is incompatible with input shape Summary: -- No conversion rule found for operations: prim::ListConstruct -- Conversion is failed for: aten.view.default | Shape inference input shapes {[1],[4]} Requested output shape [1,24,4608,128] is incompatible with input shape Summary: -- No conversion rule found for operations: prim::ListConstruct -- Conversion is failed for: aten.view.default |
| System info | app: sdnext updated: 2025-06-16 hash: 72eb0132 url: https://github.com/vladmandic/sdnext/tree/master arch: x86_64 cpu: x86_64 system: Linux release: 6.11.0-26-generic python: 3.12.3 device: 13th Gen Intel(R) Core(TM) i7-1355U openvino: 2025.1.0 ram: free:84.49 used:9.48 total:93.97 active: cpu dtype: torch.float32 vae: torch.float32 xformers: diffusers: 0.34.0.dev0 transformers: 4.52.4 base: juggernaut_reborn [338b85bc4f] refiner: None vae: Automatic te: Default unet: Default | app: sdnext updated: 2025-06-16 hash: 72eb0132 url: https://github.com/vladmandic/sdnext/tree/master arch: x86_64 cpu: x86_64 system: Linux release: 6.11.0-26-generic python: 3.12.3 device: Intel(R) Iris(R) Xe Graphics (iGPU) openvino: 2025.1.0 ram: free:89.06 used:4.91 total:93.97 gpu: total:93.97 active: cpu dtype: torch.float32 vae: torch.float32 xformers: diffusers: 0.34.0.dev0 transformers: 4.52.4 base: juggernaut_reborn [338b85bc4f] refiner: None vae: Automatic te: Default unet: Default |
Intel Xe more than 4GB allocation error
openvino [GPU] Exceeded max size of memory object allocation: requested 8589934592 bytes, but max alloc size supported by device is 4294959104 bytes.
error occurs because the GPU memory allocator in your 13th Gen Intel iGPU (Intel Xe Graphics) cannot handle the large memory request made by Stable Diffusion (SD.Next using OpenVINO). Specifically:
Requested: 8,589,934,592 bytes (~8GB)
Max allowed per allocation: 4,294,959,104 bytes (~4GB)
This is a hardware limitation of integrated Intel GPUs — they typically have a maximum buffer allocation size around 4GB, even if more shared memory is available overall.
🔧 How to Fix It
Here are ways to address this:
✅ 1. Reduce Batch Size
In the SD.Next interface:
Find the Batch size setting (not to be confused with "Batch count")
Set Batch size = 1 (this will reduce the single memory block size)
This is usually the most effective and easiest fix.
✅ 2. Use Lower Precision (FP16)
OpenVINO can support FP16 (half precision), which drastically reduces memory usage.
If you are using a model in FP32, try:
Converting your model to FP16
Or downloading an FP16 version of the model (if available)
Some UIs (like SD.Next, ComfyUI, or WebUI) allow selecting FP16 models directly or during import.
✅ 3. Use a Smaller Model
Stable Diffusion 1.5 requires less VRAM than SD 2.1 or SDXL.
Try:
Switching from SDXL → SD 1.5
Or using Lite versions of models if available (like "v1.5-pruned-emaonly" or "tiny" models)
✅ 4. Use CPU Instead (as fallback)
If the GPU is not sufficient, you can:
Select
OpenVINO [CPU]backendIt will be slower but more compatible
Use this as a last resort if GPU execution isn't viable.
⚠️ Notes
This limitation is not something that can be bypassed by modifying system settings — it's enforced by the Intel OpenVINO GPU runtime and the GPU hardware.
SDXL and models with high-resolution pipelines or large attention layers are more likely to trigger this problem.
IPEX
app: sdnext updated: 2025-06-19 hash: 7fc7797a url: https://github.com/vladmandic/sdnext/tree/dev arch: x86_64 cpu: x86_64 system: Linux release: 6.11.0-26-generic python: 3.12.3 2.7.1+xpu 2.7.1+xpu ram: free:85.83 used:8.14 total:93.97 xformers: diffusers: 0.34.0.dev0 transformers: 4.52.4 active: xpu dtype: torch.bfloat16 vae: torch.bfloat16 unet: torch.bfloat16 base: model.safetensors refiner: None vae: Automatic te: Default unet: Default
