github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #13338] RTX 5090 on Windows – GPU sometimes detected, sometimes reported as “0 B VRAM”, causing CPU fallback + dolphin3 crash (graph_reserve: failed to allocate compute buffers) #70869

New Issue
Closed
opened 2026-05-04 23:17:04 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-05-04 23:17:04 -05:00
Owner
Copy Link

Originally created by @jason-witter on GitHub (Dec 5, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/13338

What is the issue?

RTX 5090 on Windows – GPU intermittently detected, sometimes reported as “0 B VRAM”, causing CPU fallback + dolphin3 crash

Environment

  • OS: Windows 10 (build 19045)
  • Ollama version: 0.13.1
  • GPU: NVIDIA GeForce RTX 5090 (Blackwell), 31.8 GiB VRAM
  • NVIDIA Driver: 591.44
  • CUDA: using bundled Ollama runtime (no standalone CUDA toolkit installed)
  • Hardware: Single-GPU desktop

Expected Behavior

  • Ollama consistently detects and uses the RTX 5090 GPU.

  • Server log contains a CUDA entry similar to:
    inference compute id=GPU-xxxx library=CUDA name="NVIDIA GeForce RTX 5090"

  • Models (including dolphin3) run on GPU without crashing.

  • Ollama does not intermittently drop into CPU-only “low VRAM mode”.

Actual Behavior

GPU detection succeeds sometimes and fails other times, and dolphin3 crashes when it does attempt GPU execution.

Sometimes GPU is detected correctly:

inference compute id=GPU-7345c94c-464c-a8ab-fa48-9229002dda06 filter_id="" library=CUDA compute=12.0 name="NVIDIA GeForce RTX 5090" libdirs=ollama,cuda_v13 driver=13.0 pci_id=0000:0b:00.0 type=discrete total="31.8 GiB" available="30.8 GiB"

Other times, Ollama reports 0 B of VRAM and falls back to CPU:

discovering available GPUs...
inference compute id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="63.9 GiB" available="53.7 GiB"
entering low vram mode "total vram"="0 B" threshold="20.0 GiB"

At this point, all models run on CPU only, even if OLLAMA_LLM_LIBRARY=cuda is set.

dolphin3 crash when GPU is detected:

llama_kv_cache: size = 16384.00 MiB (131072 cells, 32 layers, 1/1 seqs), K (f16): 8192.00 MiB, V (f16): 8192.00 MiB
graph_reserve: failed to allocate compute buffers
Exception 0xc0000005 ... signal arrived during external code execution
llama runner process has terminated: exit status 2

After the crash:

  • GPU detection frequently stops working entirely.
  • Server log begins reporting total vram 0 B.
  • All models run exclusively on CPU.

Steps to Reproduce

  1. Install Ollama 0.13.1 on Windows 10.
  2. Pull dolphin3:
    ollama pull dolphin3
  3. Run dolphin3 repeatedly:
    ollama run dolphin3
  4. Observe the following pattern:
    • Sometimes the GPU is detected and dolphin3 attempts to load on GPU.
    • dolphin3 then crashes with graph_reserve / access violation.
    • After the crash, GPU detection fails and server log reports total vram 0 B.
    • Restarting Ollama does not always restore GPU detection.

Notes

  • Changing OLLAMA_* environment variables (flash attention, cuda graphs, force slow, etc.) does not influence GPU detection — the issue occurs even with all GPU-related env vars removed.
  • Other models (e.g., mistral) also run on CPU once Ollama enters the 0 B VRAM state.
  • GPU works normally in other CUDA applications.
  • Issue appears consistent with incomplete Blackwell (sm_120) support, based on other user reports.

Related Issues

  • #10402 – Official RTX 5090 Support
  • #13163 – RTX 5070 Ti (Blackwell) falling back to CPU
  • #12116 – Models not loading into VRAM on RTX 5090
  • #12895 – GPU discovery inconsistency on Windows
  • #13083 – Low GPU utilization on 5090 (Windows)

Summary

Ollama 0.13.1 on Windows intermittently fails to detect the RTX 5090 GPU. When dolphin3 attempts GPU execution, the runner sometimes crashes with a buffer allocation error. After the crash, Ollama frequently reports total vram 0 B and switches into CPU-only mode. This behavior appears related to early Blackwell GPU support on Windows.

I am happy to provide additional logs or run diagnostic builds if helpful.

Relevant log output

time=2025-12-04T23:29:48.783-08:00 level=INFO source=routes.go:1544 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:262144 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY:cuda OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:C:\\Users\\Jason\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES:]"
time=2025-12-04T23:29:48.785-08:00 level=INFO source=images.go:522 msg="total blobs: 6"
time=2025-12-04T23:29:48.785-08:00 level=INFO source=images.go:529 msg="total unused blobs removed: 0"
time=2025-12-04T23:29:48.786-08:00 level=INFO source=routes.go:1597 msg="Listening on 127.0.0.1:11434 (version 0.13.1)"
time=2025-12-04T23:29:48.787-08:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2025-12-04T23:29:48.805-08:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="63.9 GiB" available="53.7 GiB"
time=2025-12-04T23:29:48.805-08:00 level=INFO source=routes.go:1638 msg="entering low vram mode" "total vram"="0 B" threshold="20.0 GiB"
[GIN] 2025/12/04 - 23:29:49 | 200 |            0s |       127.0.0.1 | HEAD     "/"
[GIN] 2025/12/04 - 23:29:49 | 200 |      43.619ms |       127.0.0.1 | POST     "/api/show"
[GIN] 2025/12/04 - 23:29:49 | 200 |     39.9348ms |       127.0.0.1 | POST     "/api/show"
time=2025-12-04T23:29:49.340-08:00 level=INFO source=cpu_windows.go:148 msg=packages count=1
time=2025-12-04T23:29:49.340-08:00 level=INFO source=cpu_windows.go:195 msg="" package=0 cores=16 efficiency=0 threads=32
llama_model_loader: loaded meta data with 77 key-value pairs and 292 tensors from C:\Users\Jason\.ollama\models\blobs\sha256-1eee6953530837b2b17d61a4e6f71a5aa31c9714cfcf3cb141aa5c1972b5116b (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = llama
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                               general.name str              = Dolphin 3.0 Llama 3.1 8B
llama_model_loader: - kv   3:                       general.organization str              = Cognitivecomputations
llama_model_loader: - kv   4:                           general.basename str              = Dolphin-3.0-Llama-3.1
llama_model_loader: - kv   5:                         general.size_label str              = 8B
llama_model_loader: - kv   6:                            general.license str              = llama3.1
llama_model_loader: - kv   7:                   general.base_model.count u32              = 1
llama_model_loader: - kv   8:                  general.base_model.0.name str              = Llama 3.1 8B
llama_model_loader: - kv   9:          general.base_model.0.organization str              = Meta Llama
llama_model_loader: - kv  10:              general.base_model.0.repo_url str              = https://huggingface.co/meta-llama/Lla...
llama_model_loader: - kv  11:                      general.dataset.count u32              = 13
llama_model_loader: - kv  12:                     general.dataset.0.name str              = Opc Sft Stage1
llama_model_loader: - kv  13:             general.dataset.0.organization str              = OpenCoder LLM
llama_model_loader: - kv  14:                 general.dataset.0.repo_url str              = https://huggingface.co/OpenCoder-LLM/...
llama_model_loader: - kv  15:                     general.dataset.1.name str              = Opc Sft Stage2
llama_model_loader: - kv  16:             general.dataset.1.organization str              = OpenCoder LLM
llama_model_loader: - kv  17:                 general.dataset.1.repo_url str              = https://huggingface.co/OpenCoder-LLM/...
llama_model_loader: - kv  18:                     general.dataset.2.name str              = Orca Agentinstruct 1M v1
llama_model_loader: - kv  19:                  general.dataset.2.version str              = v1
llama_model_loader: - kv  20:             general.dataset.2.organization str              = Microsoft
llama_model_loader: - kv  21:                 general.dataset.2.repo_url str              = https://huggingface.co/microsoft/orca...
llama_model_loader: - kv  22:                     general.dataset.3.name str              = Orca Math Word Problems 200k
llama_model_loader: - kv  23:             general.dataset.3.organization str              = Microsoft
llama_model_loader: - kv  24:                 general.dataset.3.repo_url str              = https://huggingface.co/microsoft/orca...
llama_model_loader: - kv  25:                     general.dataset.4.name str              = Hermes Function Calling v1
llama_model_loader: - kv  26:                  general.dataset.4.version str              = v1
llama_model_loader: - kv  27:             general.dataset.4.organization str              = NousResearch
llama_model_loader: - kv  28:                 general.dataset.4.repo_url str              = https://huggingface.co/NousResearch/h...
llama_model_loader: - kv  29:                     general.dataset.5.name str              = NuminaMath CoT
llama_model_loader: - kv  30:             general.dataset.5.organization str              = AI MO
llama_model_loader: - kv  31:                 general.dataset.5.repo_url str              = https://huggingface.co/AI-MO/NuminaMa...
llama_model_loader: - kv  32:                     general.dataset.6.name str              = NuminaMath TIR
llama_model_loader: - kv  33:             general.dataset.6.organization str              = AI MO
llama_model_loader: - kv  34:                 general.dataset.6.repo_url str              = https://huggingface.co/AI-MO/NuminaMa...
llama_model_loader: - kv  35:                     general.dataset.7.name str              = Tulu 3 Sft Mixture
llama_model_loader: - kv  36:             general.dataset.7.organization str              = Allenai
llama_model_loader: - kv  37:                 general.dataset.7.repo_url str              = https://huggingface.co/allenai/tulu-3...
llama_model_loader: - kv  38:                     general.dataset.8.name str              = Dolphin Coder
llama_model_loader: - kv  39:             general.dataset.8.organization str              = Cognitivecomputations
llama_model_loader: - kv  40:                 general.dataset.8.repo_url str              = https://huggingface.co/cognitivecompu...
llama_model_loader: - kv  41:                     general.dataset.9.name str              = Smoltalk
llama_model_loader: - kv  42:             general.dataset.9.organization str              = HuggingFaceTB
llama_model_loader: - kv  43:                 general.dataset.9.repo_url str              = https://huggingface.co/HuggingFaceTB/...
llama_model_loader: - kv  44:                    general.dataset.10.name str              = Samantha Data
llama_model_loader: - kv  45:            general.dataset.10.organization str              = Cognitivecomputations
llama_model_loader: - kv  46:                general.dataset.10.repo_url str              = https://huggingface.co/cognitivecompu...
llama_model_loader: - kv  47:                    general.dataset.11.name str              = CodeFeedback Filtered Instruction
llama_model_loader: - kv  48:            general.dataset.11.organization str              = M A P
llama_model_loader: - kv  49:                general.dataset.11.repo_url str              = https://huggingface.co/m-a-p/CodeFeed...
llama_model_loader: - kv  50:                    general.dataset.12.name str              = Code Feedback
llama_model_loader: - kv  51:            general.dataset.12.organization str              = M A P
llama_model_loader: - kv  52:                general.dataset.12.repo_url str              = https://huggingface.co/m-a-p/Code-Fee...
llama_model_loader: - kv  53:                          general.languages arr[str,1]       = ["en"]
llama_model_loader: - kv  54:                          llama.block_count u32              = 32
llama_model_loader: - kv  55:                       llama.context_length u32              = 131072
llama_model_loader: - kv  56:                     llama.embedding_length u32              = 4096
llama_model_loader: - kv  57:                  llama.feed_forward_length u32              = 14336
llama_model_loader: - kv  58:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv  59:              llama.attention.head_count_kv u32              = 8
llama_model_loader: - kv  60:                       llama.rope.freq_base f32              = 500000.000000
llama_model_loader: - kv  61:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  62:                 llama.attention.key_length u32              = 128
llama_model_loader: - kv  63:               llama.attention.value_length u32              = 128
llama_model_loader: - kv  64:                          general.file_type u32              = 15
llama_model_loader: - kv  65:                           llama.vocab_size u32              = 128258
llama_model_loader: - kv  66:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv  67:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  68:                         tokenizer.ggml.pre str              = llama-bpe
llama_model_loader: - kv  69:                      tokenizer.ggml.tokens arr[str,128258]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  70:                  tokenizer.ggml.token_type arr[i32,128258]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  71:                      tokenizer.ggml.merges arr[str,280147]  = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "...
llama_model_loader: - kv  72:                tokenizer.ggml.bos_token_id u32              = 128000
llama_model_loader: - kv  73:                tokenizer.ggml.eos_token_id u32              = 128256
llama_model_loader: - kv  74:            tokenizer.ggml.padding_token_id u32              = 128001
llama_model_loader: - kv  75:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
llama_model_loader: - kv  76:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   66 tensors
llama_model_loader: - type q4_K:  193 tensors
llama_model_loader: - type q6_K:   33 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = Q4_K - Medium
print_info: file size   = 4.58 GiB (4.89 BPW) 
load: printing all EOG tokens:
load:   - 128001 ('<|end_of_text|>')
load:   - 128008 ('<|eom_id|>')
load:   - 128009 ('<|eot_id|>')
load:   - 128256 ('<|im_end|>')
load: special tokens cache size = 258
load: token to piece cache size = 0.7999 MB
print_info: arch             = llama
print_info: vocab_only       = 1
print_info: model type       = ?B
print_info: model params     = 8.03 B
print_info: general.name     = Dolphin 3.0 Llama 3.1 8B
print_info: vocab type       = BPE
print_info: n_vocab          = 128258
print_info: n_merges         = 280147
print_info: BOS token        = 128000 '<|begin_of_text|>'
print_info: EOS token        = 128256 '<|im_end|>'
print_info: EOT token        = 128256 '<|im_end|>'
print_info: EOM token        = 128008 '<|eom_id|>'
print_info: PAD token        = 128001 '<|end_of_text|>'
print_info: LF token         = 198 'Ċ'
print_info: EOG token        = 128001 '<|end_of_text|>'
print_info: EOG token        = 128008 '<|eom_id|>'
print_info: EOG token        = 128009 '<|eot_id|>'
print_info: EOG token        = 128256 '<|im_end|>'
print_info: max token length = 256
llama_model_load: vocab only - skipping tensors
time=2025-12-04T23:29:49.553-08:00 level=WARN source=server.go:167 msg="requested context size too large for model" num_ctx=262144 n_ctx_train=131072
time=2025-12-04T23:29:49.559-08:00 level=INFO source=server.go:392 msg="starting runner" cmd="C:\\Users\\Jason\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --model C:\\Users\\Jason\\.ollama\\models\\blobs\\sha256-1eee6953530837b2b17d61a4e6f71a5aa31c9714cfcf3cb141aa5c1972b5116b --port 52014"
time=2025-12-04T23:29:49.561-08:00 level=INFO source=sched.go:443 msg="system memory" total="63.9 GiB" free="53.7 GiB" free_swap="55.5 GiB"
time=2025-12-04T23:29:49.561-08:00 level=INFO source=server.go:459 msg="loading model" "model layers"=33 requested=-1
time=2025-12-04T23:29:49.561-08:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="4.3 GiB"
time=2025-12-04T23:29:49.561-08:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="16.0 GiB"
time=2025-12-04T23:29:49.561-08:00 level=INFO source=device.go:272 msg="total memory" size="20.3 GiB"
time=2025-12-04T23:29:49.593-08:00 level=INFO source=runner.go:963 msg="starting go runner"
load_backend: loaded CPU backend from C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\ggml-cpu-haswell.dll
time=2025-12-04T23:29:49.606-08:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(clang)
time=2025-12-04T23:29:49.607-08:00 level=INFO source=runner.go:999 msg="Server listening on 127.0.0.1:52014"
time=2025-12-04T23:29:49.615-08:00 level=INFO source=runner.go:893 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:false KvSize:131072 KvCacheType: NumThreads:16 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}"
time=2025-12-04T23:29:49.615-08:00 level=INFO source=server.go:1294 msg="waiting for llama runner to start responding"
time=2025-12-04T23:29:49.615-08:00 level=INFO source=server.go:1328 msg="waiting for server to become available" status="llm server loading model"
llama_model_loader: loaded meta data with 77 key-value pairs and 292 tensors from C:\Users\Jason\.ollama\models\blobs\sha256-1eee6953530837b2b17d61a4e6f71a5aa31c9714cfcf3cb141aa5c1972b5116b (version GGUF V3 (latest))
llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output.
llama_model_loader: - kv   0:                       general.architecture str              = llama
llama_model_loader: - kv   1:                               general.type str              = model
llama_model_loader: - kv   2:                               general.name str              = Dolphin 3.0 Llama 3.1 8B
llama_model_loader: - kv   3:                       general.organization str              = Cognitivecomputations
llama_model_loader: - kv   4:                           general.basename str              = Dolphin-3.0-Llama-3.1
llama_model_loader: - kv   5:                         general.size_label str              = 8B
llama_model_loader: - kv   6:                            general.license str              = llama3.1
llama_model_loader: - kv   7:                   general.base_model.count u32              = 1
llama_model_loader: - kv   8:                  general.base_model.0.name str              = Llama 3.1 8B
llama_model_loader: - kv   9:          general.base_model.0.organization str              = Meta Llama
llama_model_loader: - kv  10:              general.base_model.0.repo_url str              = https://huggingface.co/meta-llama/Lla...
llama_model_loader: - kv  11:                      general.dataset.count u32              = 13
llama_model_loader: - kv  12:                     general.dataset.0.name str              = Opc Sft Stage1
llama_model_loader: - kv  13:             general.dataset.0.organization str              = OpenCoder LLM
llama_model_loader: - kv  14:                 general.dataset.0.repo_url str              = https://huggingface.co/OpenCoder-LLM/...
llama_model_loader: - kv  15:                     general.dataset.1.name str              = Opc Sft Stage2
llama_model_loader: - kv  16:             general.dataset.1.organization str              = OpenCoder LLM
llama_model_loader: - kv  17:                 general.dataset.1.repo_url str              = https://huggingface.co/OpenCoder-LLM/...
llama_model_loader: - kv  18:                     general.dataset.2.name str              = Orca Agentinstruct 1M v1
llama_model_loader: - kv  19:                  general.dataset.2.version str              = v1
llama_model_loader: - kv  20:             general.dataset.2.organization str              = Microsoft
llama_model_loader: - kv  21:                 general.dataset.2.repo_url str              = https://huggingface.co/microsoft/orca...
llama_model_loader: - kv  22:                     general.dataset.3.name str              = Orca Math Word Problems 200k
llama_model_loader: - kv  23:             general.dataset.3.organization str              = Microsoft
llama_model_loader: - kv  24:                 general.dataset.3.repo_url str              = https://huggingface.co/microsoft/orca...
llama_model_loader: - kv  25:                     general.dataset.4.name str              = Hermes Function Calling v1
llama_model_loader: - kv  26:                  general.dataset.4.version str              = v1
llama_model_loader: - kv  27:             general.dataset.4.organization str              = NousResearch
llama_model_loader: - kv  28:                 general.dataset.4.repo_url str              = https://huggingface.co/NousResearch/h...
llama_model_loader: - kv  29:                     general.dataset.5.name str              = NuminaMath CoT
llama_model_loader: - kv  30:             general.dataset.5.organization str              = AI MO
llama_model_loader: - kv  31:                 general.dataset.5.repo_url str              = https://huggingface.co/AI-MO/NuminaMa...
llama_model_loader: - kv  32:                     general.dataset.6.name str              = NuminaMath TIR
llama_model_loader: - kv  33:             general.dataset.6.organization str              = AI MO
llama_model_loader: - kv  34:                 general.dataset.6.repo_url str              = https://huggingface.co/AI-MO/NuminaMa...
llama_model_loader: - kv  35:                     general.dataset.7.name str              = Tulu 3 Sft Mixture
llama_model_loader: - kv  36:             general.dataset.7.organization str              = Allenai
llama_model_loader: - kv  37:                 general.dataset.7.repo_url str              = https://huggingface.co/allenai/tulu-3...
llama_model_loader: - kv  38:                     general.dataset.8.name str              = Dolphin Coder
llama_model_loader: - kv  39:             general.dataset.8.organization str              = Cognitivecomputations
llama_model_loader: - kv  40:                 general.dataset.8.repo_url str              = https://huggingface.co/cognitivecompu...
llama_model_loader: - kv  41:                     general.dataset.9.name str              = Smoltalk
llama_model_loader: - kv  42:             general.dataset.9.organization str              = HuggingFaceTB
llama_model_loader: - kv  43:                 general.dataset.9.repo_url str              = https://huggingface.co/HuggingFaceTB/...
llama_model_loader: - kv  44:                    general.dataset.10.name str              = Samantha Data
llama_model_loader: - kv  45:            general.dataset.10.organization str              = Cognitivecomputations
llama_model_loader: - kv  46:                general.dataset.10.repo_url str              = https://huggingface.co/cognitivecompu...
llama_model_loader: - kv  47:                    general.dataset.11.name str              = CodeFeedback Filtered Instruction
llama_model_loader: - kv  48:            general.dataset.11.organization str              = M A P
llama_model_loader: - kv  49:                general.dataset.11.repo_url str              = https://huggingface.co/m-a-p/CodeFeed...
llama_model_loader: - kv  50:                    general.dataset.12.name str              = Code Feedback
llama_model_loader: - kv  51:            general.dataset.12.organization str              = M A P
llama_model_loader: - kv  52:                general.dataset.12.repo_url str              = https://huggingface.co/m-a-p/Code-Fee...
llama_model_loader: - kv  53:                          general.languages arr[str,1]       = ["en"]
llama_model_loader: - kv  54:                          llama.block_count u32              = 32
llama_model_loader: - kv  55:                       llama.context_length u32              = 131072
llama_model_loader: - kv  56:                     llama.embedding_length u32              = 4096
llama_model_loader: - kv  57:                  llama.feed_forward_length u32              = 14336
llama_model_loader: - kv  58:                 llama.attention.head_count u32              = 32
llama_model_loader: - kv  59:              llama.attention.head_count_kv u32              = 8
llama_model_loader: - kv  60:                       llama.rope.freq_base f32              = 500000.000000
llama_model_loader: - kv  61:     llama.attention.layer_norm_rms_epsilon f32              = 0.000010
llama_model_loader: - kv  62:                 llama.attention.key_length u32              = 128
llama_model_loader: - kv  63:               llama.attention.value_length u32              = 128
llama_model_loader: - kv  64:                          general.file_type u32              = 15
llama_model_loader: - kv  65:                           llama.vocab_size u32              = 128258
llama_model_loader: - kv  66:                 llama.rope.dimension_count u32              = 128
llama_model_loader: - kv  67:                       tokenizer.ggml.model str              = gpt2
llama_model_loader: - kv  68:                         tokenizer.ggml.pre str              = llama-bpe
llama_model_loader: - kv  69:                      tokenizer.ggml.tokens arr[str,128258]  = ["!", "\"", "#", "$", "%", "&", "'", ...
llama_model_loader: - kv  70:                  tokenizer.ggml.token_type arr[i32,128258]  = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ...
llama_model_loader: - kv  71:                      tokenizer.ggml.merges arr[str,280147]  = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "...
llama_model_loader: - kv  72:                tokenizer.ggml.bos_token_id u32              = 128000
llama_model_loader: - kv  73:                tokenizer.ggml.eos_token_id u32              = 128256
llama_model_loader: - kv  74:            tokenizer.ggml.padding_token_id u32              = 128001
llama_model_loader: - kv  75:                    tokenizer.chat_template str              = {% if not add_generation_prompt is de...
llama_model_loader: - kv  76:               general.quantization_version u32              = 2
llama_model_loader: - type  f32:   66 tensors
llama_model_loader: - type q4_K:  193 tensors
llama_model_loader: - type q6_K:   33 tensors
print_info: file format = GGUF V3 (latest)
print_info: file type   = Q4_K - Medium
print_info: file size   = 4.58 GiB (4.89 BPW) 
load: printing all EOG tokens:
load:   - 128001 ('<|end_of_text|>')
load:   - 128008 ('<|eom_id|>')
load:   - 128009 ('<|eot_id|>')
load:   - 128256 ('<|im_end|>')
load: special tokens cache size = 258
load: token to piece cache size = 0.7999 MB
print_info: arch             = llama
print_info: vocab_only       = 0
print_info: n_ctx_train      = 131072
print_info: n_embd           = 4096
print_info: n_layer          = 32
print_info: n_head           = 32
print_info: n_head_kv        = 8
print_info: n_rot            = 128
print_info: n_swa            = 0
print_info: is_swa_any       = 0
print_info: n_embd_head_k    = 128
print_info: n_embd_head_v    = 128
print_info: n_gqa            = 4
print_info: n_embd_k_gqa     = 1024
print_info: n_embd_v_gqa     = 1024
print_info: f_norm_eps       = 0.0e+00
print_info: f_norm_rms_eps   = 1.0e-05
print_info: f_clamp_kqv      = 0.0e+00
print_info: f_max_alibi_bias = 0.0e+00
print_info: f_logit_scale    = 0.0e+00
print_info: f_attn_scale     = 0.0e+00
print_info: n_ff             = 14336
print_info: n_expert         = 0
print_info: n_expert_used    = 0
print_info: causal attn      = 1
print_info: pooling type     = 0
print_info: rope type        = 0
print_info: rope scaling     = linear
print_info: freq_base_train  = 500000.0
print_info: freq_scale_train = 1
print_info: n_ctx_orig_yarn  = 131072
print_info: rope_finetuned   = unknown
print_info: model type       = 8B
print_info: model params     = 8.03 B
print_info: general.name     = Dolphin 3.0 Llama 3.1 8B
print_info: vocab type       = BPE
print_info: n_vocab          = 128258
print_info: n_merges         = 280147
print_info: BOS token        = 128000 '<|begin_of_text|>'
print_info: EOS token        = 128256 '<|im_end|>'
print_info: EOT token        = 128256 '<|im_end|>'
print_info: EOM token        = 128008 '<|eom_id|>'
print_info: PAD token        = 128001 '<|end_of_text|>'
print_info: LF token         = 198 'Ċ'
print_info: EOG token        = 128001 '<|end_of_text|>'
print_info: EOG token        = 128008 '<|eom_id|>'
print_info: EOG token        = 128009 '<|eot_id|>'
print_info: EOG token        = 128256 '<|im_end|>'
print_info: max token length = 256
load_tensors: loading model tensors, this can take a while... (mmap = false)
load_tensors:          CPU model buffer size =  4685.32 MiB
llama_context: constructing llama_context
llama_context: n_seq_max     = 1
llama_context: n_ctx         = 131072
llama_context: n_ctx_per_seq = 131072
llama_context: n_batch       = 512
llama_context: n_ubatch      = 512
llama_context: causal_attn   = 1
llama_context: flash_attn    = disabled
llama_context: kv_unified    = false
llama_context: freq_base     = 500000.0
llama_context: freq_scale    = 1
llama_context:        CPU  output buffer size =     0.50 MiB
llama_kv_cache:        CPU KV buffer size = 16384.00 MiB
llama_kv_cache: size = 16384.00 MiB (131072 cells,  32 layers,  1/1 seqs), K (f16): 8192.00 MiB, V (f16): 8192.00 MiB
llama_context:        CPU compute buffer size =  8484.01 MiB
llama_context: graph nodes  = 1158
llama_context: graph splits = 1
time=2025-12-04T23:29:52.367-08:00 level=INFO source=server.go:1332 msg="llama runner started in 2.81 seconds"
time=2025-12-04T23:29:52.367-08:00 level=INFO source=sched.go:517 msg="loaded runners" count=1
time=2025-12-04T23:29:52.367-08:00 level=INFO source=server.go:1294 msg="waiting for llama runner to start responding"
time=2025-12-04T23:29:52.367-08:00 level=INFO source=server.go:1332 msg="llama runner started in 2.81 seconds"
[GIN] 2025/12/04 - 23:29:52 | 200 |    3.1097964s |       127.0.0.1 | POST     "/api/generate"

OS

Windows

GPU

Nvidia

CPU

AMD

Ollama version

0.13.1

Originally created by @jason-witter on GitHub (Dec 5, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/13338 ### What is the issue? # RTX 5090 on Windows – GPU intermittently detected, sometimes reported as “0 B VRAM”, causing CPU fallback + dolphin3 crash ## Environment - OS: Windows 10 (build 19045) - Ollama version: 0.13.1 - GPU: NVIDIA GeForce RTX 5090 (Blackwell), 31.8 GiB VRAM - NVIDIA Driver: 591.44 - CUDA: using bundled Ollama runtime (no standalone CUDA toolkit installed) - Hardware: Single-GPU desktop --- ## Expected Behavior - Ollama consistently detects and uses the RTX 5090 GPU. - Server log contains a CUDA entry similar to: inference compute id=GPU-xxxx library=CUDA name="NVIDIA GeForce RTX 5090" - Models (including dolphin3) run on GPU without crashing. - Ollama does not intermittently drop into CPU-only “low VRAM mode”. --- ## Actual Behavior GPU detection succeeds sometimes and fails other times, and dolphin3 crashes when it does attempt GPU execution. ### Sometimes GPU is detected correctly: inference compute id=GPU-7345c94c-464c-a8ab-fa48-9229002dda06 filter_id="" library=CUDA compute=12.0 name="NVIDIA GeForce RTX 5090" libdirs=ollama,cuda_v13 driver=13.0 pci_id=0000:0b:00.0 type=discrete total="31.8 GiB" available="30.8 GiB" ### Other times, Ollama reports 0 B of VRAM and falls back to CPU: discovering available GPUs... inference compute id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="63.9 GiB" available="53.7 GiB" entering low vram mode "total vram"="0 B" threshold="20.0 GiB" At this point, all models run on CPU only, even if OLLAMA_LLM_LIBRARY=cuda is set. ### dolphin3 crash when GPU *is* detected: llama_kv_cache: size = 16384.00 MiB (131072 cells, 32 layers, 1/1 seqs), K (f16): 8192.00 MiB, V (f16): 8192.00 MiB graph_reserve: failed to allocate compute buffers Exception 0xc0000005 ... signal arrived during external code execution llama runner process has terminated: exit status 2 After the crash: - GPU detection frequently stops working entirely. - Server log begins reporting total vram 0 B. - All models run exclusively on CPU. --- ## Steps to Reproduce 1. Install Ollama 0.13.1 on Windows 10. 2. Pull dolphin3: ollama pull dolphin3 3. Run dolphin3 repeatedly: ollama run dolphin3 4. Observe the following pattern: - Sometimes the GPU is detected and dolphin3 attempts to load on GPU. - dolphin3 then crashes with graph_reserve / access violation. - After the crash, GPU detection fails and server log reports total vram 0 B. - Restarting Ollama does not always restore GPU detection. --- ## Notes - Changing OLLAMA_* environment variables (flash attention, cuda graphs, force slow, etc.) does not influence GPU detection — the issue occurs even with all GPU-related env vars removed. - Other models (e.g., mistral) also run on CPU once Ollama enters the 0 B VRAM state. - GPU works normally in other CUDA applications. - Issue appears consistent with incomplete Blackwell (sm_120) support, based on other user reports. --- ## Related Issues - #10402 – Official RTX 5090 Support - #13163 – RTX 5070 Ti (Blackwell) falling back to CPU - #12116 – Models not loading into VRAM on RTX 5090 - #12895 – GPU discovery inconsistency on Windows - #13083 – Low GPU utilization on 5090 (Windows) --- ## Summary Ollama 0.13.1 on Windows intermittently fails to detect the RTX 5090 GPU. When dolphin3 attempts GPU execution, the runner sometimes crashes with a buffer allocation error. After the crash, Ollama frequently reports total vram 0 B and switches into CPU-only mode. This behavior appears related to early Blackwell GPU support on Windows. I am happy to provide additional logs or run diagnostic builds if helpful. ### Relevant log output ```shell time=2025-12-04T23:29:48.783-08:00 level=INFO source=routes.go:1544 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:262144 OLLAMA_DEBUG:INFO OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY:cuda OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:C:\\Users\\Jason\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES:]" time=2025-12-04T23:29:48.785-08:00 level=INFO source=images.go:522 msg="total blobs: 6" time=2025-12-04T23:29:48.785-08:00 level=INFO source=images.go:529 msg="total unused blobs removed: 0" time=2025-12-04T23:29:48.786-08:00 level=INFO source=routes.go:1597 msg="Listening on 127.0.0.1:11434 (version 0.13.1)" time=2025-12-04T23:29:48.787-08:00 level=INFO source=runner.go:67 msg="discovering available GPUs..." time=2025-12-04T23:29:48.805-08:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="63.9 GiB" available="53.7 GiB" time=2025-12-04T23:29:48.805-08:00 level=INFO source=routes.go:1638 msg="entering low vram mode" "total vram"="0 B" threshold="20.0 GiB" [GIN] 2025/12/04 - 23:29:49 | 200 | 0s | 127.0.0.1 | HEAD "/" [GIN] 2025/12/04 - 23:29:49 | 200 | 43.619ms | 127.0.0.1 | POST "/api/show" [GIN] 2025/12/04 - 23:29:49 | 200 | 39.9348ms | 127.0.0.1 | POST "/api/show" time=2025-12-04T23:29:49.340-08:00 level=INFO source=cpu_windows.go:148 msg=packages count=1 time=2025-12-04T23:29:49.340-08:00 level=INFO source=cpu_windows.go:195 msg="" package=0 cores=16 efficiency=0 threads=32 llama_model_loader: loaded meta data with 77 key-value pairs and 292 tensors from C:\Users\Jason\.ollama\models\blobs\sha256-1eee6953530837b2b17d61a4e6f71a5aa31c9714cfcf3cb141aa5c1972b5116b (version GGUF V3 (latest)) llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. llama_model_loader: - kv 0: general.architecture str = llama llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.name str = Dolphin 3.0 Llama 3.1 8B llama_model_loader: - kv 3: general.organization str = Cognitivecomputations llama_model_loader: - kv 4: general.basename str = Dolphin-3.0-Llama-3.1 llama_model_loader: - kv 5: general.size_label str = 8B llama_model_loader: - kv 6: general.license str = llama3.1 llama_model_loader: - kv 7: general.base_model.count u32 = 1 llama_model_loader: - kv 8: general.base_model.0.name str = Llama 3.1 8B llama_model_loader: - kv 9: general.base_model.0.organization str = Meta Llama llama_model_loader: - kv 10: general.base_model.0.repo_url str = https://huggingface.co/meta-llama/Lla... llama_model_loader: - kv 11: general.dataset.count u32 = 13 llama_model_loader: - kv 12: general.dataset.0.name str = Opc Sft Stage1 llama_model_loader: - kv 13: general.dataset.0.organization str = OpenCoder LLM llama_model_loader: - kv 14: general.dataset.0.repo_url str = https://huggingface.co/OpenCoder-LLM/... llama_model_loader: - kv 15: general.dataset.1.name str = Opc Sft Stage2 llama_model_loader: - kv 16: general.dataset.1.organization str = OpenCoder LLM llama_model_loader: - kv 17: general.dataset.1.repo_url str = https://huggingface.co/OpenCoder-LLM/... llama_model_loader: - kv 18: general.dataset.2.name str = Orca Agentinstruct 1M v1 llama_model_loader: - kv 19: general.dataset.2.version str = v1 llama_model_loader: - kv 20: general.dataset.2.organization str = Microsoft llama_model_loader: - kv 21: general.dataset.2.repo_url str = https://huggingface.co/microsoft/orca... llama_model_loader: - kv 22: general.dataset.3.name str = Orca Math Word Problems 200k llama_model_loader: - kv 23: general.dataset.3.organization str = Microsoft llama_model_loader: - kv 24: general.dataset.3.repo_url str = https://huggingface.co/microsoft/orca... llama_model_loader: - kv 25: general.dataset.4.name str = Hermes Function Calling v1 llama_model_loader: - kv 26: general.dataset.4.version str = v1 llama_model_loader: - kv 27: general.dataset.4.organization str = NousResearch llama_model_loader: - kv 28: general.dataset.4.repo_url str = https://huggingface.co/NousResearch/h... llama_model_loader: - kv 29: general.dataset.5.name str = NuminaMath CoT llama_model_loader: - kv 30: general.dataset.5.organization str = AI MO llama_model_loader: - kv 31: general.dataset.5.repo_url str = https://huggingface.co/AI-MO/NuminaMa... llama_model_loader: - kv 32: general.dataset.6.name str = NuminaMath TIR llama_model_loader: - kv 33: general.dataset.6.organization str = AI MO llama_model_loader: - kv 34: general.dataset.6.repo_url str = https://huggingface.co/AI-MO/NuminaMa... llama_model_loader: - kv 35: general.dataset.7.name str = Tulu 3 Sft Mixture llama_model_loader: - kv 36: general.dataset.7.organization str = Allenai llama_model_loader: - kv 37: general.dataset.7.repo_url str = https://huggingface.co/allenai/tulu-3... llama_model_loader: - kv 38: general.dataset.8.name str = Dolphin Coder llama_model_loader: - kv 39: general.dataset.8.organization str = Cognitivecomputations llama_model_loader: - kv 40: general.dataset.8.repo_url str = https://huggingface.co/cognitivecompu... llama_model_loader: - kv 41: general.dataset.9.name str = Smoltalk llama_model_loader: - kv 42: general.dataset.9.organization str = HuggingFaceTB llama_model_loader: - kv 43: general.dataset.9.repo_url str = https://huggingface.co/HuggingFaceTB/... llama_model_loader: - kv 44: general.dataset.10.name str = Samantha Data llama_model_loader: - kv 45: general.dataset.10.organization str = Cognitivecomputations llama_model_loader: - kv 46: general.dataset.10.repo_url str = https://huggingface.co/cognitivecompu... llama_model_loader: - kv 47: general.dataset.11.name str = CodeFeedback Filtered Instruction llama_model_loader: - kv 48: general.dataset.11.organization str = M A P llama_model_loader: - kv 49: general.dataset.11.repo_url str = https://huggingface.co/m-a-p/CodeFeed... llama_model_loader: - kv 50: general.dataset.12.name str = Code Feedback llama_model_loader: - kv 51: general.dataset.12.organization str = M A P llama_model_loader: - kv 52: general.dataset.12.repo_url str = https://huggingface.co/m-a-p/Code-Fee... llama_model_loader: - kv 53: general.languages arr[str,1] = ["en"] llama_model_loader: - kv 54: llama.block_count u32 = 32 llama_model_loader: - kv 55: llama.context_length u32 = 131072 llama_model_loader: - kv 56: llama.embedding_length u32 = 4096 llama_model_loader: - kv 57: llama.feed_forward_length u32 = 14336 llama_model_loader: - kv 58: llama.attention.head_count u32 = 32 llama_model_loader: - kv 59: llama.attention.head_count_kv u32 = 8 llama_model_loader: - kv 60: llama.rope.freq_base f32 = 500000.000000 llama_model_loader: - kv 61: llama.attention.layer_norm_rms_epsilon f32 = 0.000010 llama_model_loader: - kv 62: llama.attention.key_length u32 = 128 llama_model_loader: - kv 63: llama.attention.value_length u32 = 128 llama_model_loader: - kv 64: general.file_type u32 = 15 llama_model_loader: - kv 65: llama.vocab_size u32 = 128258 llama_model_loader: - kv 66: llama.rope.dimension_count u32 = 128 llama_model_loader: - kv 67: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 68: tokenizer.ggml.pre str = llama-bpe llama_model_loader: - kv 69: tokenizer.ggml.tokens arr[str,128258] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 70: tokenizer.ggml.token_type arr[i32,128258] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 71: tokenizer.ggml.merges arr[str,280147] = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "... llama_model_loader: - kv 72: tokenizer.ggml.bos_token_id u32 = 128000 llama_model_loader: - kv 73: tokenizer.ggml.eos_token_id u32 = 128256 llama_model_loader: - kv 74: tokenizer.ggml.padding_token_id u32 = 128001 llama_model_loader: - kv 75: tokenizer.chat_template str = {% if not add_generation_prompt is de... llama_model_loader: - kv 76: general.quantization_version u32 = 2 llama_model_loader: - type f32: 66 tensors llama_model_loader: - type q4_K: 193 tensors llama_model_loader: - type q6_K: 33 tensors print_info: file format = GGUF V3 (latest) print_info: file type = Q4_K - Medium print_info: file size = 4.58 GiB (4.89 BPW) load: printing all EOG tokens: load: - 128001 ('<|end_of_text|>') load: - 128008 ('<|eom_id|>') load: - 128009 ('<|eot_id|>') load: - 128256 ('<|im_end|>') load: special tokens cache size = 258 load: token to piece cache size = 0.7999 MB print_info: arch = llama print_info: vocab_only = 1 print_info: model type = ?B print_info: model params = 8.03 B print_info: general.name = Dolphin 3.0 Llama 3.1 8B print_info: vocab type = BPE print_info: n_vocab = 128258 print_info: n_merges = 280147 print_info: BOS token = 128000 '<|begin_of_text|>' print_info: EOS token = 128256 '<|im_end|>' print_info: EOT token = 128256 '<|im_end|>' print_info: EOM token = 128008 '<|eom_id|>' print_info: PAD token = 128001 '<|end_of_text|>' print_info: LF token = 198 'Ċ' print_info: EOG token = 128001 '<|end_of_text|>' print_info: EOG token = 128008 '<|eom_id|>' print_info: EOG token = 128009 '<|eot_id|>' print_info: EOG token = 128256 '<|im_end|>' print_info: max token length = 256 llama_model_load: vocab only - skipping tensors time=2025-12-04T23:29:49.553-08:00 level=WARN source=server.go:167 msg="requested context size too large for model" num_ctx=262144 n_ctx_train=131072 time=2025-12-04T23:29:49.559-08:00 level=INFO source=server.go:392 msg="starting runner" cmd="C:\\Users\\Jason\\AppData\\Local\\Programs\\Ollama\\ollama.exe runner --model C:\\Users\\Jason\\.ollama\\models\\blobs\\sha256-1eee6953530837b2b17d61a4e6f71a5aa31c9714cfcf3cb141aa5c1972b5116b --port 52014" time=2025-12-04T23:29:49.561-08:00 level=INFO source=sched.go:443 msg="system memory" total="63.9 GiB" free="53.7 GiB" free_swap="55.5 GiB" time=2025-12-04T23:29:49.561-08:00 level=INFO source=server.go:459 msg="loading model" "model layers"=33 requested=-1 time=2025-12-04T23:29:49.561-08:00 level=INFO source=device.go:245 msg="model weights" device=CPU size="4.3 GiB" time=2025-12-04T23:29:49.561-08:00 level=INFO source=device.go:256 msg="kv cache" device=CPU size="16.0 GiB" time=2025-12-04T23:29:49.561-08:00 level=INFO source=device.go:272 msg="total memory" size="20.3 GiB" time=2025-12-04T23:29:49.593-08:00 level=INFO source=runner.go:963 msg="starting go runner" load_backend: loaded CPU backend from C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\ggml-cpu-haswell.dll time=2025-12-04T23:29:49.606-08:00 level=INFO source=ggml.go:104 msg=system CPU.0.SSE3=1 CPU.0.SSSE3=1 CPU.0.AVX=1 CPU.0.AVX2=1 CPU.0.F16C=1 CPU.0.FMA=1 CPU.0.BMI2=1 CPU.0.LLAMAFILE=1 CPU.1.LLAMAFILE=1 compiler=cgo(clang) time=2025-12-04T23:29:49.607-08:00 level=INFO source=runner.go:999 msg="Server listening on 127.0.0.1:52014" time=2025-12-04T23:29:49.615-08:00 level=INFO source=runner.go:893 msg=load request="{Operation:commit LoraPath:[] Parallel:1 BatchSize:512 FlashAttention:false KvSize:131072 KvCacheType: NumThreads:16 GPULayers:[] MultiUserCache:false ProjectorPath: MainGPU:0 UseMmap:false}" time=2025-12-04T23:29:49.615-08:00 level=INFO source=server.go:1294 msg="waiting for llama runner to start responding" time=2025-12-04T23:29:49.615-08:00 level=INFO source=server.go:1328 msg="waiting for server to become available" status="llm server loading model" llama_model_loader: loaded meta data with 77 key-value pairs and 292 tensors from C:\Users\Jason\.ollama\models\blobs\sha256-1eee6953530837b2b17d61a4e6f71a5aa31c9714cfcf3cb141aa5c1972b5116b (version GGUF V3 (latest)) llama_model_loader: Dumping metadata keys/values. Note: KV overrides do not apply in this output. llama_model_loader: - kv 0: general.architecture str = llama llama_model_loader: - kv 1: general.type str = model llama_model_loader: - kv 2: general.name str = Dolphin 3.0 Llama 3.1 8B llama_model_loader: - kv 3: general.organization str = Cognitivecomputations llama_model_loader: - kv 4: general.basename str = Dolphin-3.0-Llama-3.1 llama_model_loader: - kv 5: general.size_label str = 8B llama_model_loader: - kv 6: general.license str = llama3.1 llama_model_loader: - kv 7: general.base_model.count u32 = 1 llama_model_loader: - kv 8: general.base_model.0.name str = Llama 3.1 8B llama_model_loader: - kv 9: general.base_model.0.organization str = Meta Llama llama_model_loader: - kv 10: general.base_model.0.repo_url str = https://huggingface.co/meta-llama/Lla... llama_model_loader: - kv 11: general.dataset.count u32 = 13 llama_model_loader: - kv 12: general.dataset.0.name str = Opc Sft Stage1 llama_model_loader: - kv 13: general.dataset.0.organization str = OpenCoder LLM llama_model_loader: - kv 14: general.dataset.0.repo_url str = https://huggingface.co/OpenCoder-LLM/... llama_model_loader: - kv 15: general.dataset.1.name str = Opc Sft Stage2 llama_model_loader: - kv 16: general.dataset.1.organization str = OpenCoder LLM llama_model_loader: - kv 17: general.dataset.1.repo_url str = https://huggingface.co/OpenCoder-LLM/... llama_model_loader: - kv 18: general.dataset.2.name str = Orca Agentinstruct 1M v1 llama_model_loader: - kv 19: general.dataset.2.version str = v1 llama_model_loader: - kv 20: general.dataset.2.organization str = Microsoft llama_model_loader: - kv 21: general.dataset.2.repo_url str = https://huggingface.co/microsoft/orca... llama_model_loader: - kv 22: general.dataset.3.name str = Orca Math Word Problems 200k llama_model_loader: - kv 23: general.dataset.3.organization str = Microsoft llama_model_loader: - kv 24: general.dataset.3.repo_url str = https://huggingface.co/microsoft/orca... llama_model_loader: - kv 25: general.dataset.4.name str = Hermes Function Calling v1 llama_model_loader: - kv 26: general.dataset.4.version str = v1 llama_model_loader: - kv 27: general.dataset.4.organization str = NousResearch llama_model_loader: - kv 28: general.dataset.4.repo_url str = https://huggingface.co/NousResearch/h... llama_model_loader: - kv 29: general.dataset.5.name str = NuminaMath CoT llama_model_loader: - kv 30: general.dataset.5.organization str = AI MO llama_model_loader: - kv 31: general.dataset.5.repo_url str = https://huggingface.co/AI-MO/NuminaMa... llama_model_loader: - kv 32: general.dataset.6.name str = NuminaMath TIR llama_model_loader: - kv 33: general.dataset.6.organization str = AI MO llama_model_loader: - kv 34: general.dataset.6.repo_url str = https://huggingface.co/AI-MO/NuminaMa... llama_model_loader: - kv 35: general.dataset.7.name str = Tulu 3 Sft Mixture llama_model_loader: - kv 36: general.dataset.7.organization str = Allenai llama_model_loader: - kv 37: general.dataset.7.repo_url str = https://huggingface.co/allenai/tulu-3... llama_model_loader: - kv 38: general.dataset.8.name str = Dolphin Coder llama_model_loader: - kv 39: general.dataset.8.organization str = Cognitivecomputations llama_model_loader: - kv 40: general.dataset.8.repo_url str = https://huggingface.co/cognitivecompu... llama_model_loader: - kv 41: general.dataset.9.name str = Smoltalk llama_model_loader: - kv 42: general.dataset.9.organization str = HuggingFaceTB llama_model_loader: - kv 43: general.dataset.9.repo_url str = https://huggingface.co/HuggingFaceTB/... llama_model_loader: - kv 44: general.dataset.10.name str = Samantha Data llama_model_loader: - kv 45: general.dataset.10.organization str = Cognitivecomputations llama_model_loader: - kv 46: general.dataset.10.repo_url str = https://huggingface.co/cognitivecompu... llama_model_loader: - kv 47: general.dataset.11.name str = CodeFeedback Filtered Instruction llama_model_loader: - kv 48: general.dataset.11.organization str = M A P llama_model_loader: - kv 49: general.dataset.11.repo_url str = https://huggingface.co/m-a-p/CodeFeed... llama_model_loader: - kv 50: general.dataset.12.name str = Code Feedback llama_model_loader: - kv 51: general.dataset.12.organization str = M A P llama_model_loader: - kv 52: general.dataset.12.repo_url str = https://huggingface.co/m-a-p/Code-Fee... llama_model_loader: - kv 53: general.languages arr[str,1] = ["en"] llama_model_loader: - kv 54: llama.block_count u32 = 32 llama_model_loader: - kv 55: llama.context_length u32 = 131072 llama_model_loader: - kv 56: llama.embedding_length u32 = 4096 llama_model_loader: - kv 57: llama.feed_forward_length u32 = 14336 llama_model_loader: - kv 58: llama.attention.head_count u32 = 32 llama_model_loader: - kv 59: llama.attention.head_count_kv u32 = 8 llama_model_loader: - kv 60: llama.rope.freq_base f32 = 500000.000000 llama_model_loader: - kv 61: llama.attention.layer_norm_rms_epsilon f32 = 0.000010 llama_model_loader: - kv 62: llama.attention.key_length u32 = 128 llama_model_loader: - kv 63: llama.attention.value_length u32 = 128 llama_model_loader: - kv 64: general.file_type u32 = 15 llama_model_loader: - kv 65: llama.vocab_size u32 = 128258 llama_model_loader: - kv 66: llama.rope.dimension_count u32 = 128 llama_model_loader: - kv 67: tokenizer.ggml.model str = gpt2 llama_model_loader: - kv 68: tokenizer.ggml.pre str = llama-bpe llama_model_loader: - kv 69: tokenizer.ggml.tokens arr[str,128258] = ["!", "\"", "#", "$", "%", "&", "'", ... llama_model_loader: - kv 70: tokenizer.ggml.token_type arr[i32,128258] = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, ... llama_model_loader: - kv 71: tokenizer.ggml.merges arr[str,280147] = ["Ġ Ġ", "Ġ ĠĠĠ", "ĠĠ ĠĠ", "... llama_model_loader: - kv 72: tokenizer.ggml.bos_token_id u32 = 128000 llama_model_loader: - kv 73: tokenizer.ggml.eos_token_id u32 = 128256 llama_model_loader: - kv 74: tokenizer.ggml.padding_token_id u32 = 128001 llama_model_loader: - kv 75: tokenizer.chat_template str = {% if not add_generation_prompt is de... llama_model_loader: - kv 76: general.quantization_version u32 = 2 llama_model_loader: - type f32: 66 tensors llama_model_loader: - type q4_K: 193 tensors llama_model_loader: - type q6_K: 33 tensors print_info: file format = GGUF V3 (latest) print_info: file type = Q4_K - Medium print_info: file size = 4.58 GiB (4.89 BPW) load: printing all EOG tokens: load: - 128001 ('<|end_of_text|>') load: - 128008 ('<|eom_id|>') load: - 128009 ('<|eot_id|>') load: - 128256 ('<|im_end|>') load: special tokens cache size = 258 load: token to piece cache size = 0.7999 MB print_info: arch = llama print_info: vocab_only = 0 print_info: n_ctx_train = 131072 print_info: n_embd = 4096 print_info: n_layer = 32 print_info: n_head = 32 print_info: n_head_kv = 8 print_info: n_rot = 128 print_info: n_swa = 0 print_info: is_swa_any = 0 print_info: n_embd_head_k = 128 print_info: n_embd_head_v = 128 print_info: n_gqa = 4 print_info: n_embd_k_gqa = 1024 print_info: n_embd_v_gqa = 1024 print_info: f_norm_eps = 0.0e+00 print_info: f_norm_rms_eps = 1.0e-05 print_info: f_clamp_kqv = 0.0e+00 print_info: f_max_alibi_bias = 0.0e+00 print_info: f_logit_scale = 0.0e+00 print_info: f_attn_scale = 0.0e+00 print_info: n_ff = 14336 print_info: n_expert = 0 print_info: n_expert_used = 0 print_info: causal attn = 1 print_info: pooling type = 0 print_info: rope type = 0 print_info: rope scaling = linear print_info: freq_base_train = 500000.0 print_info: freq_scale_train = 1 print_info: n_ctx_orig_yarn = 131072 print_info: rope_finetuned = unknown print_info: model type = 8B print_info: model params = 8.03 B print_info: general.name = Dolphin 3.0 Llama 3.1 8B print_info: vocab type = BPE print_info: n_vocab = 128258 print_info: n_merges = 280147 print_info: BOS token = 128000 '<|begin_of_text|>' print_info: EOS token = 128256 '<|im_end|>' print_info: EOT token = 128256 '<|im_end|>' print_info: EOM token = 128008 '<|eom_id|>' print_info: PAD token = 128001 '<|end_of_text|>' print_info: LF token = 198 'Ċ' print_info: EOG token = 128001 '<|end_of_text|>' print_info: EOG token = 128008 '<|eom_id|>' print_info: EOG token = 128009 '<|eot_id|>' print_info: EOG token = 128256 '<|im_end|>' print_info: max token length = 256 load_tensors: loading model tensors, this can take a while... (mmap = false) load_tensors: CPU model buffer size = 4685.32 MiB llama_context: constructing llama_context llama_context: n_seq_max = 1 llama_context: n_ctx = 131072 llama_context: n_ctx_per_seq = 131072 llama_context: n_batch = 512 llama_context: n_ubatch = 512 llama_context: causal_attn = 1 llama_context: flash_attn = disabled llama_context: kv_unified = false llama_context: freq_base = 500000.0 llama_context: freq_scale = 1 llama_context: CPU output buffer size = 0.50 MiB llama_kv_cache: CPU KV buffer size = 16384.00 MiB llama_kv_cache: size = 16384.00 MiB (131072 cells, 32 layers, 1/1 seqs), K (f16): 8192.00 MiB, V (f16): 8192.00 MiB llama_context: CPU compute buffer size = 8484.01 MiB llama_context: graph nodes = 1158 llama_context: graph splits = 1 time=2025-12-04T23:29:52.367-08:00 level=INFO source=server.go:1332 msg="llama runner started in 2.81 seconds" time=2025-12-04T23:29:52.367-08:00 level=INFO source=sched.go:517 msg="loaded runners" count=1 time=2025-12-04T23:29:52.367-08:00 level=INFO source=server.go:1294 msg="waiting for llama runner to start responding" time=2025-12-04T23:29:52.367-08:00 level=INFO source=server.go:1332 msg="llama runner started in 2.81 seconds" [GIN] 2025/12/04 - 23:29:52 | 200 | 3.1097964s | 127.0.0.1 | POST "/api/generate" ``` ### OS Windows ### GPU Nvidia ### CPU AMD ### Ollama version 0.13.1
GiteaMirror added the bug label 2026-05-04 23:17:04 -05:00
GiteaMirror commented 2026-05-04 23:17:11 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Dec 5, 2025):

Set OLLAMA_DEBUG=2 in the server environment and post the log up to the line that says msg="inference compute". This will give more detail about the GPU discovery process.

<!-- gh-comment-id:3616080953 --> @rick-github commented on GitHub (Dec 5, 2025): Set `OLLAMA_DEBUG=2` in the server environment and post the log up to the line that says `msg="inference compute"`. This will give more detail about the GPU discovery process.
GiteaMirror commented 2026-05-04 23:17:15 -05:00
Author
Owner
Copy Link

@jason-witter commented on GitHub (Dec 5, 2025):

time=2025-12-05T10:00:49.003-08:00 level=INFO source=routes.go:1544 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:DEBUG-4 OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY:cuda OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:C:\\Users\\Jason\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES:]"
time=2025-12-05T10:00:49.016-08:00 level=INFO source=images.go:522 msg="total blobs: 6"
time=2025-12-05T10:00:49.017-08:00 level=INFO source=images.go:529 msg="total unused blobs removed: 0"
time=2025-12-05T10:00:49.019-08:00 level=INFO source=routes.go:1597 msg="Listening on 127.0.0.1:11434 (version 0.13.1)"
time=2025-12-05T10:00:49.019-08:00 level=DEBUG source=sched.go:120 msg="starting llm scheduler"
time=2025-12-05T10:00:49.021-08:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:98 msg="skipping available library at user's request" requested=cuda libDir=C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\cuda_v12
time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:98 msg="skipping available library at user's request" requested=cuda libDir=C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\cuda_v13
time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:98 msg="skipping available library at user's request" requested=cuda libDir=C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\rocm
time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:98 msg="skipping available library at user's request" requested=cuda libDir=C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\vulkan
time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0
time=2025-12-05T10:00:49.039-08:00 level=TRACE source=runner.go:174 msg="supported GPU library combinations before filtering" supported=map[]
time=2025-12-05T10:00:49.039-08:00 level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=19.5169ms
time=2025-12-05T10:00:49.039-08:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="63.9 GiB" available="47.0 GiB"
time=2025-12-05T10:00:49.039-08:00 level=INFO source=routes.go:1638 msg="entering low vram mode" "total vram"="0 B" threshold="20.0 GiB"

Posting for posterity, but the root of the problem was that I had set OLLAMA_LLM_LIBRARY = cuda as an attempt to get out of CPU only mode that I was using for debugging. Most likely I just needed to re-pull dolphin3 to solve my original problem but I ended up causing more problems for myself along the way. Thank you for your time and attention.

<!-- gh-comment-id:3618004965 --> @jason-witter commented on GitHub (Dec 5, 2025): ``` time=2025-12-05T10:00:49.003-08:00 level=INFO source=routes.go:1544 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:4096 OLLAMA_DEBUG:DEBUG-4 OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY:cuda OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:C:\\Users\\Jason\\.ollama\\models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES:]" time=2025-12-05T10:00:49.016-08:00 level=INFO source=images.go:522 msg="total blobs: 6" time=2025-12-05T10:00:49.017-08:00 level=INFO source=images.go:529 msg="total unused blobs removed: 0" time=2025-12-05T10:00:49.019-08:00 level=INFO source=routes.go:1597 msg="Listening on 127.0.0.1:11434 (version 0.13.1)" time=2025-12-05T10:00:49.019-08:00 level=DEBUG source=sched.go:120 msg="starting llm scheduler" time=2025-12-05T10:00:49.021-08:00 level=INFO source=runner.go:67 msg="discovering available GPUs..." time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:98 msg="skipping available library at user's request" requested=cuda libDir=C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\cuda_v12 time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:98 msg="skipping available library at user's request" requested=cuda libDir=C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\cuda_v13 time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:98 msg="skipping available library at user's request" requested=cuda libDir=C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\rocm time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:98 msg="skipping available library at user's request" requested=cuda libDir=C:\Users\Jason\AppData\Local\Programs\Ollama\lib\ollama\vulkan time=2025-12-05T10:00:49.038-08:00 level=DEBUG source=runner.go:124 msg="evaluating which, if any, devices to filter out" initial_count=0 time=2025-12-05T10:00:49.039-08:00 level=TRACE source=runner.go:174 msg="supported GPU library combinations before filtering" supported=map[] time=2025-12-05T10:00:49.039-08:00 level=DEBUG source=runner.go:40 msg="GPU bootstrap discovery took" duration=19.5169ms time=2025-12-05T10:00:49.039-08:00 level=INFO source=types.go:60 msg="inference compute" id=cpu library=cpu compute="" name=cpu description=cpu libdirs=ollama driver="" pci_id="" type="" total="63.9 GiB" available="47.0 GiB" time=2025-12-05T10:00:49.039-08:00 level=INFO source=routes.go:1638 msg="entering low vram mode" "total vram"="0 B" threshold="20.0 GiB" ``` Posting for posterity, but the root of the problem was that I had set `OLLAMA_LLM_LIBRARY = cuda` as an attempt to get out of CPU only mode that I was using for debugging. Most likely I just needed to re-pull dolphin3 to solve my original problem but I ended up causing more problems for myself along the way. Thank you for your time and attention.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label bug
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#70869
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?