# RTX 5090 not working with PyTorch and Stable Diffusion (sm\_120 unsupported)

**URL:** <https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015>\
**Category:** CUDA Setup and Installation\
**Created:** [July 4, 2025, 11:15am UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015 "2025-07-04T11:15:39Z")\
**Posts on this page:** 12\
**Page:** 1

<div class="post-metadata">

**Author:** ![sstmkjp](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@sstmkjp](https://forums.developer.nvidia.com/u/sstmkjp)\
**Post date:** [July 4, 2025, 11:15am UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/1 "2025-07-04T11:15:39Z")

</div>

Hello,

I recently purchased a laptop with an Hello,

I recently purchased a laptop with an RTX 5090 GPU (Blackwell architecture), but unfortunately, it’s not usable with PyTorch-based frameworks like Stable Diffusion or ComfyUI. The current PyTorch builds do not support CUDA capability sm\_120 yet, which results in errors or CPU-only fallback.

This is extremely disappointing for those of us who invested in high-end hardware expecting out-of-the-box support for AI tools.

Could NVIDIA please work closely with the PyTorch team to ensure official support as soon as possible?

Thanks for your attention.  
(Blackwell architecture), but unfortunately, it’s not usable with PyTorch-based frameworks like Stable Diffusion or ComfyUI. The current PyTorch builds do not support CUDA capability sm\_120 yet, which results in errors or CPU-only fallback.

This is extremely disappointing for those of us who invested in high-end hardware expecting out-of-the-box support for AI tools.

Could NVIDIA please work closely with the PyTorch team to ensure official support as soon as possible?

Thanks for your attention.

---

<div class="post-metadata">

**Author:** ![jlitwinetz](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/jlitwinetz/32/258530_2.png) [@jlitwinetz](https://forums.developer.nvidia.com/u/jlitwinetz)\
**Post date:** [July 28, 2025, 6:45pm UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/2 "2025-07-28T18:45:52Z")

</div>

5060ti tried nightlies pytorch 577 driver comes with 12.9 cuda tried uninstall driver and using 12.8 with driver 571.96 no luck error CUDA error: no kernel image is available for execution on the device  
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.  
For debugging consider passing CUDA\_LAUNCH\_BLOCKING=1.  
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

someone sleeping at the wheel?

can’t use forge flux

---

<div class="post-metadata">

**Author:** ![jlitwinetz](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/jlitwinetz/32/258530_2.png) [@jlitwinetz](https://forums.developer.nvidia.com/u/jlitwinetz)\
**Post date:** [July 30, 2025, 10:48pm UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/3 "2025-07-30T22:48:45Z")

</div>

I have windows 11 and ran a script with the following results:

(base) C:\Windows\System32\>conda activate forge

(forge) C:\Windows\System32\>python  
Python 3.10.18 | packaged by Anaconda, Inc. | (main, Jun 5 2025, 13:08:55) [MSC v.1929 64 bit (AMD64)] on win32  
Type “help”, “copyright”, “credits” or “license” for more information.

> > > import torch
> > > 
> > > print(“Torch version:”, torch. **version** )  
> > > Torch version: 2.6.0.dev20241112+cu121  
> > > print(“CUDA version:”, torch.version.cuda)  
> > > CUDA version: 12.1  
> > > print(“CUDA available:”, torch.cuda.is\_available())  
> > > CUDA available: True  
> > > print(“Device count:”, torch.cuda.device\_count())  
> > > Device count: 1
> > > 
> > > if torch.cuda.is\_available():  
> > > … print(“Device name:”, torch.cuda.get\_device\_name(0))  
> > > … # Try to run a CUDA operation  
> > > … try:  
> > > … x = torch.tensor([1.0], device=“cuda”)  
> > > … print(“Tensor on GPU:”, x)  
> > > … except Exception as e:  
> > > … print(“CUDA operation failed:”, e)  
> > > … else:  
> > > … print(“CUDA is not available.”)  
> > > …  
> > > C:\Users\jlitw\miniconda3\envs\forge\lib\site-packages\torch\cuda\__init_\_.py:235: UserWarning:  
> > > NVIDIA GeForce RTX 5060 Ti with CUDA capability sm\_120 is not compatible with the current PyTorch installation.  
> > > The current PyTorch install supports CUDA capabilities sm\_50 sm\_60 sm\_61 sm\_70 sm\_75 sm\_80 sm\_86 sm\_90.  
> > > If you want to use the NVIDIA GeForce RTX 5060 Ti GPU with PyTorch, please check the instructions at [https://pytorch.org/get-started/locally/](https://pytorch.org/get-started/locally/)

warnings.warn(  
Device name: NVIDIA GeForce RTX 5060 Ti  
Tensor on GPU: CUDA operation failed: CUDA error: no kernel image is available for execution on the device  
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.  
For debugging consider passing CUDA\_LAUNCH\_BLOCKING=1  
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

[![jlitz](https://global.discourse-cdn.com/nvidia/original/4X/1/f/0/1f0442851b6cfc252ab4c522453a799b7c810ec0.png)](https://github.com/jlitz)

## Add a comment

---

<div class="post-metadata">

**Author:** ![rs277](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@rs277](https://forums.developer.nvidia.com/u/rs277)\
**Post date:** [July 31, 2025, 12:00am UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/4 "2025-07-31T00:00:46Z")

</div>

50XX was supported from Cuda 12.8.

> [@jlitwinetz](#):
>
> print(“Torch version:”, torch. **version** )  
> Torch version: 2.6.0.dev20241112+cu121

You appear to have a version compiled with Cuda 12.1

> [@jlitwinetz](#):
>
> If you want to use the NVIDIA GeForce RTX 5060 Ti GPU with PyTorch, please check the instructions at [https://pytorch.org/get-started/locally/](https://pytorch.org/get-started/locally/)

Looking there, 12.8 is an option so you might want to install that.

---

<div class="post-metadata">

**Author:** ![jlitwinetz](https://sea2.discourse-cdn.com/nvidia/user_avatar/forums.developer.nvidia.com/jlitwinetz/32/258530_2.png) [@jlitwinetz](https://forums.developer.nvidia.com/u/jlitwinetz)\
**Post date:** [July 31, 2025, 1:33am UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/5 "2025-07-31T01:33:39Z")

</div>

yes i installed the 12.8 nightly just now results

I installed the 12.8 version and checked to confirm still get error on image generation attempt  
INSTALLED 12.8  
forgeflux) C:\FORGE\webui\_forge\_cu121\_torch231\>python -c “import torch; print(torch. **version** ); print(torch.version.cuda)”  
2.9.0.dev20250726+cu128  
12.8

(forge) C:\Windows\System32\>python  
Python 3.10.18 | packaged by Anaconda, Inc. | (main, Jun 5 2025, 13:08:55) [MSC v.1929 64 bit (AMD64)] on win32  
Type “help”, “copyright”, “credits” or “license” for more information.

> > > import torch
> > > 
> > > print(“Torch version:”, torch. **version** )  
> > > Torch version: 2.9.0.dev20250729+cu128  
> > > print(“CUDA version:”, torch.version.cuda)  
> > > CUDA version: 12.8  
> > > print(“CUDA available:”, torch.cuda.is\_available())  
> > > CUDA available: True  
> > > print(“Device count:”, torch.cuda.device\_count())  
> > > Device count: 1
> > > 
> > > if torch.cuda.is\_available():  
> > > … print(“Device name:”, torch.cuda.get\_device\_name(0))  
> > > … # Try to run a CUDA operation  
> > > … try:  
> > > … x = torch.tensor([1.0], device=“cuda”)  
> > > … print(“Tensor on GPU:”, x)  
> > > … except Exception as e:  
> > > … print(“CUDA operation failed:”, e)  
> > > … else:  
> > > … print(“CUDA is not available.”)  
> > > …  
> > > Device name: NVIDIA GeForce RTX 5060 Ti  
> > > Tensor on GPU: tensor([1.], device=‘cuda:0’)

THE COMPLETE RUN INFO

[Unload] Trying to free 1024.00 MB for cuda:0 with 1 models keep loaded … Current free memory is 9897.80 MB … Done.  
Traceback (most recent call last):  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\modules\_forge\main\_thread.py”, line 30, in work  
self.result = self.func(\*self.args, \*\*self.kwargs)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\modules\txt2img.py”, line 131, in txt2img\_function  
processed = processing.process\_images(p)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\modules\processing.py”, line 842, in process\_images  
res = process\_images\_inner(p)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\modules\processing.py”, line 962, in process\_images\_inner  
p.setup\_conds()  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\modules\processing.py”, line 1601, in setup\_conds  
super().setup\_conds()  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\modules\processing.py”, line 503, in setup\_conds  
self.uc = self.get\_conds\_with\_caching(prompt\_parser.get\_learned\_conditioning, negative\_prompts, total\_steps, [self.cached\_uc], self.extra\_network\_data)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\modules\processing.py”, line 474, in get\_conds\_with\_caching  
cache[1] = function(shared.sd\_model, required\_prompts, steps, hires\_steps, shared.opts.use\_old\_scheduling)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\modules\prompt\_parser.py”, line 189, in get\_learned\_conditioning  
conds = model.get\_learned\_conditioning(texts)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\system\python\lib\site-packages\torch\utils\_contextlib.py”, line 115, in decorate\_context  
return func(\*args, \*\*kwargs)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\backend\diffusion\_engine\flux.py”, line 86, in get\_learned\_conditioning  
cond\_l, pooled\_l = self.text\_processing\_engine\_l(prompt)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\backend\text\_processing\classic\_engine.py”, line 272, in **call**  
z = self.process\_tokens(tokens, multipliers)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\backend\text\_processing\classic\_engine.py”, line 305, in process\_tokens  
z = self.encode\_with\_transformers(tokens)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\webui\backend\text\_processing\classic\_engine.py”, line 128, in encode\_with\_transformers  
self.text\_encoder.transformer.text\_model.embeddings.position\_embedding = self.text\_encoder.transformer.text\_model.embeddings.position\_embedding.to(dtype=torch.float32)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\system\python\lib\site-packages\torch\nn\modules\module.py”, line 1173, in to  
return self.\_apply(convert)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\system\python\lib\site-packages\torch\nn\modules\module.py”, line 804, in \_apply  
param\_applied = fn(param)  
File “C:\FORGE\webui\_forge\_cu121\_torch231\system\python\lib\site-packages\torch\nn\modules\module.py”, line 1159, in convert  
return t.to(  
RuntimeError: CUDA error: no kernel image is available for execution on the device  
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.  
For debugging consider passing CUDA\_LAUNCH\_BLOCKING=1.  
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

CUDA error: no kernel image is available for execution on the device  
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.  
For debugging consider passing CUDA\_LAUNCH\_BLOCKING=1.  
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions  
.

---

<div class="post-metadata">

**Author:** ![rs277](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@rs277](https://forums.developer.nvidia.com/u/rs277)\
**Post date:** [July 31, 2025, 5:06am UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/6 "2025-07-31T05:06:58Z")

</div>

I have no experience with pytorch, but I wonder if you still have part/all of some 12.1 version, as all the lines after “THE COMPLETE RUN INFO”, are prefixed:

File “C:\FORGE\webui\_forge\_cu121\_torch231

which indicates 12.1 somewhere.

---

<div class="post-metadata">

**Author:** ![hameedullah.hassan](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@hameedullah.hassan](https://forums.developer.nvidia.com/u/hameedullah.hassan)\
**Post date:** [September 7, 2025, 7:58am UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/7 "2025-09-07T07:58:42Z")

</div>

Fix: “Torch not compiled with CUDA enabled” in Automatic1111 on RTX 5090 (Windows)

This is a complete, reproducible fix for getting **Automatic1111 Stable Diffusion WebUI** to use the **GPU** on an **RTX 5090**.  
It captures the exact errors I hit, why they happened, and the step‑by‑step commands that solved them.

* * *

## TL;DR (Quick Fix)

1. **Activate your WebUI venv** (mine is `E:\Automatic111\sd-venv312`):

```bat
E:\Automatic111\stable-diffusion-webui> call E:\Automatic111\sd-venv312\Scripts\activate

```

1. **Clean old Torch installs & cache:**

```bat
pip uninstall -y torch torchvision torchaudio xformers
pip cache purge

```

1. **Install PyTorch nightly with CUDA 12.8 (sm\_120 support for RTX 50‑series):**

```bat
pip install --no-cache-dir --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu128

```

1. **Verify GPU is detected:**

```bat
python -c "import torch,torchvision; print('torch',torch. __version__ ,'cuda',getattr(torch.version,'cuda',None)); \
print('avail',torch.cuda.is_available()); print('name', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'NO GPU'); \
print('cap', torch.cuda.get_device_capability(0) if torch.cuda.is_available() else None)"

```

You should see something like:

```auto
torch 2.9.0.dev20xx+cu128 cuda 12.8
avail True
name NVIDIA GeForce RTX 5090
cap (12, 0)

```

1. **Launch WebUI** with a **simple** `webui-user.bat` (no extra Torch commands, no skip‑cuda‑test):

```bat
set COMMANDLINE_ARGS=--opt-sdp-attention
call webui.bat

```

If the UI shows steps running ~20–30 it/s and **no** “CUDA not enabled” errors, you’re good.

* * *

## My Environment (when it failed & then worked)

- Windows
- **GPU:** NVIDIA GeForce **RTX 5090**
- **Python:** 3.10.11 (64‑bit)
- **WebUI:** v1.10.1 (`82a973c0...`)
- **Venv:** `E:\Automatic111\sd-venv312`
- **Final working Torch/TV:**
  - `torch 2.9.0.dev...+cu128`
  - `torchvision 0.24.0.dev...+cu128`
  - CUDA runtime reported by Torch: **12.8**

> Note: cu124 (CUDA 12.4) **does not** include `sm_120` for RTX 50‑series, so those wheels either warn about unsupported capability or fall back to CPU. Nightly **cu128** wheels do include `sm_120` support.

* * *

## The Errors I Saw

### 1) CPU build or CUDA disabled

From WebUI and terminal:

```auto
AssertionError: Torch not compiled with CUDA enabled
Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS ...

```

and

```auto
torch 2.8.0+cpu cuda None is_available False

```

### 2) Older CUDA (12.4) wheels on a 5090

```auto
UserWarning:
NVIDIA GeForce RTX 5090 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 ... sm_90.

```

This means those wheels don’t include **sm\_120** , so the 5090 won’t be used.

* * *

## Root Cause (Why it Broke)

- I had **CPU‑only** or **older CUDA (cu124)** Torch/TV wheels installed.
- **RTX 5090** requires **sm\_120** support, which currently ships in **nightly CUDA 12.8** wheels (`cu128`).
- WebUI’s auto‑install / custom index settings can sometimes pull the wrong wheels (CPU or older CUDA).

* * *

## The Full Fix (Step by Step)

> Paths below are mine; adjust for your setup.

### 0) Open a fresh terminal and activate the correct venv

```bat
call E:\Automatic111\sd-venv312\Scripts\activate

```

Confirm you’re in the venv:

```bat
python -c "import sys; print(sys.executable)"

```

Expected:

```auto
E:\Automatic111\sd-venv312\Scripts\python.exe

```

### 1) Remove bad installs and cache

```bat
pip uninstall -y torch torchvision torchaudio xformers
pip cache purge

```

### 2) Install **nightly cu128** wheels (have `sm_120` for 50‑series)

```bat
pip install --no-cache-dir --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu128

```

### 3) Sanity‑check GPU from Python

```bat
python -c "import torch,torchvision; print('torch',torch. __version__ ); print('torchvision',torchvision. __version__ ); \
print('cuda?',torch.cuda.is_available()); print('cuda runtime',getattr(torch.version,'cuda',None)); \
print(torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'NO GPU'); \
print('cap', torch.cuda.get_device_capability(0) if torch.cuda.is_available() else None)"

```

I got:

```auto
torch 2.9.0.dev...+cu128
torchvision 0.24.0.dev...+cu128
cuda? True
cuda runtime 12.8
NVIDIA GeForce RTX 5090
cap (12, 0)

```

### 4) Keep WebUI from re‑installing the wrong Torch

Use a **minimal** `webui-user.bat`. Mine looks like this:

```bat
@echo off
rem --- Use the venv that already has the correct Torch installed ---
set PYTHON=E:\Automatic111\sd-venv312\Scripts\python.exe
set VENV_DIR=E:\Automatic111\sd-venv312

rem --- Do NOT force torch installs here ---
set TORCH_COMMAND=

rem --- Clean, safe args (no skip-cuda-test needed) ---
set COMMANDLINE_ARGS=--opt-sdp-attention

rem --- Nuke any custom pip index URLs that could fetch CPU/old wheels ---
set TORCH_INDEX_URL=
set PIP_INDEX_URL=
set PIP_EXTRA_INDEX_URL=

call webui.bat

```

> If you _must_ install through WebUI, set `TORCH_COMMAND` to the **nightly cu128** line:
> 
> ```bat
> set TORCH_COMMAND=pip install --no-cache-dir --pre torch torchvision --index-url https://download.pytorch.org/whl/nightly/cu128
> 
> ```
> 
> But I prefer keeping it empty once I’ve installed the right wheels in the venv.

### 5) Launch and confirm it’s using the GPU

On launch I see:

```auto
Applying attention optimization: sdp... done.
Model loaded in 3.1s ...
20/20 [00:00<00:00, 21–27 it/s]

```

That iteration speed is GPU‑level. No more CUDA errors.

* * *

## What Didn’t Work (and Why)

- **cu124 wheels (Torch 2.6.0+cu124, TV 0.21.0+cu124)** → missing `sm_120`, so 5090 prints warnings and/or falls back to CPU.
- **CPU wheels (torch 2.8.0+cpu)** → `torch.cuda.is_available()` is `False` and WebUI throws “not compiled with CUDA”.
- **Adding `--skip-torch-cuda-test`** → just hides the problem; it doesn’t enable GPU.

* * *

## Optional Notes

- **xFormers** is optional. With modern GPUs, PyTorch SDPA (`--opt-sdp-attention`) is fast and stable.
- The TF32 warning from PyTorch 2.9 is harmless; it’s just a heads‑up about a future API change.
- If you ever slip back to CPU, rerun the **uninstall + purge + cu128 install** steps above.

* * *

## Log Snippets (for searchability)

**Failure (CPU / no CUDA):**

```auto
AssertionError: Torch not compiled with CUDA enabled
Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
torch 2.8.0+cpu cuda None is_available False

```

**Failure (old CUDA 12.4 on 5090):**

```auto
UserWarning:
NVIDIA GeForce RTX 5090 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90.

```

**Success:**

```auto
torch 2.9.0.dev...+cu128 cuda 12.8
avail True
name NVIDIA GeForce RTX 5090
cap (12, 0)

Applying attention optimization: sdp... done.
... 20/20 [00:00<00:00, 21–27 it/s]

```

* * *

## Credit / Context

This write‑up is distilled from a live troubleshooting session.  
If it helps you, consider replying with your exact GPU / driver / Torch versions so others can compare.

---

<div class="post-metadata">

**Author:** ![martinpeter.uk](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@martinpeter.uk](https://forums.developer.nvidia.com/u/martinpeter.uk)\
**Post date:** [September 16, 2025, 10:07pm UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/8 "2025-09-16T22:07:34Z")

</div>

You are a legend, it works like a charm. only issue was to run 3) Sanity‑check GPU from Python as there is some SyntaxError: unterminated string literal (detected at line 4), but even without that its finally works. Big thanks for your help.

---

<div class="post-metadata">

**Author:** ![hameedullah.hassan](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@hameedullah.hassan](https://forums.developer.nvidia.com/u/hameedullah.hassan)\
**Post date:** [September 27, 2025, 10:11pm UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/9 "2025-09-27T22:11:19Z")

</div>

glad it helped!!

---

<div class="post-metadata">

**Author:** ![kylefoxaustin](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@kylefoxaustin](https://forums.developer.nvidia.com/u/kylefoxaustin)\
**Post date:** [November 25, 2025, 6:05am UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/10 "2025-11-25T06:05:20Z")

</div>

legendary. simply LEGENDERY

this just completely solved my training issues on my new RTX 5090… I was pulling my hair out trying to fix this myself.

so many thanks.

---

<div class="post-metadata">

**Author:** ![mahmoudeawad](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@mahmoudeawad](https://forums.developer.nvidia.com/u/mahmoudeawad)\
**Post date:** [December 12, 2025, 11:08pm UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/11 "2025-12-12T23:08:49Z")

</div>

\_available() else None)"  
torch 2.10.0.dev20251212+cu128 cuda 12.8  
avail True  
name NVIDIA GeForce RTX 5090  
cap (12, 0) thanx

---

<div class="post-metadata">

**Author:** ![nicksephton](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@nicksephton](https://forums.developer.nvidia.com/u/nicksephton)\
**Post date:** [August 21, 2026, 10:10am UTC](https://forums.developer.nvidia.com/t/rtx-5090-not-working-with-pytorch-and-stable-diffusion-sm-120-unsupported/338015/12 "2026-08-21T10:10:36Z")

</div>

Hi, I’m running a MSI GeForce RTX 5090 32G, and step 3 of this gives me an error:

“ERROR: Cannot install torch and torchvision==0.27.0.dev20260407+cu128 because these package versions have conflicting dependencies.”

“ERROR: ResolutionImpossible: for help visit [Dependency Resolution - pip documentation v26.3.dev0](https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts)”

I basically have no idea what I’m doing (I’m techy but this isn’t my field) so any help welcome!