Is it possible to add more GPU Memory

[Ivanvaagenes]

Hi. Sorry for my terrible English.
I have a Gforce 4070 with 12Gb memory and it works ok with short videos.
Can I add a Dell H56H0 NVIDIA Tesla M10 32GB GDDR5 and use this with wan 2.2 and Comfyui?
If this is possible, will I have 44Gb memory for Comfyui to use?

mvh
Ivan Vågenes

[gershu-ar]

Hey, amigo, nope, you cannot.

1. No VRAM stacking, no single memory pool possible.
2. The M10 lacks the modern CUDA features, FP16/BF16 support, and Tensor Cores required by modern video models like Wan 2.2. Your RTX 4070 is infinitely faster.

Try GGUF / quantized versions (Q4 or Q8) & run ComfyUI with the --lowvram startup argument.

[Ivanvaagenes]

Hay. Tanks for the answer :-)

[NewwieSystem]

Hi Ivan! 😊

Your English is fine, so don't worry about it!

Unfortunately, you can't simply combine the RTX 4070's 12 GB and the Tesla M10's 32 GB to get 44 GB of shared VRAM.

The Tesla M10 has four separate GPUs, each with 8 GB of VRAM. ComfyUI won't automatically combine their memory into one pool.

While both cards may work in the same system, using them together for Wan 2.2 would require a compatible multi-GPU setup, and the M10's older Maxwell architecture may cause compatibility issues with modern AI software.

I'd recommend trying quantized Wan 2.2 models (such as GGUF, where supported) and ComfyUI's low-VRAM options before investing in an M10. If you need more VRAM, a newer GPU with a larger memory capacity would generally be a better choice.

Hope this helps!

[Ivanvaagenes]

Hi.
Tanks for the explanation. :-)

Ivan