8GB GPU that works with SL12

[johncaf]

Hi,

I have an 8GB 5060 GPU that still still crashes with some modules on SL12.40

I can repurpose this into another PC for other work and could grab another GPU for the DAW. At the moment, I would only have budget for an 8gb, are there any recommendations?

John

[MrSoundman]

Have you tried this?

GPU / CPU Speed

In “Edit | Preferences | System”, the option “Use Shared Memory” seems to be unchecked by default in 12.0.40 [image] Can you try checking that box, and restarting SpectraLayers?

I have a 5060 16GB and even with that, checking “Use Shared Memory” is needed with 12.0.40 in order to get the GPU fully used. It would seem that with this version, 24GB is needed … and unfortunately, there’s no cheap option there.

The system should not crash however; I previously used a 3050 6GB without issue, other than speed. Check your RAM with something like the open-source memtest86+ and update to that latest NVIDIA “studio” driver.

[Co84]

I have a 6GB RTX3050, that made SpectraLayers crash. After upgrading my RAM from 16GB to 32GB, everything in SpectraLayers is stable.

[Jari_Junttila]

I have RTX3050 8gb, but I also have 64gb RAM. Never crashed

If i´m understanded it correctly, some AI architectures loads models on RAM, but processes data in GPU, so even if you have 24gb GPU, but only 8 gb´s of system RAM, you can´t load the model.

Correct me, if i´m wrong, this tech develops too fast for me

[ctreitzell]

I’m waaaaaaay behind all of you with a 1050Ti; but I’ve always had 64GB system RAM. I do not run many modules, tho…mainly UNS…and lately not even that, just doing manual editing

[Sophus]

@Jari_Junttila

so even if you have 24gb GPU, but only 8 gb´s of system RAM, you can´t load the model.

Technically, Windows should allocate swap memory on the SSD. This is much slower than RAM but should still work.