Hacker News Reader

Comment by D4Ha | Hacker News Reader

Do you find it useful or worthwhile to split a large LLM across two GPUs on a desktop?

If you've tried it, what worked well and what didn't? I'm especially interested in mismatched VRAM setups, e.g. a 16 GB GPU + a 24 GB GPU.

How much overhead did you see from inter-GPU transfers, and did the extra usable VRAM outweigh the performance hit?

View on HN

It depends on what model you’re running, and for what workload. For personal use (one or two convos at a time) with models that fit in gpu memory, pcie bandwidth doesn't really matter. Just try and be on gen 3 x8 or higher.

Llama is decent at auto optimizing it if you let it use both gpus. It’ll split the workload so the contiguous layers are all on one gpu. Once the model is loaded, you only transfer weights between gpus once (per token?), at the layer boundary.

I run on an 8gb 3070 and 12 gb 3060, and the only weird thing is that the weaker card gets more layers (and therefore work) because it has more ram.

Oh, if you’re barely fitting the models into your vram, you may need to explicitly adjust the layer balance between cards — sometimes it fails to realize it should have put certain things (like draft models) on the other card so you can fit one more layer in.

A hetero-GPU setup is definitely cost-effective if you don't strictly require the raw speed of a top-tier card like 5090. Just keep in mind that the total throughput will also be bottlenecked by the slower card.

To provide some anecdotal data, here is how my 5090 + 3060 setup performs with Qwen 3.8 27B (Unsloth's UD-Q4 with MTP):

  Single 5090: 101 t/s (TG), 2650 t/s (PP)
  5090 + 3060: 53 t/s (TG), 1700 t/s (PP)

For reference, here are also some numbers from my 4060ti + 3060 (16GB + 12GB) setup. [0]

[0]: https://news.ycombinator.com/item?id=48700091

I wonder if turning on GGML_CUDA_NCCL would help reduce delay between cards

As opposed to loading it up in RAM + VRAM? Pretty much always better to split it up to multiple GPUs. My priorities are load up all available VRAM, then offload MoE experts to RAM (if possible), then offload other layers.

I use an RTX 3090 (24GB) and a GTX 1080ti (11GB). Just the 3090 for 3.8 I get maybe 60tg/s (UD-Q4 quant), for both I get around 40tg/s (UD-Q6). Not apples to apples though considering it's different quants.

Here's my config: https://gitea.va.reichard.io/evan/nix/src/branch/master/modu...

Yes. I can split a model like this across 3 GPUs (a 1080 with 8GB and two Titan Vs with 12GB), and it's much faster than running it on 36 CPU cores. As long as it fits in aggregate VRAM, it seems very advantageous to do so.

That's a very deep rabbit hole involving PCIe topology on both the hardware and software (NCCL) side, among other things. It's too system-specific to answer directly, but the entrance to said hole can be found at https://github.com/local-inference-lab/rtx6kpro/blob/master/... .

Disregard references to RTX 6000 cards, most of it is generally applicable to all multiple-GPU boxes.

Depends on your pcie connection. If they're both x16 then it's pretty low overhead, x8 is ok, but x4 is too slow. Also it's a bit tricky getting an optimal setups with mismatched vram, I think you could probably still make use of the full vram if you're clever but it's trickier.

for layer parallelism (e.g. to get more vram) the bandwidth between layers is essentially nothing (like 16kb per token I think), so I don't think x4 would even be a problem!

Good point. It's much more of an issue when running dense models with tensor parallelism. In that case, I'd look for an MoE model instead.

I haven't tried this either but I'm guessing if you could pool the GPU memory over whatever the kids are using these days, I think it was SLI back in my day. The GPU memory should still be faster than the RAM?

Unfortunately I checked, SLI doesnt work for this situation. Because the program loading the LLM uses CUDA library, which doesnt account for/takes advantage of SLI for this purpose at least.

Reply on news.ycombinator.com