The point is not how fast it is now. The point is that this opens new possibilities that can be built on. Potentially models that are trained with slightly different architectures to optimize to this use case. Possibly others come to improve this path. Possibly HW manufacturers make a few small adjustments that remove bottlenecks. Who knows, the next person may combine CPU compute with this mem sharing to get another token a second. Then the next person does predictive loading into memory to keep that bandwith 100% maxed and usable. Then the next does and the next does. Before you know it there is a real thing there that never existed.

This is a great project. I love the possibilities it hints at. Thanks for building it!

ExLlamaV3 EXL3 2bpw is likely the 30b parameter GLM 4.7 Flash quantised down to 2 bits, the unstated assumption is that you need to check the 2bpw quantisation works well enough for your use case.

The reported size of the ModelOpt FP8, 16 GB, sounds wrong to me. If its 8 bits per parameter it is going to be a similar size to the glm-4.7-flash:q8_0. They repeat this a few times in the readme.

KV cache is, well, a cache that can fill up and trigger eviction. You require enough space to execute at least 1 fwd pass of 1 request at your context length. KV cache hits reduce TTFT by avoiding prefill. You don’t get to skip decode.

MoE is kinda related in terms of lower usage requirements vs a dense model of same total param size, but I think your mental model is a bit off.

> it can generate a 50 steps 512x512 image around 1 minute and 50 seconds.

I have the 4650G APU, and the best way to describe it is: lacking of support. This was even more true 3 yo than now. rocm (is) was absolutely dogshit then, I know this because I tried to do the same when that post was made. You have to compile everything from scratch, get the relevant patches, and even then, xformers which is a library that accelerate diffusion model inferencing was not supported for renoir or rocm back then. Yes, you can generate an image, but it was much slower, and rigged with bugs. You couldn'rt update rocm because it broke compatibility, and it was partly the reason I got into nixos. That being said, those APUs are a power house. Nowadays I can run decent agentic workflows on them (I have 64gb of ddr4 ram, ie APU can suck as much as it needs with the latest linux kernels).

Just note, diffusion models are still second class citizens on AMD apus even GPUs. But then again, nothing close right now on the market except for what apple offers.

The readme opens with this:

> I have an RTX 5070 with 12 GB VRAM and I wanted to run glm-4.7-flash:q8_0, which is a 31.8 GB model. The standard options are:

> Offload layers to CPU — works, but drops token/s by 5–10× because CPU RAM has no CUDA coherence. You end up waiting. Use a smaller quantization — you lose quality. At q4_0 the model is noticeably worse on reasoning tasks.

> Buy a bigger GPU — not realistic for consumer hardware. A 48 GB card costs more than a complete workstation.

> None of those felt right, so I built an alternative: route the overflow memory to DDR4 via DMA-BUF, which gives the GPU direct access to system RAM over PCIe 4.0 without a CPU copy involved.

And then limps home with this caveat on the closest thing to a benchmark:

> The PCIe 4.0 link (~32 GB/s) is the bottleneck when the model overflows VRAM. The best strategy is to shrink the model until it fits — either with EXL3 quantization or ModelOpt PTQ — and use GreenBoost's DDR4 pool for KV cache only.

I think the reason it refers it to DDR4 is because that is how the user explained it to their coding agent. LLMs are great at perpetuating unnecessary specificity.

CUDA has had managed memory that pages between VRAM and system RAM for a decade. Problem is doing so is unusably slow for AI purposes. Seems like an unnecessary layer here.

Presumably it means that software doesn’t have to write the same sort of layer offloading support. It’ll “just work” as if you had X GB of VRAM all along.

The nvidia windows driver enables RAM swapping by default.

Great way to backstab you if you prefer inference speed.

I don't think Windows does this, but Ollama does

> it decided to "solve the problem in software"

This isn't made by nvidia

And burn the swap pagesys file to a rewritable DVD to complete the cycle. It will be super fast that way.

If you are doing video models, this is an excellent way to murder your SSD.

Do not put swap on an SSD you care about at all.

It makes the difference between being able to run a lot of machine learning tasks, and not being able at all. Pretty useful.

I would say it depends entirely on your usecase. I don't think there can be a simple "not useful" generalization that applies to everyone.

Can you elaborate beyond the shallow/superficial dismissal?