Does Hy3 UD128 fit in the AMD Ryzen AI Max+ 395 (Radeon 8060S)'s memory?

Estimate: the UD128 file is 116.7 GB and needs about 119.7 GB with an ~8K context; the AMD Ryzen AI Max+ 395 (Radeon 8060S) has about 96.0 GB usable, so it does not fully fit (part of it runs on the CPU).

Home › Models › Hy3 › AMD Ryzen AI Max+ 395 (Radeon 8060S)

Hy3 on the AMD Ryzen AI Max+ 395 (Radeon 8060S)

Hy3 runs at about 18.4 tokens/s on the AMD Ryzen AI Max+ 395 (Radeon 8060S) (measured, UD128, Vulkan, llama-bench).

Best measured generation18.4tok/s

UD128 · Vulkan

Prompt processing168tok/s

same run

For comparison, our size-scaling estimate for this combination is 12.7 tok/s.

Every run

Quant	Backend	Generate tok/s	Prompt tok/s	Test	Build / by / notes	Source
UD128	Vulkan	18.4	168	decode (llama-bench)	128 GB Strix Halo, Vulkan (RADV), q4_0 KV cache; mainline llama.cpp with the hy_v3 architecture. 2-3 bit expert mix. File size 116,667,103,040 bytes as stated by the source.	YanissAmz/Hy3-295B-A21B-GGUF model card (Hugging Face)
UD128	ROCm	12.4	105	decode (llama-bench)	Same machine and build as the Vulkan row, ROCm backend. File size 116,667,103,040 bytes as stated by the source.	YanissAmz/Hy3-295B-A21B-GGUF model card (Hugging Face)

tg128 = generating 128 tokens; pp512 = reading a 512-token prompt (llama-bench). LocalScore averages several prompt and output lengths, so its numbers run lower than tg128. "decode, llama-server" = single-stream generation speed measured against a running llama-server (no speculative decoding).

Check prices

Amazon Newegg

As an Amazon Associate I earn from qualifying purchases.

No card yet?

Try it on a rented GPU (RunPod)

Affiliate link: if you sign up, RigBench may earn a referral credit. It never affects rankings or numbers.

All results for the AMD Ryzen AI Max+ 395 (Radeon 8060S) · Hy3 on other hardware

Embed this result

HTML

<a href="https://rigbench.pages.dev/model/hy3/on/amd-ryzen-ai-maxplus-395-radeon-8060s/"><img src="https://rigbench.pages.dev/badge/hy3--amd-ryzen-ai-maxplus-395-radeon-8060s.svg" alt="Hy3 · AMD Ryzen AI Max+ 395 (Radeon 8060S): 18.4 tok/s (UD128, measured)"></a>

Markdown (GitHub README, forums)

[![Hy3 · AMD Ryzen AI Max+ 395 (Radeon 8060S): 18.4 tok/s (UD128, measured)](https://rigbench.pages.dev/badge/hy3--amd-ryzen-ai-maxplus-395-radeon-8060s.svg)](https://rigbench.pages.dev/model/hy3/on/amd-ryzen-ai-maxplus-395-radeon-8060s/)