More
Best rig by budgetRent before you buySubmit a buildNewsletterFree datasetAdvertise
Submit build

Builds / AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU)

Strix Halo (Ryzen AI Max+ 395, 128 GB): Qwen3.8 Flash-Next Q4_K_XL at 20.8 tok/s, 36.1 with MTP (EngramHalo.cpp)

Last updated:

Source: kksoftwareag.github.io ↗More source: github.com ↗📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
GPU(s)
AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU)
Amazon · eBay (used) · Newegg
VRAM / unified memory (GB)
128
CPU
AMD Ryzen AI Max+ 395
Amazon · eBay (used) · Newegg
Motherboard
not stated
RAM
128 GB unified memory
PSU
not stated
Other
Fedora 44, ROCm/HIP 7.1; EngramHalo.cpp (Strix Halo llama.cpp fork that keeps the 26.8 GB n-gram table on NVMe); dzannotti MTP head
Reported total cost
not stated
Power
not stated

Measured speed (generation tokens/sec)

ModelQuantEnginetok/sPrompt processing (tok/s)Details
Qwen3.8 Flash-NextQ4_K_XL, KV q8_0, no MTPEngramHalo.cpp (llama.cpp fork, ROCm/HIP)20.8260 @ 38,144-token prompt
est. TTFT @8k ≈ 31.5 s (calculated from measured prefill)
prefill source: kksoftwareag.github.io ↗		context: 32K ctx📏 Single user, as printed in the source "Decode speed" panel
Qwen3.8 Flash-NextQ4_K_XL + MTP (dzannotti head), KV q8_0EngramHalo.cpp (llama.cpp fork, ROCm/HIP)36.1not stated⚡ spec-decode (MTP)context: 32K ctx, code prompts📏 MTP on code prompts; the source shows 41 on reasoning prompts, we list the lower code figure

Prompt processing = prefill speed (how fast the prompt is read) as reported by the source, with the prompt length it was measured at; blank = the source does not report it. est. TTFT @8k is calculated from the measured prefill speed (8,192 ÷ prefill tok/s), shown only when prefill was measured on a prompt of 512+ tokens. It is not a measurement and ignores load time and caching. measured TTFT is the source's own time-to-first-token. Why prefill matters for coding assistants.

kksoftwareag project docs, as of 16 Sep 2026. The source also runs a second engine (halogen-flash-server, 35.9 t/s single user with its own 5.53-bit checkpoint); not listed here because whether that figure uses its MTP head is not stated. MTP lowers total throughput once two or more users share the box.

Source: https://kksoftwareag.github.io/strix-halo-qwen-3.8-flash-next/
https://github.com/kksoftwareag/strix-halo-qwen-3.8-flash-next

Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.

Embed this result

Copy into a README or blog. Links back to this build. See badge docs (CC BY 4.0 attribution). The SVG does not invent speeds — numbers stay on the build page.

Markdown

[![Measured on RigBase](https://getrigbase.com/badge/strix-halo-128gb-qwen38-flash-next-engramhalo-kksoftware.svg)](https://getrigbase.com/builds/strix-halo-128gb-qwen38-flash-next-engramhalo-kksoftware/)

HTML

<a href="https://getrigbase.com/builds/strix-halo-128gb-qwen38-flash-next-engramhalo-kksoftware/" rel="noopener"><img src="https://getrigbase.com/badge/strix-halo-128gb-qwen38-flash-next-engramhalo-kksoftware.svg" alt="Measured on RigBase" height="20"/></a>

Preview: Measured on RigBase · How to use

Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)

Quick answers: Can Strix Halo (Radeon 8060S) run Qwen3.8 Flash-Next?