The 16 GB RX 9060 XT is a separate memory variant from the 8 GB card. Its proposed Qwen3.5-9B Q8 artifact and 32K catalog context ask a different question from the 8 GB Q4 profile: whether a larger quantization leaves enough room for context, compute buffers and normal desktop use. The answer requires a card-specific test.
If comparing variants, keep the prompts and output target fixed, then disclose that the quantization differs. Equal model family names do not mean identical memory use or answer behavior. This page has no approved load, fit, quality or speed result, and the Linux candidate is not evidence for Windows or vice versa.
System setup
PCIe link. AMD lists PCIe 5.0 x16 (AMD specifications). Use the slot your motherboard manual lists as x16. On a riser rated for PCIe 4.0, set the slot to Gen 4 (our recommendation). If the link reads slow under load, see the guide's troubleshooting order.
Loading (calculated). This card's FastLLM model, Qwen3.5-9B Q8_0 (9.53 GB), loads in at least 1.4 s from a PCIe 4.0 NVMe drive and 17.3 s from a SATA SSD; the drive sets that, not the PCIe generation. A reload from system memory takes 0.15 s at PCIe 5.0 x16 (0.30 s at 4.0).
Engine. FastLLM, in development, uses llama.cpp's Vulkan engine on this card. Bitworks has no lab observation for this card yet. AMD lists this card as RDNA4, gfx1200, in its ROCm GPU specifications. FastLLM's HIP engine currently covers only the RX 7900 XT and XTX.
Resizable BAR. Turn it on; it matters most for the Vulkan engine this card uses. If FastLLM detects it off, it applies its workaround and shows how to turn it on. The workaround is a fallback, not a replacement.
Driver. Use AMD Software: Adrenalin Edition, the current WHQL Recommended release. Our test PCs use the Minimal install option.
Power and cooling. AMD lists a typical board power of 160 W and a minimum power supply recommendation of 450 W (AMD specifications). Plan airflow for sustained load, not short bursts.
Discussion
Share setups, questions and your own results. Posts are reviewed before they appear. Replies from Bitworks staff are marked.