Can my Radeon run it? · RX 9070 XT · Qwen3.6-35B · llama.cpp
Fits with RAM offload
RX 9070 XT (16 GB) × Qwen3.6-35B. Shopper file is UD-Q5_K_XL at 256K. Calculator verdict: fits with ram offload. llama.cpp first (Vulkan, then HIP). Q3 is a labeled estimate, not a house pick. Measured decode appears only where a local desktop card already exists. We do not invent a price.
How we test — Measured cells cite official amdai-bench/run@1 cards. Estimates stay labeled. Q3 is estimate-only. Q1/Q2 stay off.
Weights are official GGUF listings when the model page has them. Totals include KV at f16. 1-GPU and 2-GPU columns use the same calculator as this share URL.
| Quant | Weights | 1× 4K | 1× 32K | 1× 256K | 2× 4K | 2× 32K | 2× 256K | Speed |
|---|---|---|---|---|---|---|---|---|
|
Q6_K
Estimate |
26.7 GB
Estimate: total_params × 6.56 bpw (nominal GGUF Q6_K; UD-Q6_K_XL uses the same bucket) / 8. Active params do not shrink the file. No listed publisher file size for this quant. |
Does not fit this GPU
27.7 GB |
Does not fit this GPU
28.2 GB |
Does not fit this GPU
32.7 GB |
Fits in VRAM (estimate)
13.8 GB/GPU |
Tight — little desktop headroom
14.1 GB/GPU |
Does not fit this GPU
16.3 GB/GPU |
not measured yet
No official local desktop cell for this model × card. Empty stays not measured yet. |
|
UD-Q5_K_XL
Estimate |
24.8 GB
Hugging Face listing for Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf (26.6 GB) |
Does not fit this GPU
25.7 GB |
Does not fit this GPU
26.2 GB |
Does not fit this GPU
30.7 GB |
Fits in VRAM (estimate)
12.8 GB/GPU |
Fits in VRAM (estimate)
13.1 GB/GPU |
Tight — little desktop headroom
15.3 GB/GPU |
not measured yet
No official local desktop cell for this model × card. Empty stays not measured yet. |
|
Q4_K_M
Estimate |
19.8 GB
Estimate: total_params × 4.85 bpw (nominal GGUF Q4_K_M) / 8. Active params do not shrink the file. No listed publisher file size for this quant. |
Does not fit this GPU
20.6 GB |
Does not fit this GPU
21.2 GB |
Does not fit this GPU
25.6 GB |
Fits in VRAM (estimate)
10.3 GB/GPU |
Fits in VRAM (estimate)
10.6 GB/GPU |
Fits in VRAM (estimate)
12.8 GB/GPU |
not measured yet
No official local desktop cell for this model × card. Empty stays not measured yet. |
|
UD-Q3_K_XL
Q3 estimate |
15.6 GB
Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB) |
Does not fit this GPU
16.4 GB |
Does not fit this GPU
17.0 GB |
Does not fit this GPU
21.4 GB |
Fits in VRAM (estimate)
8.21 GB/GPU |
Fits in VRAM (estimate)
8.49 GB/GPU |
Fits in VRAM (estimate)
10.7 GB/GPU |
not measured yet
UD-Q3_K_XL is a labeled estimate — not a local desktop run. |
When the file does not stay in 16 GB VRAM, the spill is system RAM. Shopper lines stay Minimum system RAM needed and RAM needed at full context for UD-Q5_K_XL on this card. Estimate unless a measured R9700 card overlays the kit.
Minimum system RAM needed
48 GB kit
Estimate · 38.7 GiB raw · 256K context
RAM needed at full context
48 GB kit
Estimate · 38.7 GiB raw · 256K full context
Estimate. Offloaded weights + KV in RAM + runtime + OS/desktop + headroom, rounded up to a 48 GB DIMM kit.
Vulkan first, then ROCm/HIP. Copy a house recipe when one exists. llama.cpp stays first. Strata / vLLM stay on /backends.
llama.cpp Vulkan
House daily-driver recipe. Copy the recipe page for the 262K pin.
# llama.cpp Vulkan. House pin when a recipe exists; otherwise a shape only. ./build-vk/bin/llama-server \ -m ~/models/Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf \ -ngl 99 \ -c 262144 \ --cache-type-k f16 \ --cache-type-v f16 \ --host 127.0.0.1 --port 8080
llama.cpp HIP
HIP decode is not measured yet. Confirm gfx with rocminfo first.
# HIP shape only — house 35B cards are Vulkan. HIP decode is not measured yet. ./build-hip/bin/llama-server \ -m ~/models/Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf \ -ngl 99 \ -c 262144 \ --cache-type-k f16 \ --cache-type-v f16 \ --host 127.0.0.1 --port 8080
Qwen3.6-35B on the other published Radeon sheets.
Existing /go/ routes for this GPU sheet. Buttons say Check price — we do not print a figure.
Check price · no live or invented prices
Return terms vary by seller and item. Newegg's Hassle-Free page covers only Sold-and-Shipped-by-Newegg listings; marketplace items use the seller's policy. See the listing's Warranty & Returns tab.
As an Amazon Associate I earn from qualifying purchases. Some Newegg links use Rakuten Advertising and some B&H links use B&H’s affiliate program when those IDs are configured. Micro Center and AMD Store links are plain “also available at” links — no tracking. Buttons say “Check price” and never invent or live-quote a price. Full disclosure.
Measure the pair on your box. Doctor and planner are allowlisted. House daily / Flash profiles appear only for those house models.
If the app is not installed, packages are not published yet. There is no public download URL. Read the Stack Lab install notes . If Stack Lab is already running on this machine, open the local app .
If the app is not installed, packages are not published yet. There is no public download URL. Read the Stack Lab install notes . If Stack Lab is already running on this machine, open the local app .
If the app is not installed, packages are not published yet. There is no public download URL. Read the Stack Lab install notes . If Stack Lab is already running on this machine, open the local app .
If the app is not installed, packages are not published yet. There is no public download URL. Read the Stack Lab install notes . If Stack Lab is already running on this machine, open the local app .
Also: All pairs · RX 9070 XT · Qwen3.6-35B · Official GGUF · Official card · llama.cpp · Speed up llama.cpp · Dual / multi-GPU