Models · Daily driver on one R9700
the house daily driver: Unsloth UD-Q5_K_XL on llama.cpp Vulkan, one R9700, 262K, f16 KV. Every tok/s and measured GB comes from official amdai-bench/run@1 cards. Empty cells stay contribute. Q3 is a labeled estimate.
How we test — Measured cells cite official amdai-bench/run@1 cards. Estimates stay labeled. Q3 is estimate-only. Q1/Q2 stay off.
Official card: 35B total / 3B active, 40 layers, 10 full-attention (GQA 2×256). Only those 10 layers grow a token-length KV cache. The other 30 hold a fixed linear-attention state.
Official card: Qwen/Qwen3.6-35B-A3B · GGUF: unsloth/Qwen3.6-35B-A3B-GGUF. Abliterated derivatives are educational only — read /abliterated, never the default on this page.
Kit sizes are real DIMM steps. Measured peaks come from run cards when we have them; otherwise the line is an estimate. Wherever RAM offload appears, both shopper labels stay on the page.
Daily driver · 35B llama.cpp · 262K
Minimum system RAM needed
42.9 GB
Measured · 42.9 GiB raw · 256K context
local-desktop-20261012-2314-35b-udq5kxl-1gpu-256k-llamacpp-f16
RAM needed at full context
42.9 GB
Measured · 42.9 GiB raw · 256K full context
local-desktop-20261012-2314-35b-udq5kxl-1gpu-256k-llamacpp-f16
Measured peak system RAM on the local desktop: 42.9 GB (Linux GiB labeled GB). Measured, pending re-verify after RAM fix.
Q3 estimate · Qwen3.6-35B-A3B
Minimum system RAM needed
32 GB kit
Estimate · 24.6 GiB raw · 4K context
RAM needed at full context
32 GB kit
Estimate · 24.7 GiB raw · 256K full context
Estimate. Offloaded weights + KV in RAM + runtime + OS/desktop + headroom, rounded up to a 32 GB DIMM kit.
Measured cells cite official amdai-bench/run@1 cards. Empty cells stay contribute.
Q6 / Q5 / Q4, plus Q3 as a labeled estimate. 1 GPU and 2 GPU.
Decode, prefill, peak VRAM per card, VRAM total, and peak system RAM come from
amdai-bench/run@1 cards when a cell is measured. A null field stays not on the card.
Untested cells say no data — contribute
.
Q1 and Q2 stay off.
| Quant | GPUs | Context | Decode | Prefill | VRAM / card | VRAM total | System RAM | Source |
|---|---|---|---|---|---|---|---|---|
| UD-Q5_K_XL | 1 · single | short | 102.39 tok/s | 2761.6 tok/s | 25.45 GB | 25.97 GB | 37.32 GB | local-desktop-20261012-2311-35b-udq5kxl-1gpu-4k-llamacpp-f16 |
| UD-Q5_K_XL | 1 · single | 32K | 90.52 tok/s | 1937.9 tok/s | 26.16 GB | 26.7 GB | 37.83 GB | local-desktop-20261012-2312-35b-udq5kxl-1gpu-32k-llamacpp-f16 |
| UD-Q5_K_XL | 1 · single | 128K | 66.43 tok/s | 942.8 tok/s | 28.22 GB | 28.82 GB | 40.18 GB | local-desktop-20261012-2312-35b-udq5kxl-1gpu-128k-llamacpp-f16 |
| UD-Q5_K_XL | 1 · single | 256K | 44.4 tok/s | 519.2 tok/s | 28.73 GB | 29.34 GB | 42.9 GB | local-desktop-20261012-2314-35b-udq5kxl-1gpu-256k-llamacpp-f16 |
| UD-Q5_K_XL | 2 · split | short | 82.78 tok/s | 2637.3 tok/s | 13.28 / 13.11 GB | 26.39 GB | 37.6 GB | local-desktop-20261012-2319-35b-udq5kxl-2gpu-4k-llamacpp-f16 |
| UD-Q5_K_XL | 2 · split | 32K | 74.8 tok/s | 1916 tok/s | 13.77 / 13.64 GB | 27.41 GB | 38.16 GB | local-desktop-20261012-2319-35b-udq5kxl-2gpu-32k-llamacpp-f16 |
| UD-Q5_K_XL | 2 · split | 128K | 57.94 tok/s | 1071.1 tok/s | 15.47 / 15.48 GB | 30.94 GB | 41.02 GB | local-desktop-20261012-2320-35b-udq5kxl-2gpu-128k-llamacpp-f16 |
| UD-Q5_K_XL | 2 · split | 256K | 42.03 tok/s | 671.7 tok/s | 17.8 / 17.99 GB | 35.74 GB | 44.71 GB | local-desktop-20261012-2321-35b-udq5kxl-2gpu-256k-llamacpp-f16 |
| Q5_K_M | 1 · single | short | 109.73 tok/s | 2710.8 tok/s | 24.16 GB | 24.68 GB | 35.83 GB | local-desktop-20261012-2327-35b-q5km-1gpu-4k-llamacpp-f16 |
| Q5_K_M | 1 · single | 32K | 96.21 tok/s | 1902.6 tok/s | 24.87 GB | 25.39 GB | 36.46 GB | local-desktop-20261012-2327-35b-q5km-1gpu-32k-llamacpp-f16 |
| Q5_K_M | 1 · single | 128K | 69.64 tok/s | 940.3 tok/s | 26.93 GB | 27.43 GB | 38.65 GB | local-desktop-20261012-2328-35b-q5km-1gpu-128k-llamacpp-f16 |
| Q5_K_M | 1 · single | 256K | 40.54 tok/s | 541.7 tok/s | 28.56 GB | 29.15 GB | 42.79 GB | local-desktop-20261012-2329-35b-q5km-1gpu-256k-llamacpp-f16 |
| Q5_K_M | 2 · split | short | 88.19 tok/s | 2647.2 tok/s | 12.71 / 12.41 GB | 25.13 GB | 37.17 GB | local-desktop-20261012-2334-35b-q5km-2gpu-4k-llamacpp-f16 |
| Q5_K_M | 2 · split | 32K | 78.91 tok/s | 1920.9 tok/s | 13.2 / 12.95 GB | 26.13 GB | 37.8 GB | local-desktop-20261012-2335-35b-q5km-2gpu-32k-llamacpp-f16 |
| Q5_K_M | 2 · split | 128K | 59.86 tok/s | 1067.3 tok/s | 14.98 / 14.79 GB | 29.76 GB | 40.73 GB | local-desktop-20261012-2335-35b-q5km-2gpu-128k-llamacpp-f16 |
| Q5_K_M | 2 · split | 256K | 42.33 tok/s | 664.6 tok/s | 17.45 / 17.3 GB | 34.54 GB | 46.21 GB | local-desktop-20261012-2337-35b-q5km-2gpu-256k-llamacpp-f16 |
| Q4_K_M | 1 · single | short | 117.69 tok/s | 2733.9 tok/s | 20.84 GB | 21.43 GB | 33.74 GB | local-desktop-20261012-2341-35b-q4km-1gpu-4k-llamacpp-f16 |
| Q4_K_M | 1 · single | 32K | 101.88 tok/s | 1902.1 tok/s | 21.55 GB | 22.17 GB | 34.58 GB | local-desktop-20261012-2342-35b-q4km-1gpu-32k-llamacpp-f16 |
| Q4_K_M | 1 · single | 128K | 72.48 tok/s | 942.1 tok/s | 23.61 GB | 24.27 GB | 36.5 GB | local-desktop-20261012-2342-35b-q4km-1gpu-128k-llamacpp-f16 |
| Q4_K_M | 1 · single | 256K | 47.77 tok/s | 558.5 tok/s | 26.36 GB | 27.03 GB | 39.05 GB | local-desktop-20261012-2344-35b-q4km-1gpu-256k-llamacpp-f16 |
| Q4_K_M | 2 · split | short | 93.28 tok/s | 2660.4 tok/s | 10.83 / 10.85 GB | 21.69 GB | 33.74 GB | local-desktop-20261012-2349-35b-q4km-2gpu-4k-llamacpp-f16 |
| Q4_K_M | 2 · split | 32K | 82.57 tok/s | 1933.1 tok/s | 11.34 / 11.39 GB | 22.72 GB | 34.39 GB | local-desktop-20261012-2349-35b-q4km-2gpu-32k-llamacpp-f16 |
| Q4_K_M | 2 · split | 128K | 62.12 tok/s | 1078.4 tok/s | 13.03 / 13.23 GB | 26.26 GB | 38.04 GB | local-desktop-20261012-2349-35b-q4km-2gpu-128k-llamacpp-f16 |
| Q4_K_M | 2 · split | 256K | 42.88 tok/s | 673 tok/s | 15.28 / 15.74 GB | 31.02 GB | 40.84 GB | local-desktop-20261012-2351-35b-q4km-2gpu-256k-llamacpp-f16 |
| Q6_K | 1 · single | short | 103.23 tok/s | 2623.6 tok/s | 28.78 GB | 29.26 GB | 40.15 GB | local-desktop-20261012-2355-35b-q6k-1gpu-4k-llamacpp-f16 |
| Q6_K | 1 · single | 32K | 71.19 tok/s | 1683 tok/s | 28.68 GB | 29.17 GB | 44.01 GB | local-desktop-20261012-2356-35b-q6k-1gpu-32k-llamacpp-f16 Partly offloaded to DDR5 on 1 GPU |
| Q6_K | 1 · single | 128K | 54.51 tok/s | 800.3 tok/s | 28.78 GB | 29.25 GB | 45.43 GB | local-desktop-20261012-2356-35b-q6k-1gpu-128k-llamacpp-f16 Partly offloaded to DDR5 on 1 GPU |
| Q6_K | 1 · single | 256K | 37.78 tok/s | 471.6 tok/s | 28.81 GB | 29.3 GB | 43.69 GB | local-desktop-20261012-2358-35b-q6k-1gpu-256k-llamacpp-f16 Partly offloaded to DDR5 on 1 GPU |
| Q6_K | 2 · split | short | 83.5 tok/s | 2572.1 tok/s | 15.01 / 14.63 GB | 29.64 GB | 38.22 GB | local-desktop-20261013-0005-35b-q6k-2gpu-4k-llamacpp-f16 |
| Q6_K | 2 · split | 32K | 75.43 tok/s | 1890.7 tok/s | 15.5 / 15.16 GB | 30.66 GB | 38.98 GB | local-desktop-20261013-0005-35b-q6k-2gpu-32k-llamacpp-f16 Partly offloaded to DDR5 on 1 GPU |
| Q6_K | 2 · split | 128K | 58.04 tok/s | 1059.7 tok/s | 17.21 / 17 GB | 34.21 GB | 41.41 GB | local-desktop-20261013-0005-35b-q6k-2gpu-128k-llamacpp-f16 Partly offloaded to DDR5 on 1 GPU |
| Q6_K | 2 · split | 256K | 42.2 tok/s | 666.2 tok/s | 19.46 / 19.51 GB | 38.96 GB | 45.15 GB | local-desktop-20261013-0007-35b-q6k-2gpu-256k-llamacpp-f16 Partly offloaded to DDR5 on 1 GPU |
| UD-Q3_K_XL estimate | 1 · single | short | no data — contribute | no data — contribute | 30.7 GB (estimate) | 30.7 GB (estimate) | 32 GB kit (estimate) | Calculator Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB). Q3 is an option labeled estimate — not a local desktop run. |
| UD-Q3_K_XL estimate | 1 · single | 32K | no data — contribute | no data — contribute | 30.7 GB (estimate) | 30.7 GB (estimate) | 32 GB kit (estimate) | Calculator Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB). Q3 is an option labeled estimate — not a local desktop run. |
| UD-Q3_K_XL estimate | 1 · single | 128K | no data — contribute | no data — contribute | 30.7 GB (estimate) | 30.7 GB (estimate) | 32 GB kit (estimate) | Calculator Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB). Q3 is an option labeled estimate — not a local desktop run. |
| UD-Q3_K_XL estimate | 1 · single | 256K | no data — contribute | no data — contribute | 30.7 GB (estimate) | 30.7 GB (estimate) | 32 GB kit (estimate) | Calculator Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB). Q3 is an option labeled estimate — not a local desktop run. |
| UD-Q3_K_XL estimate | 2 · split | short | no data — contribute | no data — contribute | 15.3 GB (estimate) | 30.7 GB (estimate) | 32 GB kit (estimate) | Calculator Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB). Q3 is an option labeled estimate — not a local desktop run. |
| UD-Q3_K_XL estimate | 2 · split | 32K | no data — contribute | no data — contribute | 15.3 GB (estimate) | 30.7 GB (estimate) | 32 GB kit (estimate) | Calculator Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB). Q3 is an option labeled estimate — not a local desktop run. |
| UD-Q3_K_XL estimate | 2 · split | 128K | no data — contribute | no data — contribute | 15.3 GB (estimate) | 30.7 GB (estimate) | 32 GB kit (estimate) | Calculator Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB). Q3 is an option labeled estimate — not a local desktop run. |
| UD-Q3_K_XL estimate | 2 · split | 256K | no data — contribute | no data — contribute | 15.3 GB (estimate) | 30.7 GB (estimate) | 32 GB kit (estimate) | Calculator Hugging Face listing for Qwen3.6-35B-A3B-UD-Q3_K_XL.gguf (16.8 GB). Q3 is an option labeled estimate — not a local desktop run. |
llama.cpp first. Strata only where this site already documents it. Flags marked on the recipe stay shape-only. Recommended context is 256K.
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 .
llama.cpp · UD-Q5_K_XL · 256K
llama.cpp · Vulkan flags not recorded
# Shape only — exact flags TODO on stacks.
./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 · UD-Q5_K_XL · 32K
llama.cpp · Vulkan flags not recorded
# Shape only — exact flags TODO on stacks.
./build-vk/bin/llama-server \
-m ~/models/Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf \
-ngl 99 \
-c 32768 \
--cache-type-k f16 \
--cache-type-v f16 \
--host 127.0.0.1 --port 8080
llama.cpp · Q4_K_M · short
llama.cpp · Vulkan flags not recorded
# Shape only — exact flags TODO on stacks.
./build-vk/bin/llama-server \
-m ~/models/Qwen3.6-35B-A3B-Q4_K_M.gguf \
-ngl 99 \
-c 4096 \
--cache-type-k f16 \
--cache-type-v f16 \
--host 127.0.0.1 --port 8080
llama.cpp · Q5_K_M · short
llama.cpp · Vulkan flags not recorded
# Shape only — exact flags TODO on stacks.
./build-vk/bin/llama-server \
-m ~/models/Qwen3.6-35B-A3B-Q5_K_M.gguf \
-ngl 99 \
-c 4096 \
--cache-type-k f16 \
--cache-type-v f16 \
--host 127.0.0.1 --port 8080
llama.cpp · UD-Q5_K_XL · short
llama.cpp · Vulkan flags not recorded
# Shape only — exact flags TODO on stacks.
./build-vk/bin/llama-server \
-m ~/models/Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf \
-ngl 99 \
-c 4096 \
--cache-type-k f16 \
--cache-type-v f16 \
--host 127.0.0.1 --port 8080
Calculator VRAM at the shopper quant, 1 GPU and 2 GPUs (split). The R9700 35B one-card badge is the measured card; other SKUs stay estimates. Speed is not invented here. Offload rows still show both RAM labels on this page.
| GPU | 1 GPU | 2 GPU |
|---|---|---|
| Can R9700 run Qwen3.6-35B-A3B? R9700 sheet · Measured card | Fits in VRAM (estimate) 25.7 GB measured card | Fits in VRAM (estimate) 25.7 GB estimate |
| Can PRO W7900 run Qwen3.6-35B-A3B? PRO W7900 sheet · Estimate | Fits in VRAM (estimate) 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can PRO W7800 run Qwen3.6-35B-A3B? PRO W7800 sheet · Estimate | Fits in VRAM (estimate) 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can PRO W7700 run Qwen3.6-35B-A3B? PRO W7700 sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can PRO W6800 run Qwen3.6-35B-A3B? PRO W6800 sheet · Estimate | Fits in VRAM (estimate) 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 7900 XTX run Qwen3.6-35B-A3B? RX 7900 XTX sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 7900 XT run Qwen3.6-35B-A3B? RX 7900 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 9070 XT run Qwen3.6-35B-A3B? RX 9070 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 9070 run Qwen3.6-35B-A3B? RX 9070 sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 9070 GRE run Qwen3.6-35B-A3B? RX 9070 GRE sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Does not fit this GPU 25.7 GB estimate |
| Can RX 9060 XT run Qwen3.6-35B-A3B? RX 9060 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 7900 GRE run Qwen3.6-35B-A3B? RX 7900 GRE sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 7800 XT run Qwen3.6-35B-A3B? RX 7800 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 7700 XT run Qwen3.6-35B-A3B? RX 7700 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Does not fit this GPU 25.7 GB estimate |
| Can RX 7600 XT run Qwen3.6-35B-A3B? RX 7600 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 7600 run Qwen3.6-35B-A3B? RX 7600 sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Does not fit this GPU 25.7 GB estimate |
| Can RX 6800 XT run Qwen3.6-35B-A3B? RX 6800 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 6950 XT run Qwen3.6-35B-A3B? RX 6950 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
| Can RX 6700 XT run Qwen3.6-35B-A3B? RX 6700 XT sheet · Estimate | Does not fit this GPU 25.7 GB estimate | Does not fit this GPU 25.7 GB estimate |
| Can Strix Halo run Qwen3.6-35B-A3B? Strix Halo sheet · Estimate | Fits in VRAM (estimate) 25.7 GB estimate | Fits in VRAM (estimate) 25.7 GB estimate |
Head-to-head pages use the same sizes, cited context, and calculator fit already on the model pages. Q3 is a labeled estimate. No invented prices. Qwen3.6-35B against the other cited /models pages:
Also: Qwen3.8 27B
| Model | Setup | Backend | Quant | Ctx | Decode | Evidence |
|---|---|---|---|---|---|---|
| Qwen3.6-35B-A3B Qwen3.6-35B-A3B-Q4_K_M.gguf | 1× r9700 | llama.cpp vulkan | Q4_K_M | short | 117.69 | Tested on our rig |
| Qwen3.6-35B-A3B Qwen3.6-35B-A3B-Q5_K_M.gguf | 1× r9700 | llama.cpp vulkan | Q5_K_M | short | 109.73 | Tested on our rig |
| Qwen3.6-35B-A3B Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf | 1× r9700 | llama.cpp vulkan | UD-Q5_K_XL | 256K | 44.4 | Tested on our rig |
| Qwen3.6-35B-A3B Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf | 1× r9700 | llama.cpp vulkan | UD-Q5_K_XL | 32K | 90.52 | Tested on our rig |
| Qwen3.6-35B-A3B Qwen3.6-35B-A3B-UD-Q5_K_XL.gguf | 1× r9700 | llama.cpp vulkan | UD-Q5_K_XL | short | 102.39 | Tested on our rig |
Measured cells come from verified amdai-bench/run@1 cards credited to the submitter.
amdai-bench/run@1 · single · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:11:50-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 32K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 32,768 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:12:00-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 128K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 131,072 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:12:23-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 256K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:14:07-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:19:38-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 32K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 32,768 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:19:48-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 128K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 131,072 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:20:12-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 256K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:21:50-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:26:34-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes. · Q6_K single-GPU 4K table cell is 103.23 tok/s from the 23:55 card (the 4K point in the complete Q6_K single-GPU series). The earlier 23:26 card at 82.62 tok/s stays on /results and is not averaged into the cell.
amdai-bench/run@1 · single · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:27:26-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 32K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 32,768 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:27:39-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 128K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 131,072 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:28:02-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 256K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:29:46-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:34:56-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 32K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 32,768 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:35:07-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 128K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 131,072 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:35:31-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 256K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:37:08-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:41:50-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 32K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 32,768 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:42:03-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 128K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 131,072 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:42:26-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · 256K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:44:09-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:49:02-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 32K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 32,768 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:49:11-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 128K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 131,072 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:49:35-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 256K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:51:11-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · single · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:55:52-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes. · Q6_K single-GPU 4K table cell is 103.23 tok/s from the 23:55 card (the 4K point in the complete Q6_K single-GPU series). The earlier 23:26 card at 82.62 tok/s stays on /results and is not averaged into the cell.
amdai-bench/run@1 · single · 32K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 32,768 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:56:08-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes. · Partly offloaded to DDR5 on one GPU (llama.cpp --fit-target 1024). Decode is not an in-VRAM figure.
amdai-bench/run@1 · single · 128K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 131,072 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:56:39-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes. · Partly offloaded to DDR5 on one GPU (llama.cpp --fit-target 1024). Decode is not an in-VRAM figure.
amdai-bench/run@1 · single · 256K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 1 GPU · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-12T23:58:54-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes. · Partly offloaded to DDR5 on one GPU (llama.cpp --fit-target 1024). Decode is not an in-VRAM figure.
amdai-bench/run@1 · split · short
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-13T00:05:21-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 32K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 32,768 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-13T00:05:32-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 128K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 131,072 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-13T00:05:58-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · split · 256K
llama.cpp llama.cpp c060ca974 (build 10603, 2026-08-23) Vulkan; cmake Release GGML_VULKAN=ON GGML_NATIVE=ON GGML_OPENMP=ON · 2 GPUs · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10-13T00:07:38-05:00
VRAM sampled every 1 s via amdgpu mem_info_vram_used; Vulkan0 (display) includes desktop baseline ~0.5 GB · DDR5 = MemTotal-MemAvailable + resident page cache of the model files (mincore), 1 s samples; GB = GiB · RAM at DDR5-5200: one transient bad read in 10 earlier sha256 passes; sanity prompt cross-checked · Measured, pending re-verify after RAM fix. DDR5-5200 later proved unstable; a kernel oops at 00:14 CT aborted the sweep. All 33 cards predate the oops. Each card already notes one transient bad read in earlier sha256 passes.
amdai-bench/run@1 · dual-instance · 256K
llama.cpp · 2 GPUs · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10 (date not recorded)
Measured on our rig, 2026-10. Two copies (one per card). About 100 tok/s per request.
amdai-bench/run@1 · dual-instance · 256K
llama.cpp · 2 GPUs · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10 (date not recorded)
Measured on our rig, 2026-10. Two concurrent requests across two copies. 172 tok/s total.
amdai-bench/run@1 · dual-instance · 256K
llama.cpp · 2 GPUs · 262,144 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10 (date not recorded)
Measured on our rig, 2026-10. 4–8 concurrent, about 177 tok/s total. Soak: 423 requests, 20.6 min, 0 errors. Peak VRAM 27.7 / 29.7 GB (Linux GiB labeled GB). DDR5 about 13 GB.
amdai-bench/run@1 · single · short
vLLM stilldeadcode/vllm-radiance:0.5.8 · 1 GPU · 4,096 tokens
Credited to AMDai @amdai · local-desktop verified submitter
2026-10 (date not recorded)
Measured on our rig, 2026-10. Band only — no single decode_tps. Image stilldeadcode/vllm-radiance:0.5.8. Exact serve flags not recorded.