Skip to content

Gemma-4 31B-it

Q4_K_M·31B params·GGUF
checkpoint: unsloth/gemma-4-31B-it-GGUF:Q4_K_M

All runs (10)

legacystack comparable
GeForce RTX 3090 · 24 GiBcap 300 Wdrv 590
llama.cpp cuda-4f13cb7 (cuda)baselinechat1
37.5
34.3132.5264ms26.7361002.91s0.000 GiB
legacystack comparable
GeForce RTX 3090 · 24 GiBcap 300 Wdrv 590
llama.cpp cuda-4f13cb7 (cuda)baselinecodegen1
36.5
34.990.7847ms27.471100028.68s0.000 GiB
legacystack comparable
GeForce RTX 3090 · 24 GiBcap 300 Wdrv 590
llama.cpp cuda-4f13cb7 (cuda)baselineagent1
36.2
31.4275.81.90s27.761850015.93s0.000 GiB
legacystack comparable
GeForce RTX 3090 · 24 GiBcap 300 Wdrv 590
llama.cpp cuda-4f13cb7 (cuda)baselineagent4
36.1
13.329.724.88s27.761850038.84s0.000 GiB
legacystack comparable
GeForce RTX 3090 · 24 GiBcap 300 Wdrv 590
llama.cpp cuda-4f13cb7 (cuda)baselinerag1
36.1
24.5259.72.43s27.78532008.16s0.000 GiB
legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinechat1
10.6
9.9740ms94.4979.80s0.002 GiB
legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinecodegen1
10.3
10.2989ms97.099697.56s0.003 GiB
legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent1
10.2
9.63.26s98.549751.86s0.010 GiB
legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselinerag1
10.0
9.12.34s99.519721.55s0.007 GiB
legacystack comparable
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)unified
llama.cpp b1203 (rocm)baselineagent4
3.3
3.114.37s306.3497162.75s0.016 GiB

Environment

GeForce RTX 3090 · 24 GiB
cpuAMD EPYC 7302P 16-Core Processor
gpuNVIDIA GeForce RTX 3090
archNVIDIA
vram24 GiB (system 64.0 GiB)
power300 W / 450 W max(67% cap)
pcieGen 4 x16 / Gen 4 x16 max
clocksgfx 1920/2100 MHz · mem 9501 MHz
temp53°C idle · 68°C peak
peak draw287 W
hardware probes
copy 42% of theoryFP16 peak 65.4 TFcopy/math flat across caps
384-bit9751 MHz82 SM/CU
Microbenchmarks for memory copy and tensor math; raw-engine decode and API workload rows measure model-serving speed.
captheorycopyfp16bf16
200 W936 GB/s391 GB/s65.4 TF65.4 TF
300 W936 GB/s391 GB/s65.4 TF65.3 TF
450 W936 GB/s391 GB/s65.4 TF65.4 TF
compute: 8.6
backendllama.cpp cuda-4f13cb7 (cuda)
osUbuntu 24.04 LTS
kernel6.17.13-7-pve
driverNVIDIA 590.48.01 + CUDA 13.1
libc2.39
python3.12.3
llama.cppversion: 18 (4f13cb7) built with GNU 13.3.0 for Linux x86_64
build flagsGGML_CUDA=ON CMAKE_BUILD_TYPE=Release
runs/cell5
warmups2
endpoint/v1/chat/completions
streamingtrue
Strix Halo · Radeon 8060S · 128 GiB unified (96 GiB VRAM)
cpuAMD RYZEN AI MAX+ 395 w/ Radeon 8060S
gpuAMD Radeon 8060S
archStrix Halo (gfx1151)
vram96 GiB (system 31.1 GiB, unified)
hardware probes
copy 41% of theoryFP16 peak 30.3 TF
256-bit8000 MHz20 SM/CU
Microbenchmarks for memory copy and tensor math; raw-engine decode and API workload rows measure model-serving speed.
captheorycopyfp16bf16
fixed256 GB/s106 GB/s30.3 TF-
compute: 11.5
backendllama.cpp b1203 (rocm)
osUbuntu 24.04.4 LTS
kernel7.0.2-2-pve
python3.12.3
runs/cell3
warmups1
endpoint/v1/chat/completions
streamingtrue