Last Update 2026/02/07
低スペック寄りのPCでローカルLLMを動作させた際の記録です。
LLM以外の仮想マシンなどが起動され、多少負荷がかかった状態で実行しています。
ベンチマークなどでLLMの性能を評価する内容ではありません。
LLM以外の仮想マシンなどが起動され、多少負荷がかかった状態で実行しています。
ベンチマークなどでLLMの性能を評価する内容ではありません。
検証用PC
|
OS |
Debian GNU/Linux 12 (bookworm) |
|
CPU |
Intel(R) Core(TM) i5-14400F |
|
GPU |
GeForce RTX 3060 12GB |
|
メモリ |
DDR4 PC4-25600 32GB × 4 |
|
SSD |
crucial P310 CT1000P310SSD8-JP |
構築環境 : Docker + Ollama (特別な設定などは無い状態)
検証用プロンプト
おすすめの日本の絶景を教えてください。東西南北、10箇所程度。
Gemma 3 (QAT)
GPU無し 事前のモデルのロード無し GPU使用 事前のモデルのロード無し GPU使用 事前にモデルをロード済みTPS(tokens/s) は eval_count / eval_duration により算出
gemma3:1b-it-qat(GPU無し 事前のモデルのロード無し)
Model
parameters 999.89M
context length 32768
embedding length 1152
quantization Q4_0
2026-02-07
total_duration(合計時間) : 23114286785 (23.114s)
load_duration(モデルのロード時間) : 745412115 ( 0.745s)
prompt_eval_count(評価されたプロンプトのトークン数) : 27
prompt_eval_duration(プロンプトの評価時間) : 128995863 ( 0.013s)
eval_count(生成トークン数) : 808
eval_duration(生成時間) : 21869615838 (21.870s)
real 0m23.125s
user 0m0.026s
sys 0m0.012s
メモリ使用量(RSS) : 1735204 KB
gemma3:4b-it-qat(GPU無し 事前のモデルのロード無し)
Model
parameters 4.3B
context length 131072
embedding length 2560
quantization Q4_0
2026-02-07
total_duration(合計時間) : 73934311903 (73.934s)
load_duration(モデルのロード時間) : 2051269517 ( 2.051s)
prompt_eval_count(評価されたプロンプトのトークン数) : 27
prompt_eval_duration(プロンプトの評価時間) : 419686236 ( 0.420s)
eval_count(生成トークン数) : 851
eval_duration(生成時間) : 71080709852 (71.081s)
real 1m13.945s
user 0m0.022s
sys 0m0.016s
メモリ使用量(RSS) : 5672308 KB
gemma3:12b-it-qat(GPU無し 事前のモデルのロード無し)
Model
parameters 12.2B
context length 131072
embedding length 3840
quantization Q4_0
2026-02-07
total_duration(合計時間) : 134438344073 (134.438s)
load_duration(モデルのロード時間) : 3365638166 ( 3.366s)
prompt_eval_count(評価されたプロンプトのトークン数) : 43
prompt_eval_duration(プロンプトの評価時間) : 2025412282 ( 2.025s)
eval_count(生成トークン数) : 621
eval_duration(生成時間) : 128753220603 (128.753s)
real 2m14.450s
user 0m0.033s
sys 0m0.014s
メモリ使用量(RSS) : 11710796 KB
gemma3:27b-it-qat(GPU無し 事前のモデルのロード無し)
Model
parameters 27.4B
context length 131072
embedding length 5376
quantization Q4_0
2026-02-07
total_duration(合計時間) : 731943453087 (731.943s)
load_duration(モデルのロード時間) : 3128573447 ( 3.129s)
prompt_eval_count(評価されたプロンプトのトークン数) : 43
prompt_eval_duration(プロンプトの評価時間) : 4754642855 ( 4.755s)
eval_count(生成トークン数) : 1607
eval_duration(生成時間) : 723332071807 (723.332s)
real 12m11.955s
user 0m0.095s
sys 0m0.030s
メモリ使用量(RSS) : 21615816 KB
gemma3:1b-it-qat(GPU使用 事前のモデルのロード無し)
Model
parameters 999.89M
context length 32768
embedding length 1152
quantization Q4_0
2026-02-07
total_duration(合計時間) : 6168774421 (6.169s)
load_duration(モデルのロード時間) : 752172466 (0.752s)
prompt_eval_count(評価されたプロンプトのトークン数) : 27
prompt_eval_duration(プロンプトの評価時間) : 11596649 (0.012s)
eval_count(生成トークン数) : 959
eval_duration(生成時間) : 4978427788 (4.978s)
real 0m6.180s
user 0m0.031s
sys 0m0.000s
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 0% 46C P2 139W / 170W | 1579MiB / 12288MiB | 83% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1259 G /usr/lib/xorg/Xorg 107MiB |
| 0 N/A N/A 1934 G xfwm4 2MiB |
| 0 N/A N/A 2458 G /usr/bin/x-www-browser 200MiB |
| 0 N/A N/A 31063 C /usr/bin/ollama 1256MiB |
+---------------------------------------------------------------------------------------+
メモリ使用量(RSS) : 1073432 KB
gemma3:4b-it-qat(GPU使用 事前のモデルのロード無し)
Model
parameters 4.3B
context length 131072
embedding length 2560
quantization Q4_0
2026-02-07
total_duration(合計時間) : 12730883298 (12.731s)
load_duration(モデルのロード時間) : 1473597891 ( 1.474s)
prompt_eval_count(評価されたプロンプトのトークン数) : 27
prompt_eval_duration(プロンプトの評価時間) : 22769313 ( 0.023s)
eval_count(生成トークン数) : 917
eval_duration(生成時間) : 10828147333 (10.828s)
real 0m12.742s
user 0m0.020s
sys 0m0.010s
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 0% 53C P2 163W / 170W | 4831MiB / 12288MiB | 94% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1259 G /usr/lib/xorg/Xorg 107MiB |
| 0 N/A N/A 1934 G xfwm4 2MiB |
| 0 N/A N/A 2458 G /usr/bin/x-www-browser 200MiB |
| 0 N/A N/A 31136 C /usr/bin/ollama 4508MiB |
+---------------------------------------------------------------------------------------+
メモリ使用量(RSS) : 1978560 KB
gemma3:12b-it-qat(GPU使用 事前のモデルのロード無し)
Model
parameters 12.2B
context length 131072
embedding length 3840
quantization Q4_0
2026-02-07
total_duration(合計時間) : 25037799624 (25.038s)
load_duration(モデルのロード時間) : 2250079527 ( 2.250s)
prompt_eval_count(評価されたプロンプトのトークン数) : 43
prompt_eval_duration(プロンプトの評価時間) : 58829442 ( 0.059s)
eval_count(生成トークン数) : 808
eval_duration(生成時間) : 22361751819 (22.362s)
real 0m25.048s
user 0m0.031s
sys 0m0.000s
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 31% 60C P2 169W / 170W | 9983MiB / 12288MiB | 97% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1259 G /usr/lib/xorg/Xorg 107MiB |
| 0 N/A N/A 1934 G xfwm4 2MiB |
| 0 N/A N/A 2458 G /usr/bin/x-www-browser 200MiB |
| 0 N/A N/A 31235 C /usr/bin/ollama 9660MiB |
+---------------------------------------------------------------------------------------+
メモリ使用量(RSS) : 2711480 KB
gemma3:27b-it-qat(GPU使用 事前のモデルのロード無し)
Model
parameters 27.4B
context length 131072
embedding length 5376
quantization Q4_0
2026-02-07
total_duration(合計時間) : 129983417904 (129.983s)
load_duration(モデルのロード時間) : 3036630444 ( 3.037s)
prompt_eval_count(評価されたプロンプトのトークン数) : 43
prompt_eval_duration(プロンプトの評価時間) : 475172873 ( 0.475s)
eval_count(生成トークン数) : 638
eval_duration(生成時間) : 126192161008 (126.192s)
real 2m10.002s
user 0m0.060s
sys 0m0.009s
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 0% 54C P2 67W / 170W | 11736MiB / 12288MiB | 17% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1259 G /usr/lib/xorg/Xorg 107MiB |
| 0 N/A N/A 1934 G xfwm4 2MiB |
| 0 N/A N/A 2458 G /usr/bin/x-www-browser 201MiB |
| 0 N/A N/A 31328 C /usr/bin/ollama 11412MiB |
+---------------------------------------------------------------------------------------+
メモリ使用量(RSS) : 10938060 KB
gemma3:1b-it-qat(GPU使用 事前にモデルをロード済み)
Model
parameters 999.89M
context length 32768
embedding length 1152
quantization Q4_0
2026-02-07
total_duration(合計時間) : 4930318282 (4.930s)
load_duration(モデルのロード時間) : 73286852 (0.073s)
prompt_eval_count(評価されたプロンプトのトークン数) : 27
prompt_eval_duration(プロンプトの評価時間) : 13136396 (0.013s)
eval_count(生成トークン数) : 855
eval_duration(生成時間) : 4466034708 (4.466s)
real 0m4.941s
user 0m0.021s
sys 0m0.009s
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 0% 53C P2 143W / 170W | 1580MiB / 12288MiB | 87% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1259 G /usr/lib/xorg/Xorg 107MiB |
| 0 N/A N/A 1934 G xfwm4 2MiB |
| 0 N/A N/A 2458 G /usr/bin/x-www-browser 201MiB |
| 0 N/A N/A 37878 C /usr/bin/ollama 1256MiB |
+---------------------------------------------------------------------------------------+
メモリ使用量(RSS) : 1073776 KB
gemma3:4b-it-qat(GPU使用 事前にモデルをロード済み)
Model
parameters 4.3B
context length 131072
embedding length 2560
quantization Q4_0
2026-02-07
total_duration(合計時間) : 7888331877 (7.888s)
load_duration(モデルのロード時間) : 69904156 (0.070s)
prompt_eval_count(評価されたプロンプトのトークン数) : 27
prompt_eval_duration(プロンプトの評価時間) : 30720280 (0.031s)
eval_count(生成トークン数) : 631
eval_duration(生成時間) : 7507783869 (7.508s)
real 0m7.899s
user 0m0.031s
sys 0m0.001s
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 0% 58C P2 165W / 170W | 4832MiB / 12288MiB | 94% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1259 G /usr/lib/xorg/Xorg 107MiB |
| 0 N/A N/A 1934 G xfwm4 2MiB |
| 0 N/A N/A 2458 G /usr/bin/x-www-browser 201MiB |
| 0 N/A N/A 37968 C /usr/bin/ollama 4508MiB |
+---------------------------------------------------------------------------------------+
メモリ使用量(RSS) : 1969820 KB
gemma3:12b-it-qat(GPU使用 事前にモデルをロード済み)
Model
parameters 12.2B
context length 131072
embedding length 3840
quantization Q4_0
2026-02-07
total_duration(合計時間) : 27324603146 (27.325s)
load_duration(モデルのロード時間) : 67569592 ( 0.068s)
prompt_eval_count(評価されたプロンプトのトークン数) : 43
prompt_eval_duration(プロンプトの評価時間) : 60669586 ( 0.061s)
eval_count(生成トークン数) : 959
eval_duration(生成時間) : 26800558403 (26.801s)
real 0m27.335s
user 0m0.025s
sys 0m0.006s
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 36% 64C P2 169W / 170W | 9984MiB / 12288MiB | 97% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1259 G /usr/lib/xorg/Xorg 107MiB |
| 0 N/A N/A 1934 G xfwm4 2MiB |
| 0 N/A N/A 2458 G /usr/bin/x-www-browser 201MiB |
| 0 N/A N/A 38053 C /usr/bin/ollama 9660MiB |
+---------------------------------------------------------------------------------------+
メモリ使用量(RSS) : 2711660 KB
gemma3:27b-it-qat(GPU使用 事前にモデルをロード済み)
Model
parameters 27.4B
context length 131072
embedding length 5376
quantization Q4_0
2026-02-07
total_duration(合計時間) : 126546740980 (126.547s)
load_duration(モデルのロード時間) : 71230491 ( 0.071s)
prompt_eval_count(評価されたプロンプトのトークン数) : 43
prompt_eval_duration(プロンプトの評価時間) : 476466407 ( 0.476s)
eval_count(生成トークン数) : 637
eval_duration(生成時間) : 125730518973 (125.731s)
real 2m6.558s
user 0m0.032s
sys 0m0.012s
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.261.03 Driver Version: 535.261.03 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 33% 44C P2 66W / 170W | 11735MiB / 12288MiB | 18% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+
+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| 0 N/A N/A 1259 G /usr/lib/xorg/Xorg 107MiB |
| 0 N/A N/A 1934 G xfwm4 2MiB |
| 0 N/A N/A 2458 G /usr/bin/x-www-browser 200MiB |
| 0 N/A N/A 38148 C /usr/bin/ollama 11412MiB |
+---------------------------------------------------------------------------------------+
メモリ使用量(RSS) : 10929988 KB