ヒートシンクを取り付けると何か変わる？
CPUの保護機能による処理能力低下を防止することができる可能性があります。
ヒートシンクは半導体の放熱を助けます。ヒートシンクを半導体に取り付けることにより、半導体の放熱特性を高めることができます。CPUは計算負荷を高くすると発熱します。あまり発熱するとCPU自身が壊れてしまうため、保護機能が働き処理能力を落とします。
従って、保護機能が動作するような場合、ヒートシンクを取り付けることにより、処理能力を落とさずに計算をすることができるかも知れません。
（１）試験環境
H/W : Raspberry Pi3 Model B+ S/W : Raspbian (4.14.98-v7+ #1200 SMP Tue Feb 12 20:27:48 GMT 2019 armv7l) 電源 : Anker PowerPort Speed 5 (5V, 2.4A)
（２）計算負荷
"unixbench"を使用しました。
Version 5.1.3 Multi-CPU version January 13, 2011
- インストール＆実行手順
$ git clone https://github.com/kdlucas/byte-unixbench $ cd byte-unixbench/UnixBench/ $ ./Run"git"がない場合は"apt install git"でインストールする。
（３）結果
ヒートシンク取り付け有無ではたいして変わりませんでした。
ヒートシンクを取り付けると確かに温度上昇が鈍くなるのですが、高負荷が続くと結局のところ温度が上がってしまい保護機能が動作してしまうようです。そのためヒートシンクなしに比べてわずかですが、スコアがよくなっています。
そこでヒートシンクに風を当ててみたところ、4並列の場合かなり改善しました。1並列が改善しないのは、保護機能が動作するほどの高負荷ではなかったと思われます。
（４）まとめ
全CPUコアをフルに動作させるような重い計算を長時間を行う場合は、ヒートシンクを取り付けて強制空冷すると効果あり。ヒートシンクのみ（無風）はあまり効果がない。
（５）結果詳細
上表の棒グラフは、項目毎の増減を示している。「ヒートシンクあり＋強制空冷」はどの項目にも効果があるようです。
- ヒートシンクなし
========================================================================
BYTE UNIX Benchmarks (Version 5.1.3)
System: XXXX: GNU/Linux
OS: GNU/Linux -- 4.14.98-v7+ -- #1200 SMP Tue Feb 12 20:27:48 GMT 2019
Machine: armv7l (unknown)
Language: en_US.utf8 (charmap="ANSI_X3.4-1968", collate="ANSI_X3.4-1968")
CPU 0: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 1: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 2: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 3: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
15:34:50 up 1:42, 2 users, load average: 0.00, 0.00, 0.64; runlevel Mar
------------------------------------------------------------------------
Benchmark Run: 金 5月 03 2019 15:34:50 - 16:02:57
4 CPUs in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 5098208.2 lps (10.0 s, 7 samples)
Double-Precision Whetstone 1232.0 MWIPS (9.8 s, 7 samples)
Execl Throughput 1129.2 lps (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 156299.0 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 44759.6 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 382678.9 KBps (30.0 s, 2 samples)
Pipe Throughput 330624.4 lps (10.0 s, 7 samples)
Pipe-based Context Switching 69988.3 lps (10.0 s, 7 samples)
Process Creation 2855.0 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 2533.1 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 585.0 lpm (60.1 s, 2 samples)
System Call Overhead 702698.6 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 5098208.2 436.9
Double-Precision Whetstone 55.0 1232.0 224.0
Execl Throughput 43.0 1129.2 262.6
File Copy 1024 bufsize 2000 maxblocks 3960.0 156299.0 394.7
File Copy 256 bufsize 500 maxblocks 1655.0 44759.6 270.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 382678.9 659.8
Pipe Throughput 12440.0 330624.4 265.8
Pipe-based Context Switching 4000.0 69988.3 175.0
Process Creation 126.0 2855.0 226.6
Shell Scripts (1 concurrent) 42.4 2533.1 597.4
Shell Scripts (8 concurrent) 6.0 585.0 975.1
System Call Overhead 15000.0 702698.6 468.5
========
System Benchmarks Index Score 362.8
------------------------------------------------------------------------
Benchmark Run: 金 5月 03 2019 16:02:57 - 16:31:29
4 CPUs in system; running 4 parallel copies of tests
Dhrystone 2 using register variables 17411803.4 lps (10.0 s, 7 samples)
Double-Precision Whetstone 4207.0 MWIPS (9.9 s, 7 samples)
Execl Throughput 2357.3 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 224108.8 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 60076.9 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 554471.0 KBps (30.0 s, 2 samples)
Pipe Throughput 1128343.7 lps (10.0 s, 7 samples)
Pipe-based Context Switching 236336.8 lps (10.0 s, 7 samples)
Process Creation 5330.6 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 4803.0 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 664.3 lpm (60.2 s, 2 samples)
System Call Overhead 2330013.7 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 17411803.4 1492.0
Double-Precision Whetstone 55.0 4207.0 764.9
Execl Throughput 43.0 2357.3 548.2
File Copy 1024 bufsize 2000 maxblocks 3960.0 224108.8 565.9
File Copy 256 bufsize 500 maxblocks 1655.0 60076.9 363.0
File Copy 4096 bufsize 8000 maxblocks 5800.0 554471.0 956.0
Pipe Throughput 12440.0 1128343.7 907.0
Pipe-based Context Switching 4000.0 236336.8 590.8
Process Creation 126.0 5330.6 423.1
Shell Scripts (1 concurrent) 42.4 4803.0 1132.8
Shell Scripts (8 concurrent) 6.0 664.3 1107.1
System Call Overhead 15000.0 2330013.7 1553.3
========
System Benchmarks Index Score 784.4
- ヒートシンクあり
========================================================================
BYTE UNIX Benchmarks (Version 5.1.3)
System: XXXX: GNU/Linux
OS: GNU/Linux -- 4.14.98-v7+ -- #1200 SMP Tue Feb 12 20:27:48 GMT 2019
Machine: armv7l (unknown)
Language: en_US.utf8 (charmap="ANSI_X3.4-1968", collate="ANSI_X3.4-1968")
CPU 0: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 1: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 2: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 3: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
18:22:07 up 9 min, 2 users, load average: 0.00, 0.00, 0.00; runlevel May
------------------------------------------------------------------------
Benchmark Run: 金 5月 03 2019 18:22:07 - 18:50:43
4 CPUs in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 5094109.5 lps (10.0 s, 7 samples)
Double-Precision Whetstone 1229.4 MWIPS (9.8 s, 7 samples)
Execl Throughput 1138.1 lps (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 156185.2 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 44857.8 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 382938.0 KBps (30.0 s, 2 samples)
Pipe Throughput 329867.3 lps (10.0 s, 7 samples)
Pipe-based Context Switching 70211.5 lps (10.0 s, 7 samples)
Process Creation 2822.8 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 2546.4 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 618.0 lpm (60.1 s, 2 samples)
System Call Overhead 703036.0 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 5094109.5 436.5
Double-Precision Whetstone 55.0 1229.4 223.5
Execl Throughput 43.0 1138.1 264.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 156185.2 394.4
File Copy 256 bufsize 500 maxblocks 1655.0 44857.8 271.0
File Copy 4096 bufsize 8000 maxblocks 5800.0 382938.0 660.2
Pipe Throughput 12440.0 329867.3 265.2
Pipe-based Context Switching 4000.0 70211.5 175.5
Process Creation 126.0 2822.8 224.0
Shell Scripts (1 concurrent) 42.4 2546.4 600.6
Shell Scripts (8 concurrent) 6.0 618.0 1030.0
System Call Overhead 15000.0 703036.0 468.7
========
System Benchmarks Index Score 364.6
------------------------------------------------------------------------
Benchmark Run: 金 5月 03 2019 18:50:43 - 19:19:36
4 CPUs in system; running 4 parallel copies of tests
Dhrystone 2 using register variables 17417773.8 lps (10.0 s, 7 samples)
Double-Precision Whetstone 4206.2 MWIPS (9.8 s, 7 samples)
Execl Throughput 2372.8 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 231473.5 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 60338.2 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 556413.0 KBps (30.0 s, 2 samples)
Pipe Throughput 1128152.9 lps (10.0 s, 7 samples)
Pipe-based Context Switching 232249.2 lps (10.0 s, 7 samples)
Process Creation 5325.5 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 4798.5 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 665.0 lpm (60.3 s, 2 samples)
System Call Overhead 2328416.2 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 17417773.8 1492.5
Double-Precision Whetstone 55.0 4206.2 764.8
Execl Throughput 43.0 2372.8 551.8
File Copy 1024 bufsize 2000 maxblocks 3960.0 231473.5 584.5
File Copy 256 bufsize 500 maxblocks 1655.0 60338.2 364.6
File Copy 4096 bufsize 8000 maxblocks 5800.0 556413.0 959.3
Pipe Throughput 12440.0 1128152.9 906.9
Pipe-based Context Switching 4000.0 232249.2 580.6
Process Creation 126.0 5325.5 422.7
Shell Scripts (1 concurrent) 42.4 4798.5 1131.7
Shell Scripts (8 concurrent) 6.0 665.0 1108.3
System Call Overhead 15000.0 2328416.2 1552.3
========
System Benchmarks Index Score 786.2
- ヒートシンクあり＋強制空冷
========================================================================
BYTE UNIX Benchmarks (Version 5.1.3)
System: XXXX: GNU/Linux
OS: GNU/Linux -- 4.14.98-v7+ -- #1200 SMP Tue Feb 12 20:27:48 GMT 2019
Machine: armv7l (unknown)
Language: en_US.utf8 (charmap="ANSI_X3.4-1968", collate="ANSI_X3.4-1968")
CPU 0: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 1: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 2: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
CPU 3: ARMv7 Processor rev 4 (v7l) (0.0 bogomips)
19:25:50 up 1:13, 2 users, load average: 0.04, 3.88, 4.83; runlevel May
------------------------------------------------------------------------
Benchmark Run: 金 5月 03 2019 19:25:50 - 19:53:53
4 CPUs in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 5087638.3 lps (10.0 s, 7 samples)
Double-Precision Whetstone 1232.6 MWIPS (9.8 s, 7 samples)
Execl Throughput 1134.2 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 155559.8 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 44994.3 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 387643.5 KBps (30.0 s, 2 samples)
Pipe Throughput 329627.0 lps (10.0 s, 7 samples)
Pipe-based Context Switching 69850.0 lps (10.0 s, 7 samples)
Process Creation 2861.7 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 2548.7 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 690.2 lpm (60.1 s, 2 samples)
System Call Overhead 703275.3 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 5087638.3 436.0
Double-Precision Whetstone 55.0 1232.6 224.1
Execl Throughput 43.0 1134.2 263.8
File Copy 1024 bufsize 2000 maxblocks 3960.0 155559.8 392.8
File Copy 256 bufsize 500 maxblocks 1655.0 44994.3 271.9
File Copy 4096 bufsize 8000 maxblocks 5800.0 387643.5 668.4
Pipe Throughput 12440.0 329627.0 265.0
Pipe-based Context Switching 4000.0 69850.0 174.6
Process Creation 126.0 2861.7 227.1
Shell Scripts (1 concurrent) 42.4 2548.7 601.1
Shell Scripts (8 concurrent) 6.0 690.2 1150.3
System Call Overhead 15000.0 703275.3 468.9
========
System Benchmarks Index Score 368.5
------------------------------------------------------------------------
Benchmark Run: 金 5月 03 2019 19:53:53 - 20:22:29
4 CPUs in system; running 4 parallel copies of tests
Dhrystone 2 using register variables 20320487.3 lps (10.0 s, 7 samples)
Double-Precision Whetstone 4918.8 MWIPS (9.8 s, 7 samples)
Execl Throughput 2649.4 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 259254.7 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 69622.1 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 607469.5 KBps (30.0 s, 2 samples)
Pipe Throughput 1315613.3 lps (10.0 s, 7 samples)
Pipe-based Context Switching 276083.6 lps (10.0 s, 7 samples)
Process Creation 5979.0 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 5359.5 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 738.1 lpm (60.1 s, 2 samples)
System Call Overhead 2717452.9 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 20320487.3 1741.3
Double-Precision Whetstone 55.0 4918.8 894.3
Execl Throughput 43.0 2649.4 616.1
File Copy 1024 bufsize 2000 maxblocks 3960.0 259254.7 654.7
File Copy 256 bufsize 500 maxblocks 1655.0 69622.1 420.7
File Copy 4096 bufsize 8000 maxblocks 5800.0 607469.5 1047.4
Pipe Throughput 12440.0 1315613.3 1057.6
Pipe-based Context Switching 4000.0 276083.6 690.2
Process Creation 126.0 5979.0 474.5
Shell Scripts (1 concurrent) 42.4 5359.5 1264.0
Shell Scripts (8 concurrent) 6.0 738.1 1230.2
System Call Overhead 15000.0 2717452.9 1811.6
========
System Benchmarks Index Score 896.6
0 件のコメント:
コメントを投稿