SpacemiT K3を採用したBanana Pi BPI-SM10 (RAM8G)を手に入れた。関税込みで約5万5000円ほど
SpacemiT K3はおそらく市場で入手可能なものとしては初めてのRVA23準拠なSoC。Phoronixのレビューからは相当速そうなので、見てみる
Unix Bench
------------------------------------------------------------------------
Benchmark Run: Wed Jun 24 2026 16:30:33 - 16:58:43
16 CPUs in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 17624814.4 lps (10.0 s, 7 samples)
Double-Precision Whetstone 3512.4 MWIPS (10.0 s, 7 samples)
Execl Throughput 1514.2 lps (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 345067.8 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 109060.5 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 903275.2 KBps (30.0 s, 2 samples)
Pipe Throughput 711541.4 lps (10.0 s, 7 samples)
Pipe-based Context Switching 63801.2 lps (10.0 s, 7 samples)
Process Creation 3181.0 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 2040.2 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 1123.4 lpm (60.0 s, 2 samples)
System Call Overhead 647010.1 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 17624814.4 1510.3
Double-Precision Whetstone 55.0 3512.4 638.6
Execl Throughput 43.0 1514.2 352.1
File Copy 1024 bufsize 2000 maxblocks 3960.0 345067.8 871.4
File Copy 256 bufsize 500 maxblocks 1655.0 109060.5 659.0
File Copy 4096 bufsize 8000 maxblocks 5800.0 903275.2 1557.4
Pipe Throughput 12440.0 711541.4 572.0
Pipe-based Context Switching 4000.0 63801.2 159.5
Process Creation 126.0 3181.0 252.5
Shell Scripts (1 concurrent) 42.4 2040.2 481.2
Shell Scripts (8 concurrent) 6.0 1123.4 1872.3
System Call Overhead 15000.0 647010.1 431.3
========
System Benchmarks Index Score 611.2
------------------------------------------------------------------------
Benchmark Run: Wed Jun 24 2026 16:58:43 - 17:27:04
16 CPUs in system; running 16 parallel copies of tests
Dhrystone 2 using register variables 131766440.3 lps (10.0 s, 7 samples)
Double-Precision Whetstone 29711.5 MWIPS (9.2 s, 7 samples)
Execl Throughput 7837.1 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 2484861.8 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 837607.2 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 6202109.0 KBps (30.0 s, 2 samples)
Pipe Throughput 5564969.7 lps (10.0 s, 7 samples)
Pipe-based Context Switching 777388.2 lps (10.0 s, 7 samples)
Process Creation 17270.5 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 9524.0 lpm (60.1 s, 2 samples)
Shell Scripts (8 concurrent) 1222.7 lpm (60.2 s, 2 samples)
System Call Overhead 5086003.3 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 131766440.3 11291.0
Double-Precision Whetstone 55.0 29711.5 5402.1
Execl Throughput 43.0 7837.1 1822.6
File Copy 1024 bufsize 2000 maxblocks 3960.0 2484861.8 6274.9
File Copy 256 bufsize 500 maxblocks 1655.0 837607.2 5061.1
File Copy 4096 bufsize 8000 maxblocks 5800.0 6202109.0 10693.3
Pipe Throughput 12440.0 5564969.7 4473.4
Pipe-based Context Switching 4000.0 777388.2 1943.5
Process Creation 126.0 17270.5 1370.7
Shell Scripts (1 concurrent) 42.4 9524.0 2246.2
Shell Scripts (8 concurrent) 6.0 1222.7 2037.8
System Call Overhead 15000.0 5086003.3 3390.7
========
System Benchmarks Index Score 3718.2
コア自体は、SiHive P550の1.4倍くらいの速さ?
RISC-V Vector Crypto Extension
このSoCはRISC-V Vector Crypto Extensionが実装されているので、あり・なしでのベンチ (自分の都合上NSSを使ってます。RVV Cryptoのコードはまだ入れいていないけど)。AESだと5倍くらい速くて、SHA256だと2倍くらい。
NSS with RVV crypto extension
# mode in symmkey opreps cxreps context op time(sec) thrgput aes_ecb_e 4Gb 323Mb 128 289M 0 0.000 10000.000 10.000 441Mb 946Kb aes_cbc_e 5Gb 807Mb 128 194M 0 0.000 10000.000 10.000 592Mb 817Kb aes_ecb_d 4Gb 341Mb 128 290M 0 0.000 10000.000 10.000 443Mb 775Kb aes_cbc_d 5Gb 536Mb 128 185M 0 0.000 10000.000 10.000 565Mb 630Kb sha256_e 102Mb 265Kb 35M 0 0.000 10000.000 10.000 10Mb 231Kb
NSS without RVV crypto extension
# mode in symmkey opreps cxreps context op time(sec) thrgput aes_ecb_e 975Mb 395Kb 128 63M 0 0.000 10000.000 10.000 97Mb 551Kb aes_cbc_e 1Gb 16Mb 128 34M 0 0.000 10000.000 10.000 104Mb 92Kb aes_cbc_d 907Mb 193Kb 128 29M 0 0.000 10000.000 10.001 90Mb 726Kb aes_ecb_d 929Mb 629Kb 128 60M 0 0.000 10000.000 10.000 92Mb 984Kb sha256_e 53Mb 988Kb 18M 0 0.000 10000.000 10.001 5Mb 405Kb
今までのいろんなボードに比べれば、だいぶ速い印象
0 件のコメント:
コメントを投稿