Loongson 3A5000

Loongson 3A5000 (LoongArch LA464 core). 2.3 GHz, 4 cores, 12 nm. RAM: 16 GB DDR4.

L1 Data cache = 64 KB, 64 B/line, 4-WAY.
L1 Instruction cache = 64 KB, ? B/line, 4-WAY.
L2 Cache = 256 KB, 64 B/line, 16-WAY
L3 Cache = 16MB MB, 64 B/line, ?-WAY

2x 72-bit DDR4-3200 with ECC support.
2x 16-bit (4x 8-bit) HyperTransport of 3.2 GHz.
4x Decode.
4x ALU, 2x AGU, 2x VALU.
L3 Cache to core = 32B Read, 16B Write.
L3 Cache to DRAM = 16B Read, 16B Write.

L1 Data Cache Latency = 4 cycles for simple access via pointer
L1 Data Cache Latency = 6 cycles for access with complex address calculation (size_t n, *p; n = p[n]).
L2 Cache Latency = 14 cycles
L3 Cache Latency = 40 cycles
RAM Latency = 40 cycles + 87 ns

16 KB pages mode (64-bit Linux)

Data TLB L1: 64 items (cover 1 MB). full-assoc. Miss penalty = 6 cycles (L1 read), ? cycles (L2 read). Parallel miss: 1.55 cycles per access
Data TLB L2: 2048 items (cover 64 MB). 8-way. Miss penalty = 134 (L2 read). Parallel miss: 160 cycles per access

  Size        Latency       Increase   Description

  64 K     4                       4   
 128 K     9                       5   + 10 (L2)        
 256 K    12                       3
 512 K    24                      12   + 26 (L3)
   1 M    32                       8
   2 M    39                       7   + 6 (L1 TLB miss)        
   4 M    43                       4   
   8 M    44                       1
  16 M    46 +  1 ns            1
  32 M    46 + 47 ns         + 48 ns   + 87 ns (RAM)
  64 M    46 + 68 ns         + 20 ns   
 128 M   114 + 77 ns      68 +  9 ns   + 134 (TLB miss) 
 256 M   151 + 82 ns      37 +  5 ns
 512 M   176 + 84 ns      15 +  2 ns
1024 M   188 + 86 ns      12 +  2 ns
2048 M   197 + 87 ns       9 +  1 ns   +6
4096 M   202 + 87 ns       5      ns   +136 (TLB)

32 MB pages mode (Huge Pages) (64-bit Linux)

Data TLB L1: 64 items (cover 2 GB). full-assoc. Miss penalty = ? cycles (L1 read), Parallel miss: ? cycles per access
Data TLB L2: 2048 items.

  Size        Latency       Increase   Description

  64 K     4                       4   
 128 K     9                       5   + 10 (L2)        
 256 K    12                       3
 512 K    24                      12   + 26 (L3)
   1 M    32                       8
   2 M    36                       4
   4 M    38                       2   
   8 M    39                       1
  16 M    40                       1
  32 M    40 + 48 ns         + 48 ns   + 87 ns (RAM)
  64 M    40 + 68 ns         + 20 ns   
 128 M    40 + 77 ns         +  9 ns   
 256 M    40 + 82 ns         +  5 ns
 512 M    40 + 84 ns          + 2 ns
1024 M    40 + 86 ns          + 2 ns
2048 M    43 + 87 ns        3 + 1 ns   +6
4096 M    43 + 87 ns        69    ns   +134 (TLB miss)

MISC

Branch misprediction penalty = 13-20 cycles

64-bytes range cross penalty = 10 cycles

L1 B/W (Parallel Random Read) = 0.5 cycle per one access

L2->L1 B/W (Parallel Random Read) = 2.75 cycles per cache line
L2->L1 B/W (Read, 64 bytes step) = 2.5 cycles per cache line
L2->L1 B/W (Read, 64 Bytes step - pointer chasing) = 14 cycles per cache line (no HW prefetch L2->L1)
L2 Write (Write, 64 bytes step) = 2.5 cycles per cache line

L3->L1 B/W (Parallel Random Read) = 4.33 cycles per cache line
L3->L1 B/W (Read, 64 bytes step) = 4.65 cycles per cache line
L3->L1 B/W (Read, 64 Bytes step - pointer chasing) = 14 cycles per cache line (HW prefetch)
L3 Write (Write, 64 bytes step) = 5.0 cycles per cache line

RAM Read B/W (Parallel Random Read) = 9 ns / access line (64 bytes)
RAM Read B/W (Read, 8 Bytes step) = 12 GB/s
RAM Read B/W (Read, 64 Bytes step - pointer chasing) = 9 GB/s (HW prefetch)
RAM Write B/W (Write, 8 Bytes step) = 18 GB/s
RAM Write B/W (Write, 64 Bytes step) = 5.5 GB/s

7-Zip Benchmark


7-Zip (z) 21.06 (LE) : Copyright (c) 1999-2021 Igor Pavlov : 2021-11-24
 64-bit locale=en_US.UTF-8 Threads:4

 mt=*
Compiler: 12.0.0 20211202 (experimental) GCC 12.0.0
Linux : 5.15.0+ : #348 SMP PREEMPT Thu Dec 30 13:46:50 CST 2021 : loongarch64
PageSize:16KB THP:always hwcap:3FE
LE 

1T CPU Freq (MHz):  2297  2298  2298  2298  2298  2298  2298
2T CPU Freq (MHz): 202% 2297   200% 2298  

RAM size:   16152 MB,  # CPU hardware threads:   4
RAM usage:    437 MB,  # Benchmark threads:      1

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:       4049   100   3955   3940  |      30394   100   2595   2595
23:       3656   100   3739   3726  |      30222   100   2615   2616
24:       3418   100   3677   3676  |      29998   100   2638   2634
25:       3296   100   3764   3764  |      29650   100   2642   2639
----------------------------------  | ------------------------------
Avr:      3605   100   3783   3776  |      30066   100   2623   2621
Tot:             100   3203   3199


RAM usage:    444 MB,  # Benchmark threads:      2

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:       6698   157   4144   6517  |      60722   200   2595   5185
23:       6614   167   4044   6739  |      60340   200   2615   5223
24:       6570   171   4142   7064  |      59713   200   2621   5242
25:       6487   171   4324   7408  |      59088   200   2630   5259
----------------------------------  | ------------------------------
Avr:      6592   166   4163   6932  |      59966   200   2615   5227
Tot:             183   3389   6080


RAM usage:    889 MB,  # Benchmark threads:      4

                       Compressing  |                  Decompressing
Dict     Speed Usage    R/U Rating  |      Speed Usage    R/U Rating
         KiB/s     %   MIPS   MIPS  |      KiB/s     %   MIPS   MIPS

22:      11543   290   3873  11229  |     121164   400   2587  10337
23:      12771   337   3859  13013  |     118560   395   2599  10259
24:      12760   348   3939  13720  |     117885   398   2600  10345
25:      12072   338   4079  13783  |     116626   400   2597  10380
----------------------------------  | ------------------------------
Avr:     12286   328   3937  12936  |     118559   398   2596  10330
Tot:             363   3266  11633



7-Zip (z) 21.06 (LE) : Copyright (c) 1999-2021 Igor Pavlov : 2021-11-24
 64-bit locale=en_US.UTF-8 Threads:4

 mt=* m=* freq=2300
freq= 2300
Compiler: 12.0.0 20211202 (experimental) GCC 12.0.0
Linux : 5.15.0+ : #348 SMP PREEMPT Thu Dec 30 13:46:50 CST 2021 : loongarch64
PageSize:16KB THP:always hwcap:3FE
LE 

1T CPU Freq (MHz):  2297  2298  2298  2298  2298  2298  2298
2T CPU Freq (MHz): 200% 2298   200% 2297  

RAM size:   16152 MB,  # CPU hardware threads:   4
RAM usage:    227 MB,  # Benchmark threads:      1


Method           Speed Usage    R/U Rating   E/U Effec
                 KiB/s     %   MIPS   MIPS     %     %

CPU                      100   2297   2299   100   100
CPU                      100   2300   2299   100   100
CPU                      100   2300   2299   100   100

LZMA:x1          13603   100   5012   5015   218   218
                 30641   100   2451   2451   107   107
LZMA:x3           7589   100   4660   4663   203   203
                 29826   100   2504   2503   109   109
LZMA:x5:mt1       3427   100   4282   4282   186   186
                 30052   100   2536   2535   110   110
LZMA:x5:mt2       6598   170   4839   8243   210   358
                 30049   100   2536   2535   110   110
Deflate:x1       32115   100   4079   4078   177   177
                100939   100   3140   3136   137   136
Deflate:x5        9338   100   3594   3596   156   156
                101152   100   3137   3140   136   137
Deflate:x7        3315   100   3675   3673   160   160
                101932   100   3166   3163   138   138
Deflate64:x5      8707   100   3766   3763   164   164
                101116   100   3161   3163   137   138
BZip2:x1          4749   100   2869   2869   125   125
                 27398   100   2977   2970   129   129
BZip2:x5          4267   100   3561   3561   155   155
                 22476   100   4417   4411   192   192
BZip2:x5:mt2      8222   193   3552   6862   154   298
                 31933   142   4405   6267   192   272
BZip2:x7          1283   100   3324   3325   145   145
                 22659   100   4438   4444   193   193
PPMD:x1           3995   100   4133   4133   180   180
                  3178   100   3739   3743   163   163
PPMD:x5           2848   100   4821   4827   210   210
                  2362   100   4437   4427   193   192
Delta:4        1292443   100   3968   3970   173   173
               1070383   100   4383   4384   191   191
BCJ             617288   100   2530   2528   110   110
                600977   100   2461   2462   107   107
AES256CBC:1      95202   100   2341   2340   102   102
                 95100   100   2337   2337   102   102
AES256CBC:2 

AES256CBC:3 

CRC32:1         245401   100   1786   1787    78    78
CRC32:4         718778   100   1604   1604    70    70
CRC32:8        1259697   100   1708   1708    74    74
CRC32:32    
CRC32:64    
CRC64           684062   100   1401   1401    61    61
SHA256:1        169912   100   3464   3466   151   151
SHA256:2    
SHA1:1          317747   100   2977   2974   129   129
SHA1:2      
BLAKE2sp        200716   100   4416   4416   192   192

CPU                      100   2297   2299   100   100
------------------------------------------------------
Tot:                     110   3488   3937   156   171

Links

Loongson 3A5000/3B5000 Processor Reference Manual