Intel Skylake X

Intel i7-7820X (Skylake X), 8 cores, 4.3 GHz (Turbo Boost), Mesh 2.4 GHz, 14 nm. RAM: 4x 8 GB DDR4-3400 16-18-18-36.

• L1 Data cache = 32 KB, 64 B/line, 8-WAY
• L1 Instruction cache = 32 KB, 64 B/line, 8-WAY.
• L2 cache = 1024 KB, 64 B/line, 16-WAY
• L3 cache = 11 MB, 64 B/line, 11-WAY
• Snoop filter for L2 = 2048 sets, 12-ways snoop filter in each CORE TILE (it's cover 1.5 MB of L2 ?)

• L1 Data Cache Latency = 4 cycles for simple access via pointer
• L1 Data Cache Latency = 5 cycles for access with complex address calculation (size_t n, *p; n = p[n]).
• L2 Cache Latency = 14 cycles
• L3 Cache Latency = 68 cycles (3.6 GHz)
• L3 Cache Latency = 79 cycles (4.3 GHz) (77-81 cycles for different cores)
• RAM Latency = 79 cycles + 50 ns

Note: It's possible that L2 Cache Latency can be 13 cycles in some cases. But dependency chain workload shows 14 cycles.

Note: L3 Cache replacement policy doesn't look like LRU or Pseudo-LRU in tests. L3 Cache can keep some old lines for long time. It can be "Random Replacement" policy or some another combined scheme.

1 GB pages mode, Windows 10

• Data TLB: 4 entries, 4-way. Miss penalty = 9 cycles.
• L2 TLB: 16 entries, 4-way. Miss penalty = 9 cycles?

  Size       Latency        Increase   Description

  32 K     4                  
  64 K     9               5           + 10 (L2)        
 128 K    12               3  
 256 K    13               1
 512 K    14               1      
   1 M    14               0
   2 M    49              35           + 65 (L3) 
   4 M    64              15  
   8 M    72               8      
  16 M    76 + 17 ns       4 + 17 ns   + 50 ns (RAM)
  32 M    78 + 34 ns       2 + 17 ns
  64 M    79 + 42 ns       1 +  8 ns
 128 M    79 + 46 ns         +  4 ns
 256 M    79 + 48 ns         +  2 ns
 512 M    79 + 49 ns         +  1 ns
   1 G    79 + 50 ns         +  1 ns
   2 G    79 + 50 ns      
   4 G    79 + 50 ns      
   8 G    84 + 50 ns       5           + 9 (L1 TLB miss)
  16 G    86 + 50 ns       2
  32 G    92 + 50 ns       6           + 9 ? (L2 TLB miss)

2 MB pages mode, Windows 10

• Data TLB: 32 items (4-way). Miss Penalty = 9 cycles. Parallel miss: 1 cycles per access
• L2 TLB: 1536 items (12-way). Miss Penalty = ? cycles.

  Size       Latency        Increase   Description

  32 K     4                  
  64 K     9               5           + 10 (L2)        
 128 K    12               3  
 256 K    13               1
 512 K    14               1      
   1 M    14               0
   2 M    49              35           + 65 (L3) 
   4 M    64              15  
   8 M    72               8      
  16 M    76 + 17 ns       4 + 17 ns   + 50 ns (RAM)
  32 M    78 + 34 ns       2 + 17 ns
  64 M    79 + 42 ns       1 +  8 ns
 128 M    84 + 46 ns       5 +  4 ns   + 9 (L1 TLB miss)
 256 M    86 + 48 ns       2 +  2 ns
 512 M    87 + 49 ns       1 +  1 ns
   1 G    88 + 50 ns       1 +  1 ns

4 KB pages mode, Windows 10

• Data TLB L1: 64 items. 4-way. Miss penalty = 9 cycles. Parallel miss: 1 cycle per access
• Data TLB L2 (STLB): 1536 items. 12-way. Miss penalty = 17 cycles ?. Parallel miss: 17 cycles per access
• PDE cache = ? items. Miss penalty = ? cycles.

  Size       Latency        Increase   Description

  32 K     4                  
  64 K     9               5           + 10 (L2)        
 128 K    12               3  
 256 K    13               1
 512 K    18               5           + 9 (L1 TLB miss)
   1 M    22               4
   2 M    57              35           + 65 (L3) 
   4 M    72              15  
   8 M    88              16           + 17 (L2 TLB miss)
  16 M    95 + 17 ns       7 + 17 ns   + 50 ns (RAM)
  32 M   101 + 34 ns       6 + 17 ns
  64 M   104 + 42 ns       3 +  8 ns
 128 M   111 + 46 ns       7 +  4 ns
 256 M   126 + 48 ns      15 +  2 ns
 512 M   148 + 49 ns      22 +  1 ns   + 65 (TLB miss to L3 Cache) 
   1 G   162 + 50 ns      14 +  1 ns
   2 G   173 + 50 ns      11

MISC

• Branch misprediction penalty = 16.5 cycles average (if mOp cache hit).

• 64-bytes range cross penalty = 7 cycles
• 4096-bytes range cross - no additional penalty

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

• L2->L1 B/W (Parallel Random Read) = 1.5 - 2.0 cycles per cache line
• L2->L1 B/W (Read, 64 bytes step) = 1.2 - 1.9 cycles per cache line
• L2 Write (Write, 64 bytes step) = 3.2 cycles per write (cache line)
• L2 Read+Write (Parallel Random) = 3.5 cycles per read+write pair (cache line)

• L3->L1 B/W (Parallel Random Read) = 8.9 cycles per cache line (3600 MHz)
• L3->L1 B/W (Parallel Random Read) = 10.1 cycles per cache line (4300 MHz)
• L3->L1 B/W (Read, 64 bytes step) = 10.2 cycles per cache line (4300 MHz)
• L3 Write (Write, 64 bytes step) = 10.0 cycles per write (cache line) (4300 MHz)

• RAM Read B/W (Parallel Random Read) = 7.0 ns / cache line = 9 GB/s
• RAM Read B/W (Read, 16-64 Bytes step) = 17 GB/s
• RAM Write B/W (Write, 8-64 Bytes step) = 17 GB/s

7-Zip Benchmark

Notes:

7z b : MIPS values are normalized with Intel Core 2 cpu.

7z b -mm=* : MIPS and Effectiveness values are normalized with AMD K8 cpu.

7z b -mmt1 -stmC0

7-Zip 17.01 beta (x64) : Copyright (c) 1999-2017 Igor Pavlov : 2017-08-28

x64 6.5504 cpus:16 128T
CMPXCHG MMX SSE RDTSC PAE SSE2 NX SSE3 CMPXCHG16B XSAVE L2AT VIRT_FIRMWARE RDWRFSGSBASE FASTFAIL RDRAND RDTSCP
Intel(R) Core(TM) i7-7820X CPU @ 3.60GHz (50654)
CPU Freq:  4000  4000  4266  4000  4129  4063  4096  4376  4366

RAM size:   32446 MB,  # CPU hardware threads:   2 / 16 : C0 / FFFF
RAM usage:    435 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:       5745   100   5590   5590  |      53684   100   4584   4584
23:       5228   100   5330   5327  |      53064   100   4593   4593
24:       4873   100   5240   5240  |      52633   100   4621   4621
25:       4638   100   5296   5296  |      51752   100   4607   4606
----------------------------------  | ------------------------------
Avr:             100   5364   5363  |              100   4601   4601
Tot:             100   4983   4982


7z b -mmt2 -stmC0

7-Zip 17.01 beta (x64) : Copyright (c) 1999-2017 Igor Pavlov : 2017-08-28

x64 6.5504 cpus:16 128T
CMPXCHG MMX SSE RDTSC PAE SSE2 NX SSE3 CMPXCHG16B XSAVE L2AT VIRT_FIRMWARE RDWRFSGSBASE FASTFAIL RDRAND RDTSCP
Intel(R) Core(TM) i7-7820X CPU @ 3.60GHz (50654)
CPU Freq:  4000  4266  4000  4266  8000  4063  4096  4376  4231

RAM size:   32446 MB,  # CPU hardware threads:   2 / 16 : C0 / FFFF
RAM usage:    441 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:       7989   155   5013   7772  |      71098   200   3035   6070
23:       7658   168   4641   7803  |      70656   200   3058   6116
24:       7412   168   4756   7970  |      69964   200   3071   6142
25:       7246   168   4916   8273  |      69120   200   3076   6152
----------------------------------  | ------------------------------
Avr:             165   4831   7955  |              200   3060   6120
Tot:             182   3946   7037

7z b

7-Zip 17.01 beta (x64) : Copyright (c) 1999-2017 Igor Pavlov : 2017-08-28

x64 6.5504 cpus:16 128T
CMPXCHG MMX SSE RDTSC PAE SSE2 NX SSE3 CMPXCHG16B XSAVE L2AT VIRT_FIRMWARE RDWRFSGSBASE FASTFAIL RDRAND RDTSCP
Intel(R) Core(TM) i7-7820X CPU @ 3.60GHz (50654)
CPU Freq:  4266  4000  4000  4266  4129  5446  4096  4675  4366

RAM size:   32446 MB,  # CPU hardware threads:  16
RAM usage:   3530 MB,  # Benchmark threads:     16

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

22:      50297  1241   3944  48929  |     550825  1594   2948  46980
23:      45930  1253   3735  46798  |     545475  1596   2958  47195
24:      45032  1272   3808  48419  |     534713  1586   2960  46934
25:      43666  1326   3760  49857  |     525312  1589   2943  46751
----------------------------------  | ------------------------------
Avr:            1273   3812  48501  |             1591   2952  46965
Tot:            1432   3382  47733



7z b -mm=* -mmt1 -stmC0


7-Zip 17.01 beta (x64) : Copyright (c) 1999-2017 Igor Pavlov : 2017-08-28

x64 6.5504 cpus:16 128T
CMPXCHG MMX SSE RDTSC PAE SSE2 NX SSE3 CMPXCHG16B XSAVE L2AT VIRT_FIRMWARE RDWRFSGSBASE FASTFAIL RDRAND RDTSCP
Intel(R) Core(TM) i7-7820X CPU @ 3.60GHz (50654)
CPU Freq:  4266  4000  4000  4266  4129  5446  4096  4357  4231

RAM size:   32446 MB,  # CPU hardware threads:   2 / 16 : C0 / FFFF
RAM usage:    225 MB,  # Benchmark threads:      1


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

CPU                      100   4278   4278
CPU                      100   4295   4294
CPU                      100   4295   4296   100   100

LZMA:x1          22080   100   8070   8072   188   188
                 52206   100   4251   4251    99    99
LZMA:x5:mt1       4923   100   6150   6150   143   143
                 52653   100   4441   4441   103   103
LZMA:x5:mt2       7437   162   5745   9291   134   216
                 52428   100   4422   4422   103   103
Deflate:x1       59094   100   7501   7504   175   175
                181129   100   5629   5628   131   131
Deflate:x5       17770   100   6843   6842   159   159
                181190   100   5624   5625   131   131
Deflate:x7        7132   100   7903   7903   184   184
                183092   100   5681   5682   132   132
Deflate64:x5     16507   100   7136   7133   166   166
                181190   100   5666   5667   132   132
BZip2:x1         10182   100   6151   6152   143   143
                 47906   100   5193   5193   121   121
BZip2:x5          9122   100   7610   7613   177   177
                 35663   100   7000   6999   163   163
BZip2:x5:mt2     11275   196   4808   9410   112   219
                 52954   166   6274  10393   146   242
BZip2:x7          2702   100   7001   7001   163   163
                 35914   100   7042   7043   164   164
PPMD:x1           7656   100   7916   7918   184   184
                  6470   100   7622   7620   177   177
PPMD:x5           5730   100   9711   9712   226   226
                  4876   100   9139   9138   213   213
Delta:4        1296489   100   7968   7966   186   185
               1201257   100   7380   7381   172   172
BCJ            1755153   100   7187   7189   167   167
               1766238   100   7235   7235   168   168
AES256CBC:1     267862   100   6585   6583   153   153
                258645   100   6356   6356   148   148
AES256CBC:2    1091263   100   8940   8940   208   208
               4228645   100   8659   8660   202   202
CRC32:1         597031   100   4346   4346   101   101
CRC32:4        1636837   100   3653   3653    85    85
CRC32:8        3350371   100   4544   4543   106   106
CRC64          1677926   100   3436   3436    80    80
SHA256          275062   100   5610   5611   131   131
SHA1            581637   100   5444   5444   127   127
BLAKE2sp        426969   100   9398   9393   219   219

CPU                      100   4295   4295
------------------------------------------------------
Tot:                     109   5894   6405   137   149



7z b -mm=* -mmt2 -stmC0

7-Zip 17.01 beta (x64) : Copyright (c) 1999-2017 Igor Pavlov : 2017-08-28

x64 6.5504 cpus:16 128T
CMPXCHG MMX SSE RDTSC PAE SSE2 NX SSE3 CMPXCHG16B XSAVE L2AT VIRT_FIRMWARE RDWRFSGSBASE FASTFAIL RDRAND RDTSCP
Intel(R) Core(TM) i7-7820X CPU @ 3.60GHz (50654)
CPU Freq:  4266  4000  4000  4266  4000  4129  4697  4357  4231

RAM size:   32446 MB,  # CPU hardware threads:   2 / 16 : C0 / FFFF
RAM usage:    450 MB,  # Benchmark threads:      2


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

CPU                      200   3994   7987
CPU                      200   3979   7958
CPU                      200   3994   7988   100   200

LZMA:x1          27058   199   4962   9892   124   248
                 73938   200   3011   6022    75   151
LZMA:x5:mt1       7108   200   4441   8880   111   222
                 73470   200   3098   6196    78   155
LZMA:x5:mt2       7766   197   4925   9703   123   243
                 73711   200   3108   6217    78   156
Deflate:x1       77256   200   4905   9810   123   246
                239916   200   3727   7454    93   187
Deflate:x5       23557   200   4536   9070   114   227
                238173   200   3696   7394    93   185
Deflate:x7        9119   200   5051  10105   126   253
                240769   200   3736   7471    94   187
Deflate64:x5     21729   199   4708   9390   118   235
                236399   200   3704   7395    93   185
BZip2:x1         12710   200   3839   7679    96   192
                 67371   200   3652   7303    91   183
BZip2:x5         11334   200   4730   9459   118   237
                 50238   200   4930   9860   123   247
BZip2:x5:mt2     11397   199   4769   9512   119   238
                 54477   197   5416  10692   136   268
BZip2:x7          3271   200   4238   8476   106   212
                 54477   200   5341  10683   134   267
PPMD:x1           9117   200   4715   9430   118   236
                  7766   200   4573   9146   114   229
PPMD:x5           6743   200   5714  11429   143   286
                  5777   200   5413  10826   136   271
Delta:4        1305397   200   4010   8020   100   201
               1214833   200   3732   7464    93   187
BCJ            1920047   200   3933   7865    98   197
               1905508   200   3910   7805    98   195
AES256CBC:1     296156   200   3639   7278    91   182
                295081   200   3626   7252    91   182
AES256CBC:2    2170604   200   8891  17782   223   445
               4559410   200   4669   9338   117   234
CRC32:1        1175132   200   4278   8555   107   214
CRC32:4        3074749   200   3437   6863    86   172
CRC32:8        5801217   200   3933   7866    98   197
CRC64          3081606   200   3156   6311    79   158
SHA256          264472   200   2698   5395    68   135
SHA1            668185   200   3132   6254    78   157
BLAKE2sp        481527   199   5311  10594   133   265

CPU                      200   3993   7986
------------------------------------------------------
Tot:                     200   4092   8163   102   204



7z b -mm=* -mmt8


RAM usage:   1802 MB,  # Benchmark threads:      8


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

CPU                      798   3994  31854
CPU                      793   3987  31625
CPU                      796   4000  31846   100   800

LZMA:x1         161319   790   7461  58973   187  1481
                375075   781   3910  30548    98   767
LZMA:x5:mt1      31067   771   5036  38811   127   975
                372075   784   4003  31379   101   788
LZMA:x5:mt2      45829  1299   4409  57254   111  1438
                378224   792   4028  31897   101   801
Deflate:x1      386317   730   6723  49053   169  1232
               1300234   781   5176  40402   130  1015
Deflate:x5      127069   775   6314  48925   159  1229
               1256410   759   5139  39008   129   980
Deflate:x7       52116   790   7313  57743   184  1451
               1361216   799   5285  42244   133  1061
Deflate64:x5    121588   793   6627  52542   166  1320
               1340558   799   5252  41938   132  1054
BZip2:x1         74898   795   5694  45251   143  1137
                332248   767   4694  36017   118   905
BZip2:x5         59068   796   6190  49296   155  1238
                208381   799   5121  40901   129  1027
BZip2:x5:mt2     72939  1563   3894  60872    98  1529
                279620  1421   3863  54884    97  1379
BZip2:x7         19106   798   6200  49500   156  1243
                209715   789   5214  41127   131  1033
PPMD:x1          57062   800   7377  59016   185  1483
                 46611   788   6962  54890   175  1379
PPMD:x5          35841   787   7714  60741   194  1526
                 29748   770   7241  55746   182  1400
Delta:4        8293171   718   7101  50953   178  1280
               8848880   799   6805  54368   171  1366
BCJ           11638994   730   6527  47673   164  1198
              13006855   791   6738  53276   169  1338
AES256CBC:1    1951518   789   6075  47961   153  1205
               1842599   775   5840  45284   147  1138
AES256CBC:2    8027721   795   8277  65763   208  1652
              31350387   795   8072  64206   203  1613
CRC32:1        4443140   800   4046  32346   102   813
CRC32:4       12176576   799   3400  27178    85   683
CRC32:8       24838904   797   4224  33682   106   846
CRC64         12402549   796   3190  25400    80   638
SHA256         1977614   780   5172  40343   130  1013
SHA1           3599301   700   4810  33689   121   846
BLAKE2sp       3036023   778   8581  66793   216  1678

CPU                      798   3987  31803
------------------------------------------------------
Tot:                     859   5098  43151   126  1084



7z b -mm=* -mmt16


RAM usage:   3604 MB,  # Benchmark threads:     16


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

CPU                     1600   3713  59401
CPU                     1600   3713  59415
CPU                     1595   3724  59401   100  1600

LZMA:x1         198593  1596   4550  72600   123  1956
                534888  1557   2799  43562    75  1173
LZMA:x5:mt1      44146  1534   3594  55152    97  1486
                527558  1573   2829  44490    76  1198
LZMA:x5:mt2      45717  1481   3856  57114   104  1538
                537731  1595   2843  45348    77  1221
Deflate:x1      562023  1593   4481  71364   121  1922
               1737640  1569   3441  53991    93  1454
Deflate:x5      171168  1575   4184  65904   113  1775
               1666412  1505   3438  51735    93  1394
Deflate:x7       66567  1582   4663  73755   126  1987
               1643713  1474   3461  51008    93  1374
Deflate64:x5    159019  1582   4345  68717   117  1851
               1625339  1476   3444  50845    93  1370
BZip2:x1         93756  1594   3554  56644    96  1526
                482769  1565   3345  52335    90  1410
BZip2:x5         73905  1586   3890  61679   105  1661
                276159  1542   3516  54203    95  1460
BZip2:x5:mt2     71099  1510   3931  59337   106  1598
                291757  1549   3697  57265   100  1542
BZip2:x7         23015  1582   3769  59626   102  1606
                275000  1545   3490  53929    94  1453
PPMD:x1          67113  1597   4348  69411   117  1870
                 57261  1594   4230  67431   114  1816
PPMD:x5          44742  1593   4761  75825   128  2042
                 38658  1588   4563  72443   123  1951
Delta:4        9667968  1593   3729  59400   100  1600
               9033373  1600   3469  55501    93  1495
BCJ           14224572  1597   3648  58264    98  1569
              14224572  1596   3651  58264    98  1569
AES256CBC:1    1964157  1445   3341  48271    90  1300
               2201681  1600   3382  54109    91  1457
AES256CBC:2   15866258  1584   8206 129976   221  3501
              33755130  1564   4422  69131   119  1862
CRC32:1        8775726  1600   3993  63887   108  1721
CRC32:4       22937070  1600   3201  51196    86  1379
CRC32:8       43137071  1600   3655  58494    98  1576
CRC64         22986943  1600   2942  47077    79  1268
SHA256         1951244  1585   2512  39805    68  1072
SHA1           4982903  1600   2915  46640    79  1256
BLAKE2sp       3581792  1595   4942  78799   133  2123

CPU                     1594   3719  59278
------------------------------------------------------
Tot:                    1557   3541  55121    95  1485

Links

Skylake X at Wikichip