MCST Elbrus (МЦСТ Эльбрус)

MCST Elbrus-3M 300 MHz, 218M transistors, 120 nm, 8-metal layer, 189 mm2. Power: 6W.

MCST Elbrus-3S (1891VM5YA / 1891ВМ5Я) 500 MHz, 218M transistors, 90 nm, 9-metal layer, 142 mm2. Power: 13-20W.

MCST Elbrus-2C+ (1891VM7YA / 1891ВМ7Я), 2 Elbrus cores (500 MHz). 4 DSP cores ElCore-09, 368 M transistors, 90 nm, 289 mm2. 2x DDR2-800, Power: 25W.

MCST Elbrus-4C (1891VM8YA / 1891ВМ8Я) 4 cores, 800 MHz, 986 (960) M transistors, 65 nm (9-metal), 380 mm2. Power: 45W. 3x DDR3-1600.

MCST Elbrus-1C+ (1891VM11YA / 1891ВМ11Я), 1 core, 1000 MHz. MCST MGA2 2D core, Vivante GC2500 GPU. 275 (or 375 ?) M transistors, 40 nm, 122 mm2. Power: 10W. 2x DDR3-1600.

MCST Elbrus-8C (МЦСТ Эльбрус-8С) (1891VM10YA / 1891ВМ10Я) 8 cores, 1300 MHz, 2730 M transistors, 28 nm, 321 mm2. Power: 80W. 4x DDR3-1600.

MCST Elbrus-8SV (МЦСТ Эльбрус-8СВ) (1891VM12YA / 1891ВМ12Я, 1891VM12AYA / 1891ВМ12АЯ), 8 cores, 1500 MHz, 3500 M transistors, 28 nm, 350 mm2. Power: 90W. 4x DDR4-2400.

MCST Elbrus-16S (МЦСТ Эльбрус-16С) (1891VM038 / 1891ВМ038, 1891VM03A8 / 1891ВМ03A8), 16 cores, 2000 MHz, 12 B transistors, 16 nm, 618 mm2. 8x DDR4-3200.

MCST Elbrus-2S3 (МЦСТ Эльбрус-2С3) (1891VM068 / К1891ВМ068, 1891VM06A8 / К1891ВМ06A8), 2 cores, 2000 MHz, 4.3 B transistors, 16 nm, 222 mm2. 2x DDR4-3200.

Chipset KPI for 2C+/4С/R1000, 250 MHz, 30 M transistors, 130 nm, 6W, 2010.

Chipset KPI-2 for 1C+/8С/R2000, 500 MHz, 85 M transistors, 151 mm2, 65 nm.

L1 Instruction cache = 64 KB. 256 B/line (3M/3S/2C+), 4-WAY, VA indexed (47:14). 2 tag banks (bank0: tags for even lines, bank1: tags for odd lines). LRU. each bank: 2 read ports (main thread + additional_thread). total: 2 tags/cycle for main thread, and 2 tags/cycle in additional thread. Throughput: 256 b/c for one thread, or 64/c for each from 4 threads
L1 Instruction cache = 128 KB. 256 B/line (1C+/4C/8C), 4-WAY
2 clusters in execution core.
Register File: 256 registers (84 bits per register). Two Register Files (same data in each cluster).
6 ALUs (3 ALUs per cluster).
L1 Data cache = 64 KB. 32 B/line. 4-WAY. Two copies (1 copy per cluster). 2 ports per copy. LRU (3 bits per 4 lines).

2M+: L2 cache = 256 KB / core.
2C+: L2 cache = 1 MB / core. ?-WAY, 64 B/line. There are 4 bits of tag for each 8-bytes word.
4C: L2 cache = 2 MB / core. ?-WAY, 64 B/line.
1C+: L2 cache = 2 MB ?-WAY, 64 B/line.
8C: L2 cache = 512 KB / core, 4-WAY, 64 B/line.
8SV: L2 cache = 512 KB / core, 4-WAY, 64 B/line.
2S3: L2 cache = 2 MB / core.
16S: L2 cache = 1 MB / core.

Elbrus-8C: L3 cache = 16 MB, 16-WAY, 64 B/line, 8 slices (bits 8,7,6 of address), NRU replacement. Inclusive
Elbrus-16S: L3 cache = 32 MB

Elbrus v3,v4,v5:

L1 Data Cache Latency = 3 cycles for simple access via pointer
L1 Data Cache Latency = 4 cycles for access with complex address calculation (size_t n, *p; n = p[n]).

Elbrus v6:

L1 Data Cache Latency = 5 cycles for simple access via pointer

Home Memory Units:

Elbrus-8C: 2 Home Memory Units (HMU). each units : 2 RAM channels, 512 KB directory, 16-way, each tag cover two 64-bytes lines.
Elbrus-16C: 4 Home Memory Units (HMU) ?, total 2.5 MB directory.

4 KB pages mode, Elbrus-2C+

L2 Cache Latency = 9 cycles
RAM Latency = 9 cycles + 94 ns for open RAM page
RAM Latency = 9 cycles + 108 ns for random access

Data TLB: 1024 items. Miss penalty = 16 cycles. The core can't make more than one TLB miss handling (Page walk) simultaneously.

  Size        Latency       Increase   Description

  64 K     3                           
 128 K     6               3           + 6 (L2)        
 256 K     8               2       
 512 K     9 +   4 ns      1 +  4 ns   
   1 M     9 +  24 ns          20 ns
   2 M     9 +  64 ns          30 ns   + 108 ns (RAM)
   4 M    11 +  92 ns      2 + 28 ns   
   8 M    17 + 104 ns      6 + 12 ns   + 16 (TLB miss, Page walk to L1)
  16 M    21 + 107 ns      4 +  3 ns   
  32 M    24 + 108 ns      3 +  1 ns   
  64 M    27 + 108 ns      3 +  0 ns   + 6 (Page walk to L2)
 128 M    29 + 118 ns      2 + 10 ns       
 256 M    31 + 139 ns      2 + 21 ns
 512 M    31 + 166 ns          27 ns   + 108 ns (Page walk to RAM)

8-bytes range cross penalty = 6 cycles
32-bytes range cross penalty = 10 cycles
L1 B/W (Parallel Random Read with direct pointers) = 0.25 cycles per one access
L1 B/W (Parallel Random Read with complex addresses) = 0.41 cycles per one access
L2->L1 B/W (Parallel Random Read) = 1 cycles per L1 cache line (32 bytes) = 16 GB/s
L2->L1 B/W (Read, 64 bytes step) = 1.5 cycles per L1 cache line (32 bytes)
L2 Write (Write, 8 bytes step) = 0.5 cycles per write
RAM Read B/W (Parallel Random Read) = 18 ns per L2 cache line (64 bytes) = 3500 MB/s
RAM Read B/W (Read, 8 Bytes step) = 540 MB/s
RAM Read B/W (Read, 64 Bytes step) = 3500 MB/s
RAM Read B/W (Read, 8 Bytes step - pointer chasing) = 350 MB/s
RAM Read B/W (Read, 64 Bytes step - pointer chasing) = 540 MB/s
RAM Write B/W (Write, 8-64 Bytes step) = 3800 MB/s

Pipeline

A : Assiciative Search: ITAG read, ITLB tags serach
F0 : Fetch 0 : ITAG read, ITLB data read (physical address)
F1 : Fetch 1 : DATA read
S : Scattering :
D : Decode :
B : Branch and Basing :
R : Read : Register Read
E0 : Execute 0 :
E1 : Execute 1 :
E2 : Execute 2 :

7-Zip Benchmark Elbrus-8C

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.


Elbrus 801, 8 cores, 1300 MHz, 32-bit

OPTFLAGS=-O4 -ffast -mtune=elbrus-8c -fprofile-use -mptr32


$ ./bin/7za b -mmt1

7-Zip (a) 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=C,Utf16=off,HugeFiles=on,32 bits,8 CPUs LE)

LE
CPU Freq:  1297  1281  1299  1299  1299  1299  1299  1299  1297

RAM size:   32164 MB,  # CPU hardware threads:   8
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:       1720   100   1674   1674  |      20138   100   1720   1719
23:       1638   100   1669   1669  |      19910   100   1724   1723
24:       1565   100   1683   1683  |      19221   100   1688   1687
25:       1516   100   1732   1732  |      18975   100   1689   1689
----------------------------------  | ------------------------------
Avr:             100   1689   1689  |              100   1705   1705
Tot:             100   1697   1697


$ ./bin/7za b -mmt8

7-Zip (a) 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=C,Utf16=off,HugeFiles=on,32 bits,8 CPUs LE)

LE
CPU Freq:  1297  1298  1298  1299  1299  1298  1299  1299  1299

RAM size:   32164 MB,  # CPU hardware threads:   8
RAM usage:   1765 MB,  # Benchmark threads:      8

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

22:       8154   618   1284   7932  |     153178   797   1639  13065
23:       8345   650   1308   8503  |     150815   797   1638  13051
24:       8262   666   1333   8884  |     148142   796   1633  13002
25:       8430   687   1402   9625  |     144462   791   1625  12857
----------------------------------  | ------------------------------
Avr:             655   1332   8736  |              795   1634  12994
Tot:             725   1483  10865





$ ./bin/7za b -mmt=* -mm=*

7-Zip (a) 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=C,Utf16=off,HugeFiles=on,32 bits,8 CPUs LE)

LE
CPU Freq:  1297  1298  1299  1298  1299  1299  1299  1298  1299

RAM size:   32164 MB,  # CPU hardware threads:   8
RAM usage:    225 MB,  # Benchmark threads:      1


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

CPU                      100   1299   1299
CPU                      100   1299   1299
CPU                      100   1299   1299   100   100

LZMA:x1           4733   100   1731   1730   133   133
                 19462   100   1585   1585   122   122
LZMA:x5:mt1       1561   100   1951   1951   150   150
                 19222   100   1621   1621   125   125
LZMA:x5:mt2       2250   160   1753   2811   135   216
                 19219   100   1621   1621   125   125
Deflate:x1        9569   100   1215   1215    94    94
                 30257   100    940    940    72    72
Deflate:x5        3546   100   1365   1365   105   105
                 30296   100    941    941    72    72
Deflate:x7        1031   100   1143   1143    88    88
                 30388   100    945    943    73    73
Deflate64:x5      3409   100   1473   1473   113   113
                 30139   100    943    943    73    73
BZip2:x1          2324   100   1405   1405   108   108
                 10345   100   1122   1121    86    86
BZip2:x5          2031   100   1695   1695   130   130
                  8987   100   1764   1764   136   136
BZip2:x5:mt2      3871   193   1674   3231   129   249
                 13577   150   1771   2665   136   205
BZip2:x7           596   100   1547   1546   119   119
                  9029   100   1771   1771   136   136
PPMD:x1           1160   100   1201   1201    92    92
                  1499   100   1766   1766   136   136
PPMD:x5            872   100   1479   1479   114   114
                  1183   100   2218   2217   171   171
Delta:4         151931   100    934    933    72    72
                119786   100    736    736    57    57
BCJ             174279   100    714    714    55    55
                174260   100    714    714    55    55
AES256CBC:1      68809   100   1691   1691   130   130
                 71154   100   1749   1749   135   135
AES256CBC:2 

CRC32:1         174951   100   1274   1274    98    98
CRC32:4         507131   100   1132   1132    87    87
CRC32:8         844906   100   1146   1146    88    88
CRC64           391338   100    801    801    62    62
SHA256          129304   100   2638   2638   203   203
SHA1            199740   100   1870   1870   144   144
BLAKE2sp    

CPU                      100   1299   1299
------------------------------------------------------
Tot:                     109   1596   1748   124   135


RAM usage:   1802 MB,  # Benchmark threads:      8


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

CPU                      767   1301   9982
CPU                      798   1300  10370
CPU                      797   1301  10371   100   800

LZMA:x1          36894   793   1701  13487   131  1040
                153680   798   1568  12517   121   966
LZMA:x5:mt1      10655   782   1702  13312   131  1027
                140727   758   1565  11868   121   916
LZMA:x5:mt2       9193   779   1475  11485   114   886
                139959   754   1566  11803   121   911
Deflate:x1       72543   790   1166   9211    90   711
                212989   729    907   6618    70   511
Deflate:x5       28082   796   1359  10813   105   834
                213056   729    907   6615    70   510
Deflate:x7        8057   784   1139   8928    88   689
                212790   726    910   6604    70   509
Deflate64:x5     26453   782   1463  11431   113   882
                210993   727    908   6601    70   509
BZip2:x1         18452   794   1404  11149   108   860
                 78555   763   1117   8516    86   657
BZip2:x5         12586   795   1322  10504   102   810
                 58221   779   1466  11428   113   882
BZip2:x5:mt2     11878   790   1255   9913    97   765
                 54610   743   1443  10719   111   827
BZip2:x7          4302   791   1410  11145   109   860
                 58503   781   1470  11473   113   885
PPMD:x1           9155   791   1197   9469    92   730
                 11486   771   1755  13527   135  1043
PPMD:x5           6152   789   1321  10426   102   804
                  8116   793   1918  15209   148  1173
Delta:4        1181158   777    934   7257    72   560
                936768   781    737   5756    57   444
BCJ            1391181   797    715   5698    55   440
               1365748   782    715   5594    55   432
AES256CBC:1     547438   797   1688  13454   130  1038
                543529   767   1741  13358   134  1030
AES256CBC:2 

CRC32:1        1398036   799   1274  10178    98   785
CRC32:4        4050118   798   1133   9040    87   697
CRC32:8        6527119   772   1146   8851    88   683
CRC64          3060106   784    800   6267    62   483
SHA256         1032306   798   2638  21059   204  1624
SHA1           1599271   798   1875  14969   145  1155
BLAKE2sp    

CPU                      798   1300  10371
------------------------------------------------------
Tot:                     775   1477  11465   114   884





Elbrus 804, 4 sockets, 32 cores, 1200 MHz, 64-bit

# OPTFLAGS=-O4 -ffast -mtune=elbrus-8c -fprofile-use -mptr64


$ ./bin/7za b -mmt=1

7-Zip (a) 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=C,Utf16=off,HugeFiles=on,64 bits,32 CPUs LE)

LE
CPU Freq:  1195  1196  1196  1198  1198  1198  1198  1197  1198

RAM size:  257783 MB,  # CPU hardware threads:  32
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:       1526   100   1485   1485  |      18254   100   1559   1559
23:       1452   100   1480   1480  |      18011   100   1559   1559
24:       1391   100   1496   1496  |      17475   100   1534   1534
25:       1347   100   1539   1538  |      17260   100   1536   1536
----------------------------------  | ------------------------------
Avr:             100   1500   1500  |              100   1547   1547
Tot:             100   1524   1523


$ ./bin/7za b -mmt=32

7-Zip (a) 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=C,Utf16=off,HugeFiles=on,64 bits,32 CPUs LE)

LE
CPU Freq:  1194  1196  1198  1198  1198  1198  1198  1198  1198

RAM size:  257783 MB,  # CPU hardware threads:  32
RAM usage:   7060 MB,  # Benchmark threads:     32

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

22:      23266  2620    864  22633  |     494007  2996   1406  42128
23:      24732  2805    898  25200  |     465498  2978   1353  40280
24:      24759  2860    931  26621  |     454901  2995   1333  39928
25:      24917  2916    976  28449  |     448275  2878   1386  39894
----------------------------------  | ------------------------------
Avr:            2800    917  25726  |             2962   1370  40558
Tot:            2881   1143  33142



$ ./bin/7za b -mmt=* -mm=*

7-Zip (a) 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21
p7zip Version 16.02 (locale=C,Utf16=off,HugeFiles=on,64 bits,32 CPUs LE)

LE
CPU Freq:  1196  1197  1198  1198  1198  1198  1198  1198  1198

RAM size:  257783 MB,  # CPU hardware threads:  32
RAM usage:    225 MB,  # Benchmark threads:      1


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

CPU                      100   1198   1198
CPU                      100   1198   1198
CPU                      100   1198   1198   100   100

LZMA:x1           4251   100   1554   1554   130   130
                 17317   100   1410   1410   118   118
LZMA:x5:mt1       1389   100   1736   1736   145   145
                 17510   100   1477   1477   123   123
LZMA:x5:mt2       1940   163   1485   2424   124   202
                 17518   100   1478   1478   123   123
Deflate:x1        8311   100   1055   1055    88    88
                 25598   100    795    795    66    66
Deflate:x5        3207   100   1235   1235   103   103
                 25597   100    795    795    66    66
Deflate:x7         942   100   1046   1045    87    87
                 25681   100    797    797    67    67
Deflate64:x5      3046   100   1322   1317   110   110
                 25438   100    796    796    66    66
BZip2:x1          1242   100    751    750    63    63
                  8396   100    910    910    76    76
BZip2:x5          1060   100    885    885    74    74
                  7355   100   1444   1444   120   120
BZip2:x5:mt2      2048   195    876   1709    73   143
                 11255   154   1436   2209   120   184
BZip2:x7           373   100    969    968    81    81
                  7392   100   1450   1450   121   121
PPMD:x1           1059   100   1096   1096    91    91
                  1138   100   1341   1341   112   112
PPMD:x5            802   100   1359   1359   113   113
                   910   100   1706   1706   142   142
Delta:4         100765   100    619    619    52    52
                107845   100    663    663    55    55
BCJ             160616   100    658    658    55    55
                160625   100    658    658    55    55
AES256CBC:1      61314   100   1507   1507   126   126
                 59607   100   1465   1465   122   122
AES256CBC:2 

CRC32:1         126471   100    921    921    77    77
CRC32:4         359812   100    803    803    67    67
CRC32:8         623461   100    845    845    71    71
CRC64           334868   100    686    686    57    57
SHA256          124431   100   2539   2538   212   212
SHA1            106878   100   1000   1000    83    83
BLAKE2sp    

CPU                      100   1198   1198
------------------------------------------------------
Tot:                     109   1351   1478   113   123



RAM usage:   7208 MB,  # Benchmark threads:     32


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

CPU                     3056   1198  36619
CPU                     3094   1198  37068
CPU                     3131   1198  37516   102  3200

LZMA:x1         123226  3051   1477  45048   126  3842
                499803  2935   1387  40706   118  3472
LZMA:x5:mt1      30131  2750   1369  37642   117  3211
                437623  2826   1306  36904   111  3148
LZMA:x5:mt2      25691  3053   1052  32096    90  2738
                468691  3004   1316  39524   112  3371
Deflate:x1      221312  2910    966  28101    82  2397
                753847  3058    766  23423    65  1998
Deflate:x5       94799  3068   1190  36500   101  3113
                755454  3061    766  23453    65  2000
Deflate:x7       29321  3137   1036  32487    88  2771
                761962  3067    771  23645    66  2017
Deflate64:x5     91577  3086   1283  39573   109  3375
                740871  3014    769  23177    66  1977
BZip2:x1         38681  3153    741  23370    63  1993
                244518  2997    884  26507    75  2261
BZip2:x5         27236  2990    761  22730    65  1939
                157646  2819   1098  30942    94  2639
BZip2:x5:mt2     25130  2999    700  20973    60  1789
                141730  2964    939  27818    80  2373
BZip2:x7         11244  3160    923  29131    79  2485
                158609  2860   1088  31104    93  2653
PPMD:x1          32641  3129   1079  33759    92  2880
                 34060  3019   1329  40110   113  3421
PPMD:x5          21088  3019   1184  35739   101  3048
                 23737  3028   1470  44482   125  3794
Delta:4        3146702  3123    619  19333    53  1649
               3386622  3142    662  20807    56  1775
BCJ            4955746  3082    659  20299    56  1731
               5036798  3131    659  20631    56  1760
AES256CBC:1    1453889  2604   1372  35731   117  3048
               1622191  2856   1396  39867   119  3401
AES256CBC:2 

CRC32:1        3938050  3114    921  28669    79  2445
CRC32:4       10930401  3036    804  24397    69  2081
CRC32:8       19646974  3150    846  26641    72  2272
CRC64         10442229  3124    685  21386    58  1824
SHA256         3717530  2988   2538  75838   216  6469
SHA1           3308660  3078   1006  30969    86  2642
BLAKE2sp    

CPU                     3030   1198  36307
------------------------------------------------------
Tot:                    2964   1177  34727   100  2962

Links

Elbrus-8C at Wikipedia

Elbrus-1C+ at MCST (Russian)

Elbrus-8C at MCST (Russian)

Elbrus-8SV at MCST (Russian)