Knowledge (XXG)

POWER3

Source đź“ť

25: 727:, where multiplication and addition is performed simultaneously. Such instructions, along with individual add and multiply, have a four-cycle latency. Divide and square-root instructions are executed in the same FPUs, but are assisted by specialized hardware. Single-precision (32-bit) divide and square-root instructions have a 14-cycle latency, whereas double-precision (64-bit) divide and square-root instructions have an 18-cycle and a 22-cycle latency, respectively. 716:, but was simpler as it did not require an extra clock cycle to synchronize the two copies due to the POWER3's higher cycle times. The floating-point register file contains 56 registers, of which 32 are floating-point registers and 24 rename registers. Compared to the PowerPC 620, there were more rename registers, which allowed more instructions to be executed out of order, improving performance. 662: 759: 697:
instruction queues. Restrictions on instruction issue are few: of the two integer instruction queues, only one can accept one instruction, the other can accept up to four, as does the floating-point instruction queue. If the queues do not have enough unused entries, instructions cannot be issued. The front end has a short pipeline, resulting in a small three-cycle
579: 89: 709:. To reduce the number of ports required to provide data and receive results, the general purpose register file is duplicated so that there are two copies, the first supporting three integer execution units and the second supporting the two load/store units. This scheme was similar to a contemporary microprocessor, the 734:
The PowerPC 620 data cache was optimized for technical and scientific applications. Its capacity was doubled to 64 KB, to improve the cache-hit rate; the cache was dual-ported, implemented by interleaving eight banks, to enable two loads or two stores to be performed in one cycle in certain cases;
719:
Execution begins in stage four. The instruction queues dispatch up to eight instructions to the execution units. Integer instructions are executed in three integer execution units (termed "fixed-point units" by IBM). Two of the units are identical and execute all integer instructions except for
730:
After execution is completed, the instructions are held in buffers before being committed and made visible to software. Execution finishes in stage five for integer instructions and stage eight for floating-point. Committing occurs during stage six for integers, stage nine for floating-point.
696:
The front end consists of two stages: fetch and decode. During the first stage, eight instructions were fetched from a 32 KB instruction cache and placed in a 12-entry instruction buffer. During the second stage, four instructions were taken from the instruction buffer, decoded, and issued to
720:
multiply and divide. All instructions executed by them have a one-cycle latency. The third unit executes multiply and divide instructions. These instructions are not pipelined and have multi-cycle latencies. 64-bit multiply has a nine-cycle latency and 64-bit divide has a 37-cycle latency.
704:
In stage three, instructions in the instruction queues that are ready for execution have their operands read from the register files. The general-purpose register file contains 48 registers, of which 32 are general-purpose registers and 16 are rename registers for
766:
The POWER3-II was an improved POWER3 that increased the clock frequency to 450 MHz. It contains 23 million transistors and measured 170 mm. It was fabricated in the IBM CMOS7S process, a 0.22 ÎĽm CMOS process with six levels of
735:
and the line-size was increased to 128-bytes. The L2 cache bus was doubled in width to 256 bits to compensate for the larger cache line size and to retain a four-cycle latency for cache refills.
565: 742:
process that is a hybrid of 0.25 ÎĽm feature sizes and 0.35 ÎĽm metal layers. The process features five layers of aluminium. It was packaged in the same 1,088-column
739: 955: 916: 558: 673:, an earlier 64-bit PowerPC implementation that was late, under-performing and commercially unsuccessful. Like the PowerPC 620, the POWER3 has three 633:
and completed IBM's long-delayed transition from POWER to PowerPC, which was originally scheduled to conclude in 1995. The POWER3 was used in IBM
784: 610: 551: 513: 950: 46: 523: 861: 498: 68: 945: 960: 693:. It has a seven-stage integer pipeline, a minimal eight-stage load/store pipeline and a ten-stage floating-point pipeline. 710: 685:
units, and an extra load-store unit was added (for a total of two) to improve floating-point performance. The POWER3 is a
638: 940: 602: 125: 39: 33: 518: 698: 794: 738:
The POWER3 contained 15 million transistors on a 270 mm die. It was fabricated in IBM's CMOS-6S2 process, a
622: 252: 831: 50: 743: 605:(ISA), including all of the optional instructions of the ISA (at the time) such as instructions present in the 724: 682: 690: 166: 909: 879: 839: 674: 900:
O'Connell, F. P.; White, S. W. (6 November 2000). "POWER3: The next generation of PowerPC processors".
731:
Writeback occurs in the stage after commit. The POWER3 can retire up to four instructions per cycle.
723:
Floating-point instructions are executed in two floating-point units (FPUs). The FPUs are capable of
488: 431: 424: 768: 678: 625:
it replaced from the more consumer-oriented 32-bit PowerPCs. The POWER3 was the successor of the
706: 465: 448: 417: 460: 892: 852: 646: 868: 617:, a high-end graphics workstation. The POWER3 was originally supposed to be called the 591: 934: 924: 503: 217: 212: 493: 207: 202: 197: 192: 862:"RS/6000 Scientific and Technical Computing: POWER3 Introduction and Tuning Guide" 661: 613:
but not in the PowerPC ISA. It was introduced on 5 October 1998, debuting in the
713: 686: 670: 642: 618: 374: 350: 337: 332: 326: 758: 455: 343: 582:
Dual 375 MHz IBM POWER3-II processors on the CPU module of a RS/6000 44P 270.
508: 174: 578: 397: 789: 650: 634: 614: 599: 528: 385: 313: 234: 170: 130: 772: 630: 606: 380: 356: 308: 303: 298: 293: 287: 280: 266: 259: 155: 145: 757: 660: 577: 229: 88: 830:
Papermaster, M.; Dinkjian, R.; Mayfield, M.; et al. (1998).
907:
Song, Peter (17 November 1997). "IBM's Power3 to Replace P2SC".
747: 626: 471: 637:
servers and workstations at 200 MHz. It competed with the
621:
but was renamed, probably to differentiate the server-oriented
595: 18: 897:- gives more information about POWER1, POWER2, and POWER3 832:"POWER3: Next Generation 64-bit PowerPC Processor Design" 860:
Anderson, S.; Bell, R.; Hague, J.; et al. (1998).
151: 141: 136: 124: 119: 111: 103: 98: 598:, that implemented the 64-bit version of the 559: 8: 81: 917:International Business Machines Corporation 681:(FPU) was replaced with two floating-point 594:, designed and exclusively manufactured by 566: 552: 162: 87: 69:Learn how and when to remove this message 665:The logic schema of the POWER3 processor 32:This article includes a list of general 902:IBM Journal of Research and Development 806: 740:complementary metal–oxide–semiconductor 480: 440: 408: 244: 181: 165: 956:Computer-related introductions in 1998 888: 877: 848: 837: 785:IBM POWER Instruction Set Architecture 80: 182:NXP (formerly Freescale and Motorola) 16:1998 family of microprocessors by IBM 7: 689:design that executed instructions 38:it lacks sufficient corresponding 14: 750:, but with a different pin out. 23: 120:Architecture and classification 1: 639:Digital Equipment Corporation 699:branch misprediction penalty 669:The POWER3 was based on the 603:instruction set architecture 951:Superscalar microprocessors 977: 867:. IBM Corp. Archived from 771:. It was succeeded by the 795:IBM Power microprocessors 744:ceramic column grid array 86: 189:PowerPC e series (2006) 946:PowerPC microprocessors 53:more precise citations. 961:64-bit microprocessors 904:, Volume 44, Number 6. 887:Cite journal requires 847:Cite journal requires 763: 666: 583: 322:PowerPC series (1992) 93:POWER 3 microprocessor 910:Microprocessor Report 761: 664: 615:RS/6000 43P Model 260 581: 489:OpenPOWER Foundation 126:Instruction set 941:IBM microprocessors 921:New IBM POWER3 chip 814:New IBM POWER3 chip 769:copper interconnect 679:floating-point unit 99:General information 83: 919:(5 October 1998). 764: 725:fused multiply–add 683:fused multiply–add 667: 629:derivative of the 584: 542:historic in italic 370:RAD series (1997) 226:Qor series (2008) 707:register renaming 677:, but the single 675:fixed-point units 576: 575: 538:Cancelled in gray 161: 160: 79: 78: 71: 968: 896: 890: 885: 883: 875: 873: 866: 856: 850: 845: 843: 835: 817: 811: 623:POWER processors 568: 561: 554: 539: 451: 163: 91: 84: 74: 67: 63: 60: 54: 49:this article by 40:inline citations 27: 26: 19: 976: 975: 971: 970: 969: 967: 966: 965: 931: 930: 886: 876: 871: 864: 859: 846: 836: 829: 826: 821: 820: 812: 808: 803: 781: 756: 659: 647:Hewlett-Packard 609:version of the 572: 537: 449: 94: 75: 64: 58: 55: 45:Please help to 44: 28: 24: 17: 12: 11: 5: 974: 972: 964: 963: 958: 953: 948: 943: 933: 932: 929: 928: 914: 905: 898: 889:|journal= 874:on 2006-07-21. 857: 849:|journal= 825: 822: 819: 818: 805: 804: 802: 799: 798: 797: 792: 787: 780: 777: 755: 752: 658: 655: 592:microprocessor 574: 573: 571: 570: 563: 556: 548: 545: 544: 534: 533: 532: 531: 526: 521: 516: 511: 506: 501: 496: 491: 483: 482: 478: 477: 476: 475: 468: 463: 458: 453: 443: 442: 438: 437: 436: 435: 428: 421: 411: 410: 406: 405: 404: 403: 393: 392: 391: 390: 389: 388: 383: 378: 368: 367: 366: 363: 354: 347: 340: 335: 330: 319: 318: 317: 316: 311: 306: 301: 296: 291: 284: 277: 270: 263: 255:series (1990) 247: 246: 242: 241: 240: 239: 238: 237: 232: 223: 222: 221: 220: 215: 210: 205: 200: 195: 184: 183: 179: 178: 159: 158: 153: 149: 148: 143: 139: 138: 134: 133: 128: 122: 121: 117: 116: 113: 109: 108: 105: 101: 100: 96: 95: 92: 77: 76: 59:September 2017 31: 29: 22: 15: 13: 10: 9: 6: 4: 3: 2: 973: 962: 959: 957: 954: 952: 949: 947: 944: 942: 939: 938: 936: 926: 925:Press release 922: 918: 915: 912: 911: 906: 903: 899: 894: 881: 870: 863: 858: 854: 841: 833: 828: 827: 823: 815: 810: 807: 800: 796: 793: 791: 788: 786: 783: 782: 778: 776: 774: 770: 760: 753: 751: 749: 745: 741: 736: 732: 728: 726: 721: 717: 715: 712: 708: 702: 700: 694: 692: 688: 684: 680: 676: 672: 663: 656: 654: 652: 648: 644: 640: 636: 632: 628: 624: 620: 616: 612: 608: 604: 601: 597: 593: 589: 580: 569: 564: 562: 557: 555: 550: 549: 547: 546: 543: 536: 535: 530: 527: 525: 522: 520: 517: 515: 512: 510: 507: 505: 502: 500: 497: 495: 492: 490: 487: 486: 485: 484: 481:Related links 479: 474: 473: 469: 467: 464: 462: 459: 457: 454: 452: 447: 446: 445: 444: 439: 434: 433: 429: 427: 426: 422: 420: 419: 415: 414: 413: 412: 407: 401: 399: 395: 394: 387: 384: 382: 379: 377: 376: 372: 371: 369: 364: 361: 360: 358: 355: 353: 352: 348: 346: 345: 341: 339: 336: 334: 331: 329: 328: 324: 323: 321: 320: 315: 312: 310: 307: 305: 302: 300: 297: 295: 292: 290: 289: 285: 283: 282: 278: 276: 275: 271: 269: 268: 264: 262: 261: 257: 256: 254: 251: 250: 249: 248: 243: 236: 233: 231: 228: 227: 225: 224: 219: 216: 214: 211: 209: 206: 204: 201: 199: 196: 194: 191: 190: 188: 187: 186: 185: 180: 177:architectures 176: 172: 168: 164: 157: 154: 150: 147: 144: 140: 135: 132: 129: 127: 123: 118: 114: 110: 106: 102: 97: 90: 85: 73: 70: 62: 52: 48: 42: 41: 35: 30: 21: 20: 920: 908: 901: 880:cite journal 869:the original 840:cite journal 813: 809: 765: 737: 733: 729: 722: 718: 703: 695: 691:out of order 668: 587: 585: 541: 494:AIM alliance 470: 430: 423: 416: 409:IBM/Nintendo 396: 373: 349: 342: 325: 286: 279: 273: 272: 265: 258: 65: 56: 37: 834:. IBM Corp. 714:Alpha 21264 687:superscalar 671:PowerPC 620 657:Description 643:Alpha 21264 619:PowerPC 630 142:Predecessor 112:Designed by 51:introducing 935:Categories 824:References 456:PWRficient 34:references 775:in 2001. 762:POWER3-II 754:POWER3-II 611:POWER ISA 509:Power.org 504:Blue Gene 175:Power ISA 152:Successor 779:See also 645:and the 432:Espresso 425:Broadway 104:Launched 790:PowerPC 746:as the 651:PA-8500 635:RS/6000 600:PowerPC 529:AltiVec 386:RAD5500 375:RAD6000 359:(2010) 314:Power10 235:Qorivva 171:PowerPC 137:History 131:PowerPC 47:improve 773:POWER4 641:(DEC) 631:POWER2 607:POWER2 588:POWER3 402:(1996) 400:series 381:RAD750 309:POWER9 304:POWER8 299:POWER7 294:POWER6 288:POWER5 281:POWER4 274:POWER3 267:POWER2 260:POWER1 173:, and 156:POWER4 146:POWER2 82:POWER3 36:, but 872:(PDF) 865:(PDF) 801:Notes 649:(HP) 590:is a 466:Xenon 450:Titan 441:Other 418:Gekko 253:Power 230:QorIQ 218:e6500 213:e5500 167:POWER 893:help 853:help 748:P2SC 627:P2SC 586:The 524:CHRP 519:PReP 514:PAPR 499:RISC 472:X704 461:Cell 398:RS64 344:74xx 208:e600 203:e500 198:e300 193:e200 107:1998 711:DEC 596:IBM 365:A2O 362:A2I 351:970 338:7xx 333:4xx 327:6xx 245:IBM 115:IBM 937:: 923:. 884:: 882:}} 878:{{ 844:: 842:}} 838:{{ 701:. 653:. 540:, 357:A2 169:, 927:. 913:. 895:) 891:( 855:) 851:( 816:. 567:e 560:t 553:v 72:) 66:( 61:) 57:( 43:.

Index

references
inline citations
improve
introducing
Learn how and when to remove this message

Instruction set
PowerPC
POWER2
POWER4
POWER
PowerPC
Power ISA
e200
e300
e500
e600
e5500
e6500
QorIQ
Qorivva
Power
POWER1
POWER2
POWER3
POWER4
POWER5
POWER6
POWER7
POWER8

Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.

↑