Floating Point Capability Most Powerful, Architecturally Advanced

The Datasheet Archive - 100 Million Datasheets from 7500 Manufacturers.

Floating Point Capability Most Powerful, Architecturally Advanced

Datasheet Thumbnail

Top Searches for this datasheet

Floating Point Capability
Most Powerful, Architecturally Advanced Floating Point Engine Ever Delivered Microprocessor
ADVANCED MICRO DEVICES, INC.
Place Sunnyvale, 94088
Contact:
Drew Prairie Public Relations (408) 749-4581 drew.prairie@amd.com
Page
Athlon Floating Point Engine
August 2000
Introduction: Most Powerful Floating Point Engine x86-Based Platforms
Athlonprocessor designed power next generation computing platforms, delivering ultimate performance cutting-edge applications unprecedented visual computing experience. deliver this ultimate visual computing experience, designed Athlon processor include most powerful floating point engine platforms. This white paper details nextgeneration features technology that Athlon processor's floating point engine apart from floating point units (FPUs) previous-generation processors.
Need High-Performance Floating Point Engine
High-performance floating point units-once considered necessary only scientific, mathematical, engineering applications running workstations-are used variety other applications, such modeling physical objects motion, applying special lighting effects complex digital images, enabling more realistic smoother movements objects games. With proliferation 3D-intensive software, many computing platforms require powerful floating point capabilities. home, popularity entertainment gaming titles skyrocketed software developers have tapped power floating point deliver ever increasing levels realism detail their latest titles. Robust floating point performance become essential workplace where latest CAD, desktop publishing, digital content creation, spreadsheet, business graphics, mathematical, statistical applications rely floating point units perform complex calculations. Exceptional floating point performance continues critical capability scientific engineering workstation applications. meet floating point (FP) performance demands cutting-edge, FP-intensive software applications, equipped Athlon processor with most powerful architecturally advanced floating point execution engine ever implemented microprocessor1. centerpiece seventh-generation Athlon processor architecture, AMD's "no-compromise," superscalar floating point engine complements capabilities Athlon processor's enhanced 3DNow!technology (see "Enhanced 3DNow!Technology" white paper) enables performance that unparalleled computing industry.
measured industry-standard benchmarks; www.amd.com details.
Page
Athlon Floating Point Engine
August 2000
Floating Point Basics
Floating point arithmetic simply represent real number with "floating" decimal point rather than "fixed" decimal point. enables more precise mathematical calculations than possible with integers, whole numbers that lack decimal point. Floating point techniques especially useful geometric calculations common CAD/CAE processing, high-precision mathematical calculations, graphics applications involving physics, geometry, triangle setup. Processor-based floating point arithmetic been thoroughly standardized. IEEE754 standards specify numeric formats, value sets, basic floating point arithmetic works. basic floating point types include:
Single-precision format 32-bit data width, 7-digit precision Double-precision format 64-bit data width, to16-digit precision Extended-precision format 80-bit data width, 20-digit precision.
Early processors interfaced separate, dedicated floating point unit, also called "math co-processor," perform numerically intensive calculations. Beginning with processor, floating point unit integrated onto processor die, delivering on-chip floating point functionality performance application developers. next innovation floating point technology processors AMD's 3DNow! technology, introduced 1998 feature AMD-K6®-2 processor. AMD's 3DNow! first 3D-enhancement instruction architecture, well first SIMD (Single Instruction Multiple Data) floating point technology processors. essence, 3DNow! technology instructions using superscalar SIMD floating point capabilities significantly improve processor's ability handle floating point calculations, dramatically improving graphics, audio, other multimedia performance. SIMD enables processors execute multiple single-precision floating point calculations clock cycle. 3DNow! technology designed complement role floating point unit enhancing processor's overall ability accelerate certain algorithms that lend themselves SIMD single-precision floating point computation, such physics geometry operations. fact, primary rationale developing 3DNow! provide specific acceleration those applications that well-suited vector processing single-precision floating point code. Applications requiring greater than single-precision processing better handled engine.
Page
Athlon Floating Point Engine
August 2000
Today's software developers require exceptional floating performance across computing platforms power generations FP-intensive applications, from imaging modeling CAD/CAE speech recognition sophisticated scientific technical applications. seventh-generation Athlon processor delivers industry's most powerful floating point engine meet these processing needs enable superior performance next generation platforms.
Anatomy Powerful Athlon ProcessorFloating Point Engine
Athlon processor features three-issue, superscalar floating point design based three pipelined, out-of-order floating point units, each with one-cycle throughput. These three execution pipelines make Athlon processor's floating point engine both superscalar superpipelined. term superscalar means that processor multiple instruction pipelines that operate parallel, enabling processor execute multiple instructions single clock cycle. Superpipelining means that there many small pipeline stages-each performing simple, dedicated task-to optimize processor clock frequency. Athlon processor world's first processor with these capabilities, delivering unprecedented performance satisfy most demanding applications platforms. Figure (AMD Athlon architecture block diagram) details three execution pipes used Athlon processor.
2-way, 64KB Instruction Cache
24-entry TLB/256-entry
Predecode Cache
Branch Prediction Table
Fetch/Decode Control
3-Way Instruction Decoders Instruction Control Unit (72-entry)
Integer Scheduler (18-entry)
Stack Rename Scheduler (36-entry) Register File (88-entry)
Interface Unit
FStore
FADD 3DNow! FMUL 3DNow!
Cache Controller
Load Store Queue Unit
2-way, 64KB Data Cache
32-entry TLB/256-entry
System Interface
SRAMs
Figure AthlonProcessor Block Diagram
Page
Athlon Floating Point Engine
August 2000
When compared other floating point architectures, design performance Athlon processor's floating point engine apart true next-generation implementation. example, Athlon processor delivers twice peak floating point execution rate Intel Pentium® processor rivals performance many RISC processors currently used workstations servers. floating point engine Athlon processor fully compatible conforms IEEE standards single-, double-, extended-precision data types. Athlon processor's floating point engine also handle MMXTM, 3DNow!, floating point instructions. Athlon processor's floating point engine uses three separate execution pipelines superscalar x87, 3DNow!, operations:
floating point load/store (FSTORE) pipeline handles constant loads, stores, miscellaneous operations. adder pipeline (FADD) contains 3DNow! add, ALU/shifter, execution units. multiplier (FMUL) pipeline contains ALU, multiplier, reciprocal unit, FP/3DNow! instruction multiplier, support FDIV instructions.
addition superscalar design, Athlon processor's floating point engine superpipelined. This technique supports higher clock frequencies enables floating point engine process complex instructions more quickly deliver higher overall instruction throughput. Table compares number execution pipelines, well throughput, Athlon processor versus Intel Pentium III.
Table AthlonProcessor Floating Point Engine Features/Latency/Throughput Comparison Seventh Generation
Feature Athlon Processor
Intel Previous Generation
Intel Pentium Pentium Xeon execution pipelines Latency Throughput Rates Classes Instructions* FADD FMUL FDIV (single) 16/13 17/17 FDIV (double) 20/17 32/32 FDIV (extended) 24/21 37/37 FSQRT (single) 19/16 28/28 FSQRT (double) 27/24 57/57 FSQRT (extended) 35/31 68/68 FCOM
Latency/throughput rates measured processor clocks; lower numbers mean higher throughput.
Page
Athlon Floating Point Engine
August 2000
pipeline analogous automobile assembly line, which divided into many functional parts, each dedicated building particular piece entire car. This efficient assembly line model increases factory's overall throughput. Similarly, using pipeline instruction "assembly line") enhances processor's overall throughput. Table compares floating point throughput Athlon processor relative Pentium processor various instructions. Athlon processor clearly superior Pentium important multiplication, division, squareroot instructions. seventh-generation Athlon processor's floating point performance significantly better than that previous-generation processor same clock speed, including Pentium III. This higher floating point performance provides users with more productive computing experience, with such benefits faster video frame rates, greater detail, superior realism graphics computer-aided design applications, faster image rendering modeling digital content creation, faster mathematical statistical calculations business workstation applications.
Summary: Unprecedented Performance Platforms
Cutting-edge, floating point-intensive software applications next-generation computing platforms require exceptional, levels floating point performance. delivers this level performance next-generation Athlon processor, which designed offer unprecedented floating point performance high-end desktop systems, workstations, servers. Athlon processor's floating point engine most powerful floating point solution ever implemented processor planned power exciting visual computing applications computing enthusiasts, small business users, enterprise users years come.
Overview
(NYSE: AMD) global supplier integrated circuits personal networked computer communications markets. produces processors, flash memories, integrated circuits telecommunications networking applications. world's second-leading supplier Windows® compatible processors, shipped more than million microprocessors, including more than million Windows compatible CPUs. Founded 1969 based Sunnyvale, California, sales marketing offices worldwide manufacturing facilities Sunnyvale;
Page
Athlon Floating Point Engine
August 2000
Austin, Texas; Dresden, Germany; Bangkok, Thailand; Penang, Malaysia; Singapore; Aizu-Wakamatsu, Japan. revenues $2.9 billion 1999.
Cautionary Statement
This White Paper includes forward-looking statements that made pursuant safe harbor provisions Private Securities Litigation Reform 1995. Forward-looking statements generally preceded words such "expects," "plans," "believes," "anticipates," "intends." Investors cautioned that forward-looking statements this white paper involve risks uncertainties that could cause actual results differ from current expectations. Forward-looking statements this white paper about Athlon processor involve risk that Athlon processor will achieve customer market acceptance; that software applications will optimized with processor. urge investors review detail risk uncertainties company's Securities Exchange Commission filings, including most recently filed Form-10K.
AMD, logo, Athlon combinations thereof, 3DNow! trademarks, AMD-K6 registered trademark Advanced Micro Devices, Inc. Windows Windows registered trademarks Microsoft Corporation. trademark, Pentium registered trademark Intel Corporation. WinBench registered trademark Ziff-Davis, Inc. Other product names used this publication identification purposes only trademarks their respective companies.
Page
Athlon Floating Point Engine
August 2000

Floating Point Capability Most Powerful, Architecturally Advanced

Top Searches for this datasheet

Other recent searches