Apple PowerPC G5 User Manual

PowerPC G5  
The World’s First 64-Bit Desktop Processor  
White Paper  
July 2003  
White Paper  
3
PowerPC G5  
Introduction  
The revolutionary PowerPC G5 changes everything you know about personal computing.  
Suddenly, the next generation of high-performance applications for design and graphics,  
media production, and scientific research is possible and practical on the desktop. That’s  
because the PowerPC G5 brings a 64-bit architecture to the Mac platform—ushering in  
an exciting new era in personal computing.  
Key Features  
• 64-bit architecture, capable of addressing  
18 exabytes of memory  
• Clock speeds up to 2GHz  
• 1GHz frontside bus for throughput of up to  
8 GBps per processor  
• Dual independent 1GHz frontside buses in  
dual processor systems  
• Superscalar execution core supporting up to  
215 in-flight instructions  
• Velocity Engine for accelerated single-  
instruction, multiple-data (SIMD) processing  
• Two double-precision floating-point units  
for high-speed advanced computation  
• Massive three-component branch prediction  
logic to increase processing efficiency  
• Native compatibility with existing 32-bit  
application code  
The introduction of the PowerPC G5 is a product of Apple’s partnership with IBM, lever-  
aging the most advanced chip design and manufacturing expertise in the world. The  
results are phenomenal: 130-nanometer fabrication technology, 2GHz clock speeds, and  
an all-new PowerPC architecture. Together, they put enormous computing power within  
the reach of personal computer users:  
• The ability to address huge amounts of memory provides ultrafast data access, boosting  
performance for 2D imaging, 3D design, and video rendering tasks.  
• A high-bandwidth execution core with 12 functional units improves performance by  
executing multiple instructions per cycle in parallel.  
• An optimized 128-bit Velocity Engine cranks through image editing tasks, high-  
definition video transitions, and complex scientific analysis.  
• State-of-the-art process technology  
from IBM  
• Two double-precision floating-point units accelerate 64-bit calculations for 3D  
visualization, research simulations, and multitrack audio creation.  
The G5 processor is making its debut in the Power Mac G5, the world’s fastest personal  
computer according to industry-standard SPEC benchmarks.1 Now Power Mac users can  
tackle projects never before possible on a desktop system—and blaze through their  
work faster than ever. In fact, the Power Mac G5 runs Adobe Photoshop more than two  
times faster than the fastest Pentium 4–based system.2 Best of all, the PowerPC G5 runs  
32-bit code—including existing Mac OS X applications—natively, so the transition to  
64-bit power is absolutely seamless.  
Welcome to the PowerPC G5, the world’s first 64-bit desktop processor and the heart  
of the new Power Mac.  
White Paper  
4
PowerPC G5  
The World’s First 64-Bit  
Desktop Processor  
The PowerPC G5 marks the arrival of 64-bit performance to the personal computer  
market. With 64-bit-wide data paths and registers, this groundbreaking new processor  
can address vast amounts of main memory and handle multiple large integer and  
floating-point math calculations in a single clock cycle.  
32-bit processing  
Postcard =  
24 in.2 (155 cm2)  
An Exponential Leap in Computing Power  
The label “32-bit” or “64-bit” characterizes the width of a microprocessor’s data stream,  
which is a function of the sizes of its registers and the internal data paths that feed the  
registers. A 64-bit processor moves data and instructions along 64-bit-wide data paths—  
compared with the 32-bit-wide paths on 32-bit processors, such as the Pentium 4. In  
addition, 64-bit processors have wide registers that can store 64-bit numbers as well  
as 32-bit numbers.  
64-bit processing  
Manhattan =  
22 mi.2 (57 km2)  
4.3 billion times bigger  
To grasp the exponential leap from 32-bit to  
64-bit processing, imagine equating the range  
of numbers a processor can express with a  
two-dimensional area. A 32-bit processor can  
express a range of integers equal to the size  
of a postcard, while a 64-bit processor can  
express a range of integers larger than the  
island of Manhattan.  
The leap from 32-bit to 64-bit processing represents an exponential advance in com-  
puting power. With 32-bit registers, a processor has a dynamic range of 232, or 4.3  
billion—which means it can express integers from 0 to 4.3 billion. With 64-bit registers,  
the dynamic range catapults to 264, or 18 billion billion—4.3 billion times larger than  
the range of a 32-bit processor.  
Memory Addressing up to 18 Exabytes  
The move to 64-bit processing results in a similarly dramatic leap in the amount of  
memory supported. A memory address is a special kind of integer, and each address  
points to one byte in memory. Since memory addresses are computed in 64-bit registers  
capable of expressing 18 billion billion integers, the PowerPC G5 can theoretically  
address 18 exabytes (18 billion billion bytes) of virtual memory.  
In practice, memory addressing is defined by the physical address space of the processor.  
The PowerPC G5, with 42 bits of physical address space, supports a colossal 242 bytes,  
or 4 terabytes, of system memory. Although it’s not currently feasible to purchase 4 tera-  
bytes of RAM, very large quantities of memory enable a desktop system to contain a  
gigantic 3D model, a complex scientific simulation, or a sequence of HD video entirely  
in RAM—drastically reducing the time to access, modify, and render the data.  
White Paper  
5
PowerPC G5  
High-Precision Calculations in a Single Clock Cycle  
With 64-bit-wide data paths and registers, the PowerPC G5 can execute instructions  
on 64 bits of data in a single clock cycle—making it possible to perform huge integer  
calculations and highly precise floating-point mathematics. In contrast, a 32-bit pro-  
cessor would have to split up any data larger than 32 bits and process it over multiple  
clock cycles. The advanced calculation capability of the PowerPC G5 is critical in state-  
of-the-art applications for scientific simulations, 3D modeling, and video effects.  
Clock Speeds up to 2GHz  
The PowerPC G5 features a scalable design that enables it to run at clock speeds  
up to 2GHz. This represents a 600MHz jump, the largest in PowerPC history, over the  
fastest G4 processor at 1.4GHz. What’s more, this breakthrough clock speed boosts  
performance across the board, accelerating everything from gaming frame rates to  
digital audio effects.  
Industry-Leading 1GHz Frontside Bus  
PowerPC G5  
PowerPC G5  
Xeon  
Xeon  
To harness the power of the G5 processor, a 64-bit Double Data Rate (DDR) frontside  
bus maximizes throughput between the processor and the rest of the system. Unlike  
conventional processor interfaces, which carry data in only one direction at a time, the  
PowerPC G5 features two dedicated unidirectional 32-bit data paths (64 bits total): One  
travels into the processor and one travels from the processor, with no wait time while  
the processor and the system controller negotiate which will use the bus or while the  
bus switches direction. In addition, the data streams integrate clock signals along with  
the data—allowing the frontside bus to work at speeds up to 1GHz for an astounding  
8 GBps of total bandwidth.  
The bidirectional frontside bus allows data to  
travel to and from the PowerPC G5 processor  
at the same time. In dual processor systems,  
each PowerPC G5 has its own dedicated inter-  
face to maximize throughput—compared  
with dual Xeon-based systems, which force  
the processors to share a single bus.  
In dual PowerPC G5 systems, each processor has its own discrete 1GHz frontside bus.  
The result is a maximum aggregate bandwidth of 16 GBps on dual 2GHz Power Mac  
G5 systems, well over twice the 6.4-GBps maximum throughput of Pentium 4– or  
Xeon-based systems. In addition to providing fast access to main memory, this high-  
performance frontside bus architecture enables each PowerPC G5 to discover and  
access data in the other processor’s caches—a process called intervention. Cache  
intervention is made possible by cache coherency, which ensures that the processor  
always fetches the correct data, even if it has been modified and is in L2 cache.  
Full Support for Symmetric Multiprocessing  
The PowerPC G5 is designed for symmetric multiprocessing—enabling multiple  
applications to run independently on different processors or a single multithreaded  
application to perform multiple tasks simultaneously. For example, while performing  
an edit, Final Cut Pro can decode two pieces of source video, one on each processor,  
at the same time.  
With dual independent frontside buses and built-in cache coherency, a dual processor  
system manages priorities between the two processors for maximum efficiency. And  
since Mac OS X was built from the ground up for symmetric multiprocessing, no special  
optimization is required for software to take advantage of this powerful capability.  
White Paper  
6
PowerPC G5  
Native Compatibility with 32-Bit Application Code  
On other platforms, switching to a 64-bit computer requires migrating to a 64-bit  
operating system (and purchasing 64-bit applications) or running a 32-bit operating  
system in a slow emulation mode. With the PowerPC G5, the transition to a 64-bit  
system is seamless: Current 32-bit code—such as existing Mac OS X and Classic  
applications—runs natively at processor speed, with no interruptions to your work-  
flow and no additional investment in software.  
This easy compatibility is possible because the PowerPC architecture, unlike competing  
instruction sets, was designed from the beginning to run both 32-bit and 64-bit appli-  
cation code. And because the PowerPC G5 uses the same Velocity Engine instruction  
set introduced in the PowerPC G4, applications that have been optimized for Velocity  
Engine will immediately run faster on the new processor. What’s more, as applications  
are optimized and as Mac OS X is further enhanced for the PowerPC G5, performance  
gains will be even greater.  
White Paper  
7
PowerPC G5  
Next-Generation PowerPC  
Architecture  
The PowerPC G5 is a highly parallel implementation of the PowerPC architecture,  
capable of handling multiple assorted tasks at the same time. It’s based on the execu-  
tion core of IBM’s 64-bit POWER4 processor—recipient of the Microprocessor Report’s  
2001 Analyst’s Choice Award for Best Workstation/Server Processor, which recognizes  
excellence in semiconductor technology innovation, design, and implementation.  
With two double-precision floating-point units, advanced branch prediction logic,  
and a high-bandwidth frontside bus, the POWER4 drives IBM’s top-of-the-line pSeries  
690 servers.  
Apple collaborated with IBM to leverage this industry-leading design for the next  
generation of personal computing. The development of the PowerPC G5 builds on  
previous PowerPC designs, combining an optimized Velocity Engine with a superscalar,  
superpipelined execution core that supports up to 215 simultaneous in-flight instruc-  
tions. And all this power is fabricated in IBM’s state-of-the-art 130-nanometer process  
technology using high-performance silicon-on-insulator (SOI) transistors and copper  
interconnects.  
The PowerPC G5 is fabricated using state-of-  
the-art 130-nanometer circuitry with more  
than 1130 feet of ultrathin wiring—nearly  
800 times thinner than a human hair.  
Ultrafast Access to Data and Instructions  
The PowerPC G5 features processing innovations that optimize the flow of data  
and instructions—making it ideal for media streaming, HD video editing, real-time  
effects, audio synthesis, image processing, 3D rendering, numerical analysis, and  
physical modeling.  
It starts with 512K of L2 cache that provides the execution core with ultrafast access to  
data and instructions—up to 32 GBps. Instructions are prefetched from the L2 cache  
into a large, direct-mapped 64K L1 cache at up to 64 GBps. As they are accessed from  
the L1 cache, up to eight instructions are fetched per clock cycle. Next, instructions are  
decoded and divided into smaller, faster-executing operations. In addition, 32K of L1  
data cache can prefetch up to eight active data streams simultaneously.  
This efficient preparation maximizes processing speed as instructions are dispatched  
into the execution core and data is loaded into the registers.  
White Paper  
8
PowerPC G5  
PowerPC G5 Architecture  
Highly Parallel Execution Core  
At the heart of the PowerPC G5 is an entirely new superscalar, superpipelined execution  
core, composed of 12 functional units that execute different types of instructions con-  
currently for massive data throughput. Before instructions are dispatched into the  
functional units, they are arranged into groups of up to five. Within the core alone,  
the PowerPC G5 can track up to 20 groups at a time, or 100 individual instructions. This  
efficient group-tracking scheme enables the PowerPC G5 to manage an unusually large  
number of instructions “in flight”: 20 instructions in each of the five dispatch groups,  
in addition to 100-plus instructions in the various fetch, decode, and queue stages.  
Up to 215 in-flight instructions  
A wide architecture with 12 discrete processing  
units enables the PowerPC G5 to contain up  
to 215 in-flight instructions simultaneously—  
71 percent more than the 126 instructions in  
a Pentium 4.  
Aggressive Queuing and Register Renaming  
The PowerPC G5 includes extensive instruction queueing and renamable data registers  
that maximize the utilization of each functional unit. Once an instruction group is dis-  
patched into the execution core, it is broken into individual instructions, which proceed  
to the appropriate functional unit. Each functional unit has its own dedicated queue,  
where multiple instructions are sequenced for processing.  
At the same time, the load/store units load the associated data from L1 cache into the  
data registers behind the units that will be processing the data. To improve processing  
efficiency, the PowerPC G5 features a large number of rename registers that act as  
proxies, or placeholders, until the appropriate data arrives for execution. The instruction  
is held in queue, allowing other operations to take place until the actual data is loaded  
into the registers.  
When operations on the data are complete, the PowerPC G5 recombines the instructions  
into the original groups of five, and the load/store units store the data in cache or main  
memory to complete processing.  
White Paper  
9
PowerPC G5  
Optimized 128-Bit Velocity Engine  
The PowerPC G5 uses a dual-pipelined Velocity Engine optimized with two independent  
queues and dedicated 128-bit registers and data paths for efficient instruction and data  
flow. This 128-bit vector processing unit accelerates data manipulation by applying a sin-  
gle instruction to multiple data at the same time, known as SIMD processing. Originally  
implemented in the PowerPC G4, the Velocity Engine in the PowerPC G5 uses the same  
set of 162 instructions, enabling it to run—and accelerate—existing Mac OS X applica-  
tions that have been optimized for the Velocity Engine.  
128-bit  
Velocity  
Engine  
32-bit  
Processor  
Vector processing is useful for transforming large sets of data, such as manipulating an  
image or rendering a video effect. For example, when a designer uses a filter to apply  
a motion blur to an image, each pixel of the image must be changed according to the  
same set of instructions—a highly repetitive processing task. Each Velocity Engine pipe-  
line speeds up this task by processing up to 128 bits of data, in four 32-bit integers, eight  
16-bit integers, sixteen 8-bit integers, or four 32-bit single-precision floating-point values,  
all in a single clock cycle.  
The Velocity Engine can manipulate 128  
bits of data at a time, up to four times  
faster than the general processing units  
in 32-bit processors.  
Two Double-Precision Floating-Point Units  
Today’s powerful applications demand both precision and performance. That’s why the  
PowerPC G5 has two double-precision floating-point units, enabling it to complete at  
least two 64-bit mathematical calculations per clock cycle. In fact, each floating-point  
unit can perform both an add and a multiply with a single instruction. This dramatically  
accelerates highly complex computations that are critical in research simulations and in  
many of the applications used to manipulate or render 3D graphics and video content.  
Weather prediction is one example of a highly iterative computing task made possible  
by floating-point math. Large-scale models simulate weather patterns over time by mea-  
suring multiple influences, such as atmospheric pressure and airflow, at various instants  
and recalculating the model every minute. The PowerPC G5 provides the precision and  
performance to deliver accurate results within a useful timeframe.  
Two Integer Units  
Integer units perform simple integer mathematics—such as add, subtract, and compare—  
which are commonly used in many basic computer functions, as well as in imaging,  
video, and audio applications. The PowerPC G5 has two integer units capable of a broad  
range of simple and complex instructions involving both 32-bit and 64-bit data. What’s  
more, they take full advantage of the processor’s 64-bit registers and data paths to com-  
plete 64-bit integer calculations in a single pass.  
Two Load/Store Units  
Load/store units manage data as it is processed, loading it into the registers of each  
functional unit and, after processing, storing the new data in L1 cache, L2 cache, or main  
memory, as appropriate. The PowerPC G5 is generously equipped with three large sets of  
registers: A general-purpose register file contains 64-bit registers for integer calculations;  
a floating-point register file contains 64-bit registers for floating-point calculations; and  
a vector register file contains 128-bit registers for the Velocity Engine. Each register file  
holds 32 registers for architected values, as well as 48 rename, or proxy, registers. With  
two load/store units, the PowerPC G5 is able to keep these registers filled with data for  
maximum processing efficiency.  
White Paper  
10  
PowerPC G5  
Condition Register  
This special 32-bit register summarizes the states of the floating-point and integer units.  
The condition register also indicates the results of comparison operations and provides  
a means for testing them as branch conditions. By bridging information between the  
branch unit and other functional units, the condition register improves the flow of data  
throughout the execution core.  
Three-Component Branch Prediction Logic  
Advanced processors use branch prediction and speculative operation to increase  
efficiency. A branch is a question in the processing queue: Which instruction should  
go next? Branch prediction anticipates the answer; and speculative operation causes  
that instruction to be executed. If the prediction is correct, the processor works more  
efficiently—since the speculative operation has executed an instruction before it’s  
required. If the prediction is incorrect, the processor must clear the unneeded branch,  
as well as any related data and instructions, resulting in an empty space called a pipeline  
bubble. Pipeline bubbles reduce performance as the processor marks time waiting for  
the next instruction.  
The PowerPC G5 features an innovative three-component branch prediction logic  
to reduce pipeline bubbles and maximize processor efficiency. The success or failure  
of each prediction is captured in three 16K branch history tables—local, global, and  
selector—that are used to improve the accuracy of future branch predictions.  
Local branch prediction takes place as individual instructions are fetched into the pro-  
cessor and the types of branches are recorded in the local branch history table. Global  
branch prediction occurs at the same time: Branches are identified in their processing  
context, relating to preceding and subsequent operations; and the results are recorded  
in the global branch history table. The third,selector” history table identifies which  
prediction type, local or global, was more accurate in predicting the outcome of each  
branch. This dynamic local/global/selector branch history scheme can predict branch  
processes with a high degree of accuracy, allowing the PowerPC G5 to efficiently use  
every processing cycle.  
Copper interconnects improve conductivity  
and boost processor performance.  
State-of-the-Art Process Technology from IBM  
The PowerPC G5 is fabricated in one of IBM’s world-class semiconductor manufacturing  
facilities. With industry-leading build, assembly, and test technology, IBM uses a cutting-  
edge 130-nanometer process with more than 58 million silicon-on-insulator (SOI) trans-  
istors and eight layers of copper interconnects.  
SOI refers to the placement of a thin layer of insulator between transistors and bulk  
silicon. When transistors are built on this SOI layer, their capacitance, or the tendency  
to store an electrical charge, is reduced—resulting in faster operation.  
For further performance gains, an additive-copper wiring process replaces the con-  
ventional subtractive-aluminum process. As semiconductor wires are made thinner  
and narrower, aluminum resists the flow of electricity and slows down transmission  
of electrical signals. Copper wiring is less resistant and results in a 40 percent gain in  
conductivity, again for faster processor operation.  
IBM is a worldwide leader in processor  
fabrication technologies, with a new  
$3 billion, state-of-the-art facility in  
East Fishkill, New York.  
White Paper  
11  
PowerPC G5  
Industry-Leading  
Performance  
Ultrafast clock speeds and a highly parallel 64-bit architecture make the PowerPC G5  
ideal for next-generation multimedia, graphics, and scientific applications. Integer and  
floating-point math calculations are faster than ever thanks to 64-bit-wide registers  
and data paths.  
To demonstrate the performance advantages of this groundbreaking new processor,  
Apple put the dual 2GHz Power Mac G5 to the test against the top-of-the-line Pentium  
4– and Xeon-based systems, the 3GHz Dell Dimension 8300 and the dual 3.06GHz Dell  
Precision 650, respectively.  
SPEC CPU2000  
The Standard Performance Evaluation Corporation (SPEC) CPU2000 benchmark suite  
is the recognized industry standard for assessing processing performance. SPEC is a  
nonprofit organization of hardware and software vendors, universities, and consultants.  
They developed the SPEC CPU2000 benchmarks based on actual end-user applications.  
These tests depend on processor, memory subsystem, and compiler performance when  
executing integer and floating-point computations. For more information on the bench-  
Apple hired an independent laboratory, VeriTest, to conduct the SPEC CPU2000  
benchmark tests and provide documented results. Since SPEC CPU2000 measures the  
performance of both the hardware and the compiler, VeriTest normalized the compiler  
on both platforms to allow for a direct comparison of hardware performance alone.  
VeriTest used GCC—an open source compiler popular with programmers around the  
world—with similar settings on all systems. (Even though GCC cannot automatically  
generate Velocity Engine code for the PowerPC G5, settings included automatic genera-  
tion of optimized SSE/SSE2 code for the Pentium 4 and the Xeon.) The Power Mac G5  
used Mac OS X v10.2.7 (G5), and the Intel-based systems used Red Hat Linux 9.0.  
The SPEC CPU2000 benchmark suite comprises four metrics—SPECint_base2000,  
SPECfp_base2000, SPECint_rate_base2000, and SPECfp_rate_base2000—to represent  
different performance characteristics of the system.  
White Paper  
12  
PowerPC G5  
SPECint_base2000 and SPECfp_base2000 measure the speed of a single task—either  
an integer calculation or a floating-point calculation—executing on a single processor.  
Each test measures how long the processor takes to complete the benchmark set of  
single tasks relative to a SPEC-defined baseline score. SPECint_base2000 is composed  
of eleven C and one C++ benchmark applications, including a chess program, a data  
compression utility, and a place-and-route simulator. SPECfp_base2000 consists of six  
Fortran-77, four Fortran-90, and four C benchmark applications, including shallow-  
water modeling, neural-network simulation, and computational chemistry.  
SPEC CPU2000: Single-Processor Speed  
SPECint_base2000: Integer calculations  
Power Mac G5  
Dual 2GHz PowerPC G5  
800  
Dell Precision 650  
Dual 3.06GHz Xeon  
836  
Dell Dimension 8300  
889  
3GHz Pentium 4  
0
500  
1000  
SPECfp_base2000: Floating-point calculations  
Power Mac G5  
Dual 2GHz PowerPC G5  
840  
Dell Precision 650  
Dual 3.06GHz Xeon  
646  
Dell Dimension 8300  
3GHz Pentium 4  
693  
0
500  
1000  
SPEC score  
In single-processor tests, the Power Mac G5 completed the set of floating-point calculations  
21 percent faster than the Pentium 4–based system and 30 percent faster than the Xeon-  
based workstation, while it performed slightly below both systems when executing simple  
integer calculations.3  
White Paper  
13  
PowerPC G5  
For comparisons that more accurately demonstrate the performance of a dual proces-  
sor system, VeriTest used the “SPEC rate” metrics, which recognize multiple processors.  
With SPECint_rate_base2000 and SPECfp_rate_base2000, the benchmark code is  
compiled and multiple copies are run concurrently, allowing both processors to work  
in parallel. SPEC rate tests determine the number of times a system can complete the  
benchmark per hour, also referred to as system throughput.  
SPEC CPU2000: Dual Processor System Throughput  
SPECint_rate_base2000: Integer calculations  
Power Mac G5  
Dual 2GHz PowerPC G5  
17.2  
Dell Precision 650  
Dual 3.06GHz Xeon  
16.7  
Dell Dimension 8300  
10.3  
3GHz Pentium 4  
0
10  
20  
SPECfp_rate_base2000: Floating-point calculations  
Power Mac G5  
Dual 2GHz PowerPC G5  
15.7  
Dell Precision 650  
Dual 3.06GHz Xeon  
11.1  
Dell Dimension 8300  
8.1  
3GHz Pentium 4  
0
10  
20  
SPEC score  
These results clearly demonstrate the benefits of the dual processor Power Mac G5. With  
full support for symmetric multiprocessing, dual independent 1GHz frontside buses, and  
two floating-point units per processor, the dual 2GHz Power Mac G5 completed the set  
of floating-point calculations 95 percent faster than the Pentium 4–based system and  
42 percent faster than the dual Xeon-based workstation. Integer performance was also  
far superior to the Pentium 4–based system and on par with the dual Xeon-based system.3  
White Paper  
14  
PowerPC G5  
Technical Specifications  
64-bit PowerPC processor architecture  
• Virtual address range: 64 bits, or 18 exabytes  
• Physical address range: 42 bits, or 4 terabytes  
• Full 64-bit data paths and registers  
• Native support for 32-bit application code  
• 64K L1 instruction cache; 32K L1 data cache  
512K internal L2 cache  
• Dedicated data flow for dividing one instruction into two internal operations  
• Microcoded instructions for up to four internal operations  
• Support for up to eight outstanding L1 cache line misses  
• Hardware-initiated instruction prefetching from L2 cache  
• Hardware- or software-initiated data stream prefetching; support for up to eight  
active streams  
• Maximum core frequency: 2GHz  
Frontside bus  
• Throughput: up to 8 GBps per processor  
• Frequency: 1GHz DDR  
• Width: 64-bit (32-bit in, 32-bit out) bidirectional  
Wide execution core  
• Support for up to 215 in-flight instructions  
• In-order dispatch of up to five operations into distributed issue queue structure  
• Simultaneous issue of up to 10 out-of-order operations:  
One Velocity Engine permute operation  
One Velocity Engine arithmetic logic operation  
Two floating-point operations  
Two fixed-point register-to-register operations  
Two load or store operations  
One condition register operation  
One branch operation  
• Out-of-order and speculative issue of load operations  
• Dual-pipeline 128-bit Velocity Engine for single-instruction, multiple-data  
(SIMD) processing  
• Two independent floating-point units for double-precision calculations  
White Paper  
15  
PowerPC G5  
Three-component branch prediction logic  
• Speculative superscalar inner core organization  
• Fast, selective flush of incorrect speculative instructions and results  
• Prediction of up to two branches per cycle  
• Support for up to 16 predicted branches in flight  
• Prediction hints added to branch instructions  
• Prediction support for branch direction and branch addresses  
Physical specifications  
• 58 million transistors  
130-nanometer, silicon-on-insulator (SOI) process  
• Die size: 118 square millimeters  
Comparison of PowerPC G4 and PowerPC G5  
PowerPC G4  
PowerPC G5  
Architecture  
32-bit  
64-bit  
Addressable memory  
Maximum clock speed  
Frontside bus  
4 gigabytes  
4 terabytes  
1.4GHz  
2GHz  
Up to 167MHz shared  
Up to 1GHz per processor  
In-flight instructions  
Floating-point units  
Integer units  
16  
215  
One  
Two  
One  
Two  
Load/Store units  
L1 data cache  
One  
Two  
32K  
32K  
L1 instruction cache  
L2 cache  
32K  
64K  
256K  
512K  
Branch prediction logic  
Process technology  
Die size  
Local  
Local/Global/Selector  
130-nanometer  
118 square millimeters  
180-nanometer  
106 square millimeters  
1
“World’s fastest” based on SPEC CPU2000 benchmark results and leading professional application performance tests  
against 3GHz Pentium 4–based Dell Dimension 8300 and dual 3.06GHz Xeon-based Dell Precision Workstation 650.  
SPEC CPU2000 benchmarks run with GCC 3.3 compiler and independently tested; professional applications tested by  
For More Information  
For more information about the PowerPC  
For more information about the Power  
2
Apple, June 2003. Tests performed by Apple in June 2003 using preproduction Power Mac G5 units. The Power Mac  
systems ran a PowerPC G5–optimized version of Photoshop 7.0.1 including optimized AltiVecCore, ACE, and BIB Carbon  
Libraries; the Dell Dimension ran Photoshop 7.0.1. Based on SPEC CPU2000 benchmark results against 3GHz Pentium  
4–based Dell Dimension 8300 and dual 3.06GHz Xeon-based Dell Precision Workstation 650m, performed by VeriTest,  
June 2003.  
3
© 2003 Apple Computer, Inc. All rights reserved. Apple, the Apple logo, Final Cut Pro, Mac, Mac OS, and Power Mac are  
trademarks of Apple Computer, Inc., registered in the U.S. and other countries. Velocity Engine is a trademark of Apple  
Computer, Inc. Adobe is a trademark or registered trademark of Adobe Systems Incorporated in the U.S. and/or other  
countries. PowerPC is a trademark of International Business Machines Corporation, used under license therefrom. SPEC  
information. Other product and company names mentioned herein may be trademarks of their respective companies.  
Product specifications are subject to change without notice. This material is provided for informational purposes only;  
Apple assumes no liability related to its use. July 2003 L29540B  

Hanns G HX 191DPB User Manual
GE Monogram ZGU366LPSS User Manual
GE 2 930SST User Manual
Frigidaire LA04CT80425 1A User Manual
Elo TouchSystems 1925L Series User Manual
Daewoo KOC 624Q User Manual
Dacor Cooktop RR30NIS User Manual
Audiovox 1286970 User Manual
AT T CL82100 User Manual
Acer TravelMate 3220 User Manual