Compaq Evo Workstation w6000, Evo Workstation w8000 White Paper

White Paper

February 2002

167T-0202A-WWEN

Prepared by:

Workstations Division

Compaq Computer Corporation

Contents

Introduction................................. 3

Hyper-Threading Technology .... 3

Advantages ............................... 3

Overview ................................... 4

Compaq Evo Workstations

W6000 and W8000....................... 8

Enabling/Disabling Hyper-

Threading.................................. 8

Hyper-Threading and

Microsoft Windows XP .............. 9

Hyper-Threading and

Microsoft Windows 2000 ......... 10

Performance Analyses ............. 11

SYSmark 2001 ........................ 11

Winstone ................................. 13

Pro/ENGINEER 2001i²............ 14

Excel Benchmark .................... 16

Solid Edge V10 ....................... 17

Montecarlo Benchmark ........... 18

Verilog Compiled Simulation

Benchmark .............................. 19

Summary ................................... 20

Appendix A —

Enable/Disable Hyper-

Threading .................................. 22

Appendix B — Verify HAL

Version ...................................... 28

References ................................ 35

Hyper-Threading Technology,

New Feature of Intel Xeon

Processor

Abstract: New refreshes of Compaq Evo Workstations W6000 and

W8000 will feature the leading edge Intel Xeon processor with

Hyper Threading technology. This new processor is fabricated with

the latest .13µ (micron) technology, 512KB-L2 cache, and the ability to support frequencies ranging from 1.8 GHz to greater than 2.6

GHz. This processor also includes support for multi-threaded execution, known as Hyper-Threading technology.

The purpose of this paper is to provide an overview of Hyper-

Threading technology, to analyze and study the performance impacts of the system with Hyper-Threading technology, enabled as well as disabled, and to provide the technical information and benefits of this feature.

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Notice

The information in this publication is subject to change without notice and is provided “AS IS” WITHOUT

WARRANTY OF ANY KIND. THE ENTIRE RISK ARISING OUT OF THE USE OF THIS

INFORMATION REMAINS WITH RECIPIENT. IN NO EVENT SHALL COMPAQ BE LIABLE FOR

ANY DIRECT, CONSEQUENTIAL, INCIDENTAL, SPECIAL, PUNITIVE, OR OTHER DAMAGES

WHATSOEVER (INCLUDING, WITHOUT LIMITATION, DAMAGES FOR LOSS OF BUSINESS

PROFITS, BUSINESS INTERRUPTION, OR LOSS OF BUSINESS INFORMATION), EVEN IF

COMPAQ HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.

The limited warranties for Compaq products are exclusively set forth in the documentation accompanying such products. Nothing herein should be construed as constituting a further or additional warranty.

This publication does not constitute an endorsement of the product or products that were tested. The configuration or configurations tested or described may or may not be the only available solution. This test is not a determination of product quality or correctness, nor does it ensure compliance with any federal, state or local requirements.

Compaq and Evo are trademarks of Compaq Information Technologies Group, L.P. in the U.S. and/or other countries.

Microsoft, Windows, and Windows NT are trademarks of Microsoft Corporation in the U.S. and/or other countries.

Intel, Pentium, and Xeon are trademarks of Intel Corporation in the U.S. and/or other countries.

All other product names mentioned herein may be trademarks of their respective companies.

©2002 Compaq Information Technologies Group, L.P.

Hyper-Threading Technology, New Feature of Intel Xeon Processor

White Paper prepared by Workstations Division

First Edition (February 2002)

Document Number 167T-0202A-WWEN

2

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Introduction

The Compaq Evo Workstations W6000 and W8000 incorporate the new leading-edge Intel Xeon

Multi-processor (MP) processor and the Intel i860 chipset to achieve the next level of system performance. The purpose of this paper is to explain the impacts on processor performance with the new Intel Hyper-Threading (HT) technology enabled versus disabled in Compaq Evo

Workstations W6000 and W8000.

The performance of the system is directly related to Instructions Per Clock (IPC). The two ways to increase system performance today are the following.

•

The operating system (OS) of multi-processor (MP) system configurations, such as

Microsoft Windows NT, Windows 2000, Windows XP and any later OS that supports MP, take advantage of Thread Level Parallelism (TLP). These operating systems divide their workloads into processes and threads, which can be independently scheduled and dispatched to run on the processors that are idle or available.

Note: Microsoft Windows 98 and Microsoft ME do not support multiple processors.

•

Instruction Level Parallelism (ILP) increases the IPC. Very Long Instruction Word (VLIW) and Explicitly Parallel Instruction-set Computing (EPIC) are designed from the ground up to support IPC. However, these instructions do not maintain binary compatibility with the Intel

Architecture (IA), known as the IA-32 Instruction-set.

Hyper-Threading Technology

Intel has developed a new feature called Hyper-Threading technology for the IA-32 architecture, which provides a performance boost to the processors.

Hyper-Threading technology signals a fundamental shift in processor design. The multi-threading design techniques allow an OS to view a single physical processor as if it were two logical processors. To accomplish this, processors enabled with Hyper-Threading technology can manage incoming tasks from different software applications and continuously alternate from one task to the other, without losing track of the data processing status of each task.

Current processor technologies queue up individual instructions and handle them one at a time in a logical order (out of order execution, in-order retirement), much like a person who is trying to watch television while talking on the phone. In the current scenario, the person takes in data from each medium, handling it piece by piece. Hyper-Threading technology allows the viewer to accomplish both activities simultaneously without impacting the quality or speed of either experience.

Advantages

Some of the advantages of Hyper-Threading technology include the following:

•

Improved overall system performance

•

Increased number of users that a platform can support (server applications)

•

Improved reaction and response time because tasks can be run on separate threads

•

Increased number of transactions that can be executed

3

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

For workstation applications that use a lot of memory intensive, multi-tasked or resource bound tasks, there seems to be no apparent benefit to using Hyper-Threading technology. This will be evident from the results of some benchmarks in this paper. There seems to be more benefit running two physical processors as opposed to running a single processor with Hyper-Threading enabled.

4

Overview

Hyper-Threading technology enables a single physical processor to appear as two independent

Logical Processors to the OS. This enables the OS to execute two separate code streams (called threads) concurrently, either from two different applications or from the same application. After power up and initialization, each logical processor can be individually halted, interrupted or directed to execute a specified thread, independently from the other logical processor on the chip.

Unlike a traditional dual processor (DP) configuration (see Figure 1) that uses two separate physical IA-32 processors (such as two Intel Xenon processors), the logical processors (see

Figure 2) in a processor with Hyper-Threading technology share the execution resources of the processor core, which include the rapid execution engine, the caches, the system bus interface, and the firmware.

Each logical processor has its own set of general purpose registers (including a separate Program Counter and local Advanced Programmable Interrupt Controller [APIC]) but, in order to minimize the complexity of the technology, the Intel Hyper-Threading technology does not attempt to simultaneously fetch/decode instructions corresponding to two threads. Instead, the

Central Processing Unit (CPU) will alternate the fetch/decode stages between the two logical

CPUs and only attempt to execute operations from two threads simultaneously, thus addressing the problem of poor execution unit utilization.

Hyper-Threading is available in a Simultaneous Multi-Threaded (SMT) class processor, which has dual Architectural State

1

. Simply stated, there are two logical processors on one die.

Therefore, two threads can be launched simultaneously on the same processor, which reduces overhead on the thread-switches. The Architectural State, which includes the associated register set for the second logical processor, is only about 5% of the total die area.

1

Architectural State represents the current thread context that consists of the IA-32 registers that are visible to the programmer such as data registers, segment registers, control registers, debug registers, and most of the MSRs as well as its own APIC. The conventional microprocessor such as P3 provide only one set of AS. These single threaded processors are used to support multiple threads application today. However, before another thread can begin, the current thread’s state must be saved in the memory so it can properly resume later. Depending on the number of registers involved and cache misses incurred, a thread-switch operation involving saving and restoring registers can take hundreds of cycles. Consequentially, it is unprofitable to support thread switching on the operations that take less than a hundred or so cycles.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 5

Using Hyper-Threading technology, Instruction-Level Parallelism can boost performance from a processor by utilizing otherwise idle resources. (See Figure 3.) For simplicity, assume that there are only three Executing units: A, B, and C, and only two threads: T1 and T2. In a single processor system, usually the T2 thread has to wait until the T1 thread is finished since the system can only execute one thread at a time. The MP system has more than one processor that the OS can use to launch TLP, so the T1 thread runs on one CPU and dispatches the T2 thread to another

CPU.

Hyper-Threading technology can allow up to two active threads in the pipeline. Since it shares the same rapid execution engine, each thread will need to arbitrate to use the resources in case of conflict. The T1 thread with µop (micro-op) A1 will need to compete with the T2 thread because both threads would have to use the same execution unit A at the same time. Hyper-Threading will give the same priority to each thread and depending on how the code is written, (using Spin-loops

[8]*) one will win over the other. Ideally, it could perform as well as MP system configuration but, on an average, Intel claims that there is a 30% performance boost in some applications.

Note: *Numbers in brackets throughout this paper refer to documents listed in the References section of this paper.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 6

Figure 3

Figure 4 shows the micro-architecture block diagram of the Intel Pentium 4 processor. A thread is initialized by the OS in the same way as in a MP system. There is no distinction in the OS between a logical processor and a true physical processor. The two active threads are interleaved at the instruction fetch and issue stages of the pipeline. (These are the first six blocks on the left in

Figure 4.) The trace cache (decoded instruction cache) has an additional tag bit for each instruction, to represent the thread number (0 or 1). If one thread takes an exception, the instructions from only that thread are flushed. The rapid execution engine (the Scheduler,

Floating–Point (FP-RF), Integer–RF, and arithmetic logic unit [ALU] stages) is competitively shared between two threads with equal priority, which can be an issue when a low-priority task is scheduled simultaneously with a high-priority, time-critical task. The rapid execution engine is the key to the performance gain. Higher utilization of these resources improves the IPC.

Instructions from both threads are simultaneously dispatched for execution by the processor core.

The processor core executes these two threads concurrently, using out-of-order instruction scheduling to keep as many of its execution units as busy as possible during each clock cycle.

Reorder/retire stage again alternates between two logical processors to commit state in program order.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 7

Figure 4:

In a Hyper-Threaded processor, the processor resources are shared so that each thread uses a different resource during each clock cycle. The goal is that the additional thread will fill the voids in the pipeline and take over processor resources when one thread is stalled, while waiting on a cache miss to be filled. With this limitation, some applications will not benefit from the Hyper-

Threading technology.

A majority of single-threaded applications will not benefit. This can be seen from the Sysmark and Winstone Benchmark results. (See the Performance Analyses section for details.)

Additionally, some multi-threaded applications that are not specifically optimized for Hyper-

Threading technology might also show negative or no benefit. For example, the interleaved thread will contend for resources and destroy the effectiveness of the tight code optimizations.

Both threads may also come into conflict over precious space in the trace cache.

Applications such as Internet audio and video streaming, high-performance 3D graphics, digital photography and digital video encoding/decoding, speech recognition, multimedia, MP3 encoding, and database applications do benefit from Hyper-Threading technology. Naturally, these applications are already multi-threaded and each thread is handling a different task that works well with a Hyper-Threaded processor. In addition, these applications can take advantage of the new 142 SSE2 new instructions. These new instructions reduce the overall number of instructions required to execute a particular program task and, as a result, can contribute to an increase in overall system performance.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Compaq Evo Workstations W6000 and W8000

The Compaq Evo Workstations W6000 and W8000 incorporate the new leading-edge Intel Xeon

MP processors and the Intel i860 chipset to achieve the next level of system performance. A detailed discussion of the Xeon processor and the i860 chipset, including performance advantages, can be found in the Compaq white paper [6] 1527-0501A-WWEN (April 2001)

Architectural Comparison of Compaq Evo Workstation W6000/W8000 with Compaq Workstation

AP550/SP750.

Compaq Evo Workstations systems arrive with Hyper-Threading technology disabled. This is done for several reasons.

First, in order to see the real advantage of Hyper-Threading technology, most applications need to be rewritten and recompiled for TLP and ILP. Recompiling with Streaming Single Instruction

Multiple Data [SIMD] Extensions (SSE2) can deliver additional performance gains. Not many of the applications that are available today are recompiled to take advantage of HT. Therefore, the majority of the existing applications will perform better with the two physical CPUs versus two logical CPUs, which Hyper Thread Technology provides.

Secondly, the system with Microsoft Windows 2000 Professional Version and dual physical processors cannot take advantage of Hyper-Threading technology since the Professional version only supports up to two CPUs. Only the server version of Windows 2000 supports Hyper

Threading technology.

Finally, Windows NT does not support Hyper-Threading technology.

IMPORTANT: If a system was originally shipped with one CPU, the system may have an

Advanced Configuration and Power Interface (ACPI) Uniprocessor Hardware Abstraction Layer

(HAL). You will need to switch to ‘ACPI Multiprocessor’ HAL in order to use Hyper-Threading technology.

8

Enabling/Disabling Hyper-Threading

This section outlines the steps to enable and disable the Hyper-Threading technology on Compaq

Evo Workstations W6000 and W8000.

Preliminary Steps

Before enabling or disabling the Hyper-Threading technology, perform these preliminary steps:

1. Check the firmware version at the boot up screen in the lower left corner to ensure that version v01.12 is installed.

Note: If the firmware version is earlier than v01.12, then it is possible that the CPU in the system does not have support for Hyper-Threading technology.

2. Flash the system with the firmware version v01.15 or the latest available on the Web.

IMPORTANT: If you are not familiar with the Compaq ROM-based setup, refer to Appendix A for detailed procedures.

3. Check to see if the ACPI Multi-processor PC is installed on the system. See Appendix B for details to check and switch to a different HAL.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

After completing the preliminary steps, follow these steps to enable or disable the Hyper-

Threading technology on the Compaq Workstation W6000 and Workstation W8000:

1. Reboot the system.

2. When the screen appears with the red COMPAQ logo and the F10 prompt at the bottom of the screen appears, press F10 to enter the setup mode. Once the system completes Power On

Self Test (POST), your system will go into setup mode.

3. When prompted for the language selection, select the appropriate language and press Enter.

4. Move to the Advanced Tab. Select Device Options and Enter.

5. Scroll to the Hyper-Threading option and use the left or right arrow key to toggle between enable/disable.

6. Once the desired option is selected, press F10 to save that option and exit that window.

9

The system saves your configuration and reboots.

Hyper-Threading and Microsoft Windows XP

If the Hyper-Threading technology option is enabled, verify that the change was made. With

Hyper-Threading enabled under Microsoft Windows XP, any physical CPU will be seen by the

OS as two logical CPUs.

1. Verify by opening the Task Manager.

2. Right-click on the task bar, and select Task Manager to bring up the Windows Task

Manager.

3. Under the Performance tab, you will see four CPUs in the Task Manager if your system has two Physical CPUs and Hyper-Threading is enabled.

Figure 5 shows the Task Manager of a system that has two physical Processors with Hyper-

Threading enabled. Figure 6 shows the Task Manager of a system that has either one physical processor with Hyper-Threading enabled or a system that has two physical processors with

Hyper-Threading disabled.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 10

5: Figure

Figure 7 shows the Task Manager on a system that has only one processor with Hyper-Threading disabled.

Figure 7:

Hyper-Threading and Microsoft Windows 2000

Microsoft Windows 2000 Professional Edition only supports up to a maximum of two processors.

The OS can only recognize two CPUs, even-though the system has two physical CPUs that support Hyper-Threading and have Hyper-Threading enabled. The OS does not recognize the case of one physical CPU with Hyper-threading enabled versus the case of two physical CPUs, regardless of Hyper-Threading enabled or disabled. The Task Manager in Figure 8 reflects these three cases. The first case with two logical CPUs will have less performance than the other two cases. Figure 9 shows a single processor system with Hyper-Threading disabled.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 11

.

Performance Analyses

Applications generally need to be tuned and recompiled

2

to gain the maximum performance with

Hyper-Threading technology. Most of the benchmarks that are used to collect the performance data in this paper are horizontal and single threaded. The purpose of this paper is to explain the impacts of processor performance with Hyper-Threading technology enabled versus disabled.

SYSmark 2001

Productivity performance should represent current business usage models of multi-tasking with background computing. The industry standard productivity benchmark SYSmark 2001 incorporates mainstream applications for office productivity and Internet content creation as well as the latest business usage models to reflect platform productivity performance.

SYSmark 2001 is a suite of application software and associated benchmark workloads developed by the Business Applications Performance Corporation (BAPCO), a non-profit consortium of leading computer industry publications, independent testing labs, PC hardware manufacturers, semiconductor manufacturers, and software publishers. SYSmark 2001 is a tool that measures system performance on popular business-oriented applications in using the Microsoft Windows operating environment.

SYSmark 2001 contains fourteen application workloads that are divided into the following two categories:

Office Productivity:

•

Dragon Naturally Speaking Preferred Version 5

•

McAfee Virus Scan 5.13

•

Microsoft Access 2000

•

Microsoft Excel 2000

2

Intel is working to provide Hyper-Threading development tool for software optimization and multi-threaded code that tune for

NetBurst microarchitecture processors with Hyper-Threading Technology

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 12

•

Microsoft Outlook 2000

•

Microsoft PowerPoint 2000

•

Microsoft Word 2000

•

Netscape Communicator 6.0

•

WinZip 8.0

Internet Content Creation:

•

Adobe Photoshop 6.0

•

Adobe Premiere 6.0

•

Macromedia Dreamweaver 4

•

Macromedia Flash 5

•

Microsoft Windows Media Encoder 7

Analysis

SYSmark 2001 is a horizontal benchmark and runs only one application at a time. Therefore only the application thread is running for any given window time. To run the same benchmark on the true MP system does not yield any performance gain or degradation since most of the time the other processor is idle. In this case, when the Hyper-Threading technology is enabled to emulate multi-processing, performance is degraded about 5% since the application thread now is shared and competing with the OS threads for the same processor resources. That does not happen with a true multiprocessor system since each of the processors has its own resources. See Figure 10.

SYSmark 2001 on Windows XP

250

212

222

200 188

200

167

180

150

100

50

0

Enable

Disable

Sysmark Rate

188

200

Figure 10

Internet Content

Creation

212

222

Office Productivity

167

180

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 13

Winstone

Business Winstone 2001

Business Winstone 2001 is a system-level, application-based benchmark that measures the overall performance of a PC when running today’s top-selling Windows-based 32-bit applications on

Microsoft Windows 98 SE, Windows NT 4.0 (SP6 or later), Windows 2000, Windows Me or

Windows XP. Business Winstone 2001 does not mimic what these packages do. It runs real applications through a series of scripted activities and measures the time a PC takes to complete those activities to produce its performance scores. Higher scores mean better performance. When

Business Winstone 2001 runs the test, it runs at least a portion of each application through a script that eTesting Labs developed. The script automatically executes commands within that application with no input from the user.

Business Winstone 2001 uses the following applications in its tests:

•

Norton Antivirus 2000 from Symantec

•

WinZip 7.0

•

Microsoft FrontPage® 2000

•

Lotus Notes R5

•

Microsoft Access 2000

•

Microsoft Excel 2000

•

Microsoft PowerPoint 2000

•

Microsoft Project 98

•

Microsoft Word 2000

•

Netscape Communicator 4.73

Content Creation Winstone 2001

Content Creation Winstone is a system-level, application-based benchmark that measures a PC's overall performance when it is running under a 32-bit OS, such as Windows 2000 or Windows

XP. Content Creation Winstone 2001 uses the following applications:

•

Adobe Photoshop 5.5

•

Adobe Premiere 5.1

•

Macromedia Director 8.0

•

Macromedia Dreamweaver 3.0

•

Netscape Navigator 4.73

•

Sonic Foundry Sound Forge 4.5

Following the lead of real users, Content Creation Winstone 2001 keeps multiple applications open at once and switches amongst those applications. Content Creation Winstone 2001 is a single large test that runs the above applications through a series of scripted activities and returns a single score. Those activities focus on so called "hot spots," periods of activity that make your

PC really work—the times where you are likely to see an hourglass or a progress bar.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Analysis

The Winstone test results are the same as the results for SYSmark 2001. No performance loss or gain is seen with Hyper-threading enabled. See Figure 11.

14

Winstone Benchmark on Windows XP

0

Enable

Disable

40

30

20

10

80

70

60

50 49.4

49.5

Business

49.4

49.5

69

71

Content Creation

69

71

Figure 11

Pro/ENGINEER 2001i²

The Standard Performance Evaluation Corporation (SPEC) SPEC/GPC Application Performance

Characterization (SPECapc) project group offers performance results and free downloads for a benchmark based on Pro/ENGINEER 2000i² (ProE 2000i²). However, the results for SPECapc for Pro/E 2000i² cannot be compared to those of previous versions of the benchmark. The model used in the benchmark is a realistic rendering of a complete photocopy machine consisting of approximately 370,000 triangles. See Figure 12.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 15

ProE2001i2

2

1.5

1

0.5

5

4.5

4.09

4.11

4

3.5

3

2.5

3.81

3.96

4.69

4.61

4.21

4.27

4.26

4.18

0

Enable

Disable

Figure 12

Composite

Score

4.09

4.11

Wireframe

Composite

3.81

3.96

Shaded

Composite

4.69

4.61

Sub Assembly

Composite

4.21

4.27

Other

Composite

4.26

4.18

The benchmark is comprised of 17 tests. Startup and initialization time is measured, but given no weight (0.0) within the composite score for the benchmark. There are 16 graphics tests, each of which measures a different rendering mode or feature. The first three graphics tests measure wireframe performance using the entire model. The next four measure different aspects of shaded performance, using the same model. Each of these tests executes exactly the same sequence of 3D transformations to provide a direct comparison of different rendering modes. The next four tests use a sub-assembly, and compare the two FASTHLR modes, the default shading mode, and shading with edges. These tests also execute a common sequence of 3D transformations. The last five graphics tests use two different instances of the model—the first three without its outer skins

(to illustrate the effect of FASTHLR and level-of-detail operations), and the last two to illustrate complex lighting modes and surface curvature display. The last test is an aggregate of all time not accounted for by the previous 16 tests, and is a mix of CPU and graphics operations.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Scores are generated for all 17 tests. Composite numbers are provided for each set of graphics tests (shaded, sub-assembly, wire-frame, and other) and there is an overall composite score for graphics and CPU operations. Start-up and initiation time is not included in the composite score.

You must have a 3D graphics display device recognized by Pro/E 2000i² in order to run the benchmark. A fully licensed, released version of Pro/E 2000i² is required. If a floating license is used, the workstation's network must be configured as documented in the Pro/ installation guide.

16

Analysis

ProE 2002i2 is not a multi-threaded application but does contain different tasks for comparing.

Wireframe and sub-assembly, which are more computationally intensive, take more advantage of

Hyper-Threading technology by utilizing the idle processor resources. Otherwise, shading, texturing, and Wireframe, which are more memory intensive, see a negative performance impact due to competitively shared caches. A trace cache miss results in a memory latency that forces the execution unit to use the X86 instruction de

FRGHULQVWHDGRIWKH RSVPLFUR

-operation) that are already decoded in the trace cache.

Excel Benchmark

Excel benchmark uses the Microsoft Excel application to measure the performance of computer system. Like other benchmarks, the Excel benchmark measures performance by managing a large amount of data. A lower score indicates better system performance. The Excel benchmark includes four tasks. Each task will have roughly about 40MB of data. Excel benchmark will run four tasks at the same time from a batch file. When running from a batch file, there will be four

Excel windows open. The first Excel window does the sorting. All columns are sorted in ascending and descending order 15 times. The second Excel window does the multiplication.

Two sheets are multiplied together 400 times. The third Excel window does the Cosine of data in a nested loop 24 times and sorts the data in ascending and descending order. The fourth Excel window sorts the amount of data in ascending and descending order 22 times. When all four tasks finish running, the run time of each task is recorded. The longest time is chosen for performance comparisons.

Analysis

Excel benchmark is a multi-task application and each task utilizes different resources from each other. This type of application is ideal for Hyper-Threading technology. Performance gain from a single physical processor to dual physical processors is about 80% since dual physical processors do not have to share and compete for the same cache. The result is that the system has twice as large a cache when compared to the single physical processor with Hyper-Threading enabled. For a single physical processor with Hyper-Threading enabled, there is a 30% performance gain, whereas an 18% performance increase in a dual physical processor case is realized with Hyper-

Threading enabled. This is because the application is not as computationally intensive as others for utilizing all of the processor resources. See Figure 13.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Excel Benchmark on Windows XP

35.00

30.00

25.00

20.00

15.00

10.00

23.05

29.78

14.17

16.68

5.00

0.00

Enable

Disable

Figure 13:

1P

23.05

29.78

2P

14.17

16.68

Solid Edge V10

SPECapc for Solid Edge V10 was developed by EDS PLM Solutions in cooperation with the

SPECapc group. It represents typical user operations that are valuable in evaluating the performance of systems running Solid Edge. The new benchmark runs on Microsoft Windows

NT and Windows 2000. Two Solid Edge models are used for the benchmark:

•

A well-head assembly with 1,320 parts

•

A 1.3-million polygons and a 64-part hydraulic jack assembly with 66,000 polygons

Solid Edge V10 measures the following four basic areas of performance:

•

Graphics—measures commonly used commands in three modes: o Smooth shading o Wireframe o Smooth shading with textures and reflections

•

File I/O—measures opening and saving the well-head assembly model using three different display configurations: o All inactive o All active—a combination of active, inactive, and hidden parts o CPU—exercises real-time VHL representation and re-computing an assembly

•

Workflow—represents a typical set of operations that users will execute during the course of creating a project: o Real-time drawing view placement

17

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 18 o In-place assembly activation o In-place part activation o Modification of a model o Update of all the files due to the model change

Note: EDS PLM Solutions consider all four portions of the benchmark to be equally important in evaluating performance using Solid Edge, so they have been given equal weightings when calculating the geometric mean.

Analysis

From this benchmark, the results show that the SolidEdge V10 CPU and graphics tests are computation intensive and will benefit from the Hyper-Threading technology. The file I/O test again has to share and compete for resources with the alternate threads, including the OS threads that require file I/O operations and thus performs poorly. See Figure 14.

5

SolidEdge V10 Benchmark on Windows XP

4.68

4.5

2.5

2

1.5

1

0.5

0

4

3.5

3

1.82

2.16

1.64

1.92

1.38

1.32

3.14

Enable

Disable

Figure 14:

Overall

1.82

2.16

CPU

1.64

1.92

File I/O

1.38

1.32

Graphics

3.14

4.68

Montecarlo Benchmark

The Montecarlo utility includes four independent tasks that run at the same time. Each task contains four spreadsheets. Each spreadsheet contains different Montecarlo functions, formula, and graphic calculations.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Analysis

The Montecarlo application is multi-tasked as well as computationally intensive. A task could utilize all of the processor resources, which is determined by the fact that the ratio between the physical 1P (30.1) and physical 2P (15.15) is almost 2:1. Independent of the 1P or 2P case,

Hyper-Threading technology does not yield any performance gain since the benchmark is resource bound. See Figure 15.

Montecarlo Benchmark on Windows XP

19

35.00

30.00

25.00

20.00

15.00

10.00

5.00

0.00

Enable

Disable

30.10

30.27

1P

30.10

30.27

15.15

15.00

2P

15.15

15.00

Figure 15:

Verilog Compiled Simulation Benchmark

The Verilog Compiled Simulation (VCS) benchmark script includes four independent tasks that run at the same time. Each task simulates the same reset sequence of a memory controller. The

VCS benchmark is very memory intensive.

Analysis

This VCS application is multi-tasked as well as memory intensive. A task requires about 20MB of data for simulation. The data cache is shared between the two threads, which results in frequent cache line replacement and hence shows poor performance when Hyper-Threading is enabled. In a two physical processor case with Hyper threading disabled, with cache available on each processor, there is a significant improvement in performance. With Hyper-Threading technology enabled in a dual processor configuration, the performance is worse than just a true dual processor configuration. This poor result is due to the sharing of resources between the same processor and checking the other processor’s cache validity for correct data. See Figure 16.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 20

VCS Benchmark on Windows XP

8

7

6

5

4

3

2

1

5.2

6.53

7.1

7.2

0

Disable

Enable

Figure 16

2P

5.2

6.53

1P

7.1

7.2

Summary

Hyper-Threading technology represents a new approach to improving the instruction throughput of processors (instruction-level parallelism) that is targeted for servers and high-performance workstations. One of the principal benefits of Hyper-Threading technology is that it enables the processor to use on-die resources that may otherwise be idle. Although existing operating system and application code will run correctly on a processor with Hyper-Threading technology, to obtain performance gain, some relatively simple code modifications are needed to get the optimum benefit from Hyper-Threading technology. To refrain from wasting resources with the idle loops, the OS needs to issue immediately the Halt instruction to the logical processor that has a thread going idle. A thread that is using spin lock code, which aggressively captures processor resources, may need to add the pause instruction within the loop to allow the other threads a fair access to the processor resources. The limited trace cache size requires tight code loops in the trace cache for Hyper-Threading technology to be effective. Programmers now avoid loop unrolling, as it increases the amount of code created and increases trace cache misses. A trace cache miss forces the execution to the X86 instruction decoder, which can only decode one X86 instruction per cycle.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 21

On the down side, the existing single threaded applications do not experience the benefit of

Hyper-Threading technology. In fact, there can sometimes be disruption by the alternative thread, such as the thread that uses a spin lock without pause instruction. Multi-threaded applications that are optimized for MP systems also need to consider resource contention that destroys the effectiveness of hand-tuned highly optimized code (games and streaming media applications).

Hyper-Threading technology can provide a performance gain of up to 30% when executing under multi-threaded OS and applications with code that is tuned for Hyper-Threaded3 technology over a comparable IA-32 processor without Hyper-Threading technology. For workstation applications that are more multi-tasked and more memory intensive, there is more benefit to using two physical processors as compared to using a single Hyper-Threading technology enabled processor.

3

Tuned instructions for Hyper-threaded is minimize the resource contention

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 22

Appendix A — Enable/Disable Hyper-Threading

IMPORTANT: Check and verify that your system was flashed with the latest Compaq

ROMPAQ version v01.15 or later, as shown in the lower left corner at the boot screen. Download the latest ROMPAQ at the Compaq website http://www.compaq.com

To enable or disable the Hyper-Threading technology on the Compaq Evo Workstations W6000 and W8000, follow these steps:

Figure 17:

1. When you power-up the system, you see the screen shown in Figure 17. Press F10 to enter the setup option. After a few seconds, your screen goes into the setup mode.

2. Select the language according to your region by using the Up or Down arrow key. In this case, select English and press the Enter key. See Figure 18.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 23

Figure 18:

Figure 19:

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 24

Figure 20:

4. Use either the Up or Down arrow key to select Device Options. See Figure 20.

Figure 21:

6. Use the Up or Down arrow key to move the white arrow cursor to select Hyper-Threading.

See Figure 22.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 25

Figure 22: depending on the option that you choose. See Figures 23 and 24.

Figure 23:

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 26

Figure 24:

Figure 25:

10. Press Enter to save any changes. Press F10 again. See Figure 26.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 27

Figure 26:

11. Your system saves the configuration and is ready for reboot. See Figure 27.

Figure 27:

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Appendix B — Verify HAL Version

This section will describe how to verify which version of the HAL has been installed on your system and how to switch to ‘ACPI Multiprocessor PC’ HAL. The procedure is for Windows

2000.

Note: The procedure for Windows XP is not included here since it is similar to the Windows

2000 procedure.

Follow these steps to verify which version of the HAL has been installed on your system:

1. At the Desktop screen, right click on My Computer and move the mouse to the Properties bar. See Figure 28.

28

Figure 28:

1. Click with the left mouse button to bring up the System Properties window. See Figure 29.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 29

Figure 29:

3. Click on the Hardware tab and Device Manager button to bring up the Device Manager window. See Figure 30.

Figure 30:

4. Double click on the Computer Icon. If you see “ACPI Multiprocessor PC” is installed, then the system is ready.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

Note: You could stop by closing all open windows and returning to the Desktop.

If you see “ACPI Uniprocessor PC,“ then follow this procedure:

1. Right click on ACPI Uniprocessor PC and click on the Properties bar to bring up the ACPI

Uniprocessor PC Properties window. See Figure 31.

30

Figure 31:

2. Click on the Driver tab and select Update Driver button to bring up the Upgrade Device

Driver Wizard window. See Figure 32.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 31

Figure 32:

Figure 33:

4. Select the second option to Select a Device Driver and click Next. See Figure 34.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 32

Figure 34:

Figure 35:

3. The Upgrade Device Driver Wizard window prompts you that the new device is ready to install; in this case, it is the ACPI Multiprocessor PC kernel. Click the Next button to continue.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 33

Figure 36:

Figure 37:

6. At this time, the system is ready for rebooting. Click Yes to restart the system. See Figure

38.

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper 34

Figure 38:

167T-0202A-WWEN

Hyper-Threading Technology, New Feature of Intel Xeon Processor White Paper

References

[1] Introduction to Hyper-Threading Technology, Intel Document Number: 250008-002

[2] Laudon J., Gupta A., Horowitz M. (1994) Interleaving: A Multi-threading Technique

Targeting Multi-processors and Workstations

[3] Smith B. J. (1985) The Architecture of HEP

[4] Thistle M., Smith B.J. (1988) A Processor Architecture for Horizon

[5] Alverson R., Callahan D., Cummings D., Koblenz B., Porterfield A., Smith J.N. (1990) The

Tera Computer System

[6] -0501A-WWEN (April 2001) Architectural Comparison of Compaq Evo Workstation

W6000/W8000 with Compaq Workstation AP550/SP750

[7] Tullsen D.M., Eggers S.J., Levy H.M. (1995) Simultaneous Multi-threading: Maximizing On

Chip Parallelism

[8] AP-949 Using Spin-loops on Intel Pentium 4 Processor and Xeon Processor

35

167T-0202A-WWEN