I was in hopes of sharing some power and benchmark data with you on an advanced AMC processor card that we are developing using the Penryn SFF processor and the small form factor Intel GS45 Express chipset in this blog. However, that will have to wait until next time. Implementing some IPMI-related code specified in the MicroTCA PICMG specification is giving us fits and a less than cooperative vendor isn’t helping our product release schedule one little bit. Ah well that’s life in processor board development land! We now have the situation under control so I should be able to share the test and product qualification data with you in my next blog.

 

In this blog I would like to share with some interesting information that compares a single board computer with two single core processors to an SBC using one quad-core CPU. We’re finding that more of our customers are asking questions such as; “Does a single quad-core CPU outperform an SBC with two, single core processors?” Of course we all know the correct answer is “It Depends…” but the following data tells an interesting story.

 

Benchmark Parameters & Board Specifications

First let me set the stage. This OEM customer has used our dual processor NLT board for a number of years and their end-user customers are very happy with the performance results in their specific application. The Trenton NLT board uses the Nocona-Lindenhurst platform. The OEM was looking for a way to increase performance, reduce cost while increasing the cooling and power efficiency of the system design. The Trenton MCXI board has a single processor with the Intel 5000P (i.e. Blackford) chipset and supports a variety of multi-core processor options. Here are the specifics of each board and the benchmark test parameters used in the performance comparison:

 

Board & Test Specs.

Trenton NLT, PICMG 1.3 SBC/SHB

Trenton MCXI, PICMG 1.3 SBC/SHB

CPU Number

Two

One

CPU Desc.

Single Core, 64-bit Intel Xeon (Nocona)

Quad-Core Intel Xeon Processor E5440 (Harpertown)

Core Speed

3.4GHz used in testing, customer actually uses the embedded 3.2GHz/1MB L2 Cache CPU in the application

2.83GHz (embedded CPU)

L2 Cache

2MB

12MB (2x6MB shared)

TDP / Tcase (max)

110W / 72° C

85W / 67° C

FSB

800MHz

1333MHz

O/S

Windows XP, SP2

Windows XP, SP3

HDD & Interface

Maxtor D70 X / UltraATA 100

WD800 / SATA II 300

System Memory

2GB / DDR2-400 DIMMs

4GB / FB-DIMMs DDR2-667

 

Benchmark Test Data

Here is a table with the raw data numbers obtained in our benchmark comparison tests:

Benchmark Test
Program

Specific Test
Name

Trenton NLT SBC
TWO - 3.4GHz with 1MB L2 Cache - Xeon (Nocona)
800MHz - System Bus

MCXI with ONE Intel Quad Xeon Harpertown 2.83GHz, 1333MHz FSB, 12GB L2 Cache (E5440), 4-1GB DDR2-667 FB DIMMs

PCMark 2004/2005

CPU Bench

 


 

Memory Bench

3463

4856

 


 


  

SISoft Sandra
2005

CPU Arithmetic
Drystone

19873

67369

CPU Arithmetic
Whetstone

13471

28588

CPU Multi-Media
Integers

49018

149222

CPU Multi-Media
Floating Point

63850

168665

Memory Bandwidth
Integers          

3586

4255

Memory Bandwidth
Floating Point

3580

4255

 


 


  

Winstone 2004

Business Suite Benchmark

22

26.4

 

Test Results

Of course benchmark data is very subjective and can vary from system-to-system and application-to-application. System memory, HDD type and software can be modified to change the benchmark test results; however, this basic set of data confirms that a single board computer using a more recent quad-core processor should be able to out perform a legacy dual processor board with two single core processors in most applications.

 

System Design Savings

Intel has done a great job in recent years of improving the thermal efficiency of high-performance processors while increasing computing performance capabilities with the introduction of multi-core processors. In this application the customer saves in a variety of ways:

>>    Lower CPU thermal design power (TDPs) results in less costly system fans

>> Tcase ratings are maintained in order to take advantage of these lower TDPs

>> Changing to a single, quad-core processor on a single board computer will result in the customer saving about 16.5\% in SBC costs

>> Power supply simplification will also lead to additional system cost savings

 

 

 

Next time look for some design information and benchmark data on the upcoming Trenton MCP6792 AMC Processor card featuring the Intel Core 2 Duo SL9400 processor and the card’s SFF Intel GS45 Express chipset.

 

 

Thanks,

Jim

 

 

Message Edited by serenajoy on 03-11-2009 08:27 PM
Message Edited by pmahler_intel on 03-12-2009 08:59 AM