Compaq BL10e - HP ProLiant - 512 MB RAM Technical manual

ISS Technology Update Volume 8, Number 4
1
Recently published industry standard server technology papers ...............................................................1
Optimizing memory performance in ProLiant G6 Intel-based servers ........................................................ 2
Unbuffered DDR-3 DIMMs are a cost-effective solution for ProLiant G6 servers .........................................5
The transition from Megahertz to Gigatransfers ........................................................................................6
Meet the Expert—Robert Elliott................................................................................................................... 7
Contact us ..................................................................................................................................................8
Recently published industry standard server technology papers
Title
URL
Technology and architecture of HP ProLiant 300–series
generation six servers technology brief
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c005026
16/c00502616.pdf
Implementing Microsoft Windows Server 2008 R2
Release Candidate on HP servers integration note
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c016395
94/c01639594.pdf
Implementing Microsoft Windows Server 2008 Hyper-V™
on HP ProLiant servers integration note
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c015161
56/c01516156.pdf
Technology and architecture of HP ProLiant 100–series
G6 (Generation 6) servers technology brief
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c017517
18/c01751718.pdf
Implementing Microsoft Windows Server 2008 SP2
Release Candidate on HP servers integration note
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c007106
06/c00710606.pdf
Serial Attached SCSI storage technology, 2nd Edition
technology brief
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c016134
20/c01613420.pdf
Industry standard server technical papers can be found at www.hp.com/servers/technology.
ISS Technology Update
Volume 8, Number 4
Keeping you informed of the latest ISS technology

ISS Technology Update Volume 8, Number 4
2
Optimizing memory performance in ProLiant G6 Intel-based servers
To achieve higher performance and better efficiency, the new HP ProLiant G6 servers featuring the Intel® Nehalem architecture
use a memory architecture that is significantly different than previous generations of servers.
Comparing traditional and non-uniform memory architecture
In traditional server memory architecture, memory controllers are part of the server chipset, and physical memory is attached to
these through the memory channels. This is sometimes referred to as a Uniform Memory Architecture since all processors in the
system have exactly the same speed access to all of system memory through the front-side bus, the system chipset, and the
memory controllers (Figure 1-1).
Figure 1-1. Traditional system architecture
Uniform memory architectures worked well. As system complexity increased, however, new challenges emerged. In order to
support large system memory footprints, the system memory controllers needed to support 4 or more DIMM’s per memory
channel. The electrical loading requirements this created led to the use of fully buffered DIMMs. This solved the electrical
loading problems but imposed significant additional burdens in terms of memory costs and power consumption. Overall
bandwidth between the processors and memory was also limited by the bandwidth of the front-side bus.
The Intel Nehalem architecture (as well as AMD Opteron architecture) has addressed these issues by using a Non-Uniform
Memory Architecture (NUMA). With NUMA, each processor has its own memory controller, memory channels, and directly
attached DIMMs, allowing it to directly access this memory at maximum bandwidth with no intermediate transport mechanism
required (Figure 1-2). The memory attached to a given processor is still a part of the overall system memory. Other processors
can access it over the new Intel®QuickPath®Interconnect links that exist between processors as well as from processors to the
system chipset.

ISS Technology Update Volume 8, Number 4
3
Figure 1-2. Intel Nehalem architecture with NUMA
Understanding and optimizing memory performance in G6 servers
With Intel Nehalem architecture, each processor has its own memory controller and three memory channels that it controls
directly. In its fastest configuration, the new DDR3 memory used with these systems operates at 1333 Megatransfers per
second. Since each transfer consists of 8 data bytes, the maximum raw throughput per memory channel can be easily
calculated.
1333 MT/s x 8 Bytes per transfer = 10667 MB/s = 10.667 GB/s raw throughput per channel
This is actually where the 10600 in PC3-10600 comes from, representing the maximum throughput for the DIMM in round
numbers.
Figure 1-3 shows both a physical and a logical block diagram of the Processor/Memory complex for a 2P ProLiant G6 system
with 12 DIMM slots and 6 total memory channels. The key to optimizing system memory performance is to install memory so
that it maximizes the number memory channels in use which maximizes the composite bandwidth.

ISS Technology Update Volume 8, Number 4
4
Figure 1-3a. Diagram of physical arrangement of 12 DIMM sockets in an HP ProLiant Intel-based G6 system with 2 processors
DIMM 1
DIMM 2
DIMM 3
DIMM 4
DIMM 5
DIMM 6
Ch 0 Ch 1 Ch 2
CP U 0
Ch 0 Ch 1 Ch 2
CP U 1
DIMM 1
DIMM 2
DIMM 3
DIMM 4
DIMM 5
DIMM 6
Figure 1-3b. Logical block diagram of HP ProLiant Intel-based G6 system with 2 processors and 12 DIMM slots
DIMM 1
DIMM 2
DIMM 3
DIMM 4
DIMM 5
DIMM 6
Ch 0 Ch 1 Ch 2
CP U 0
Ch 0 Ch 1 Ch 2
CP U 1
DIMM 1
DIMM 2
DIMM 3
DIMM 4
DIMM 5
DIMM 6
With six DIMMs available, the most optimal configuration is to populate DIMM slots 2, 4, and 6 of each processor with one
DIMM. Doing this populates all six memory channels, resulting in a potential composite bandwidth of 64 GB/s. If all three
DIMMs attached to each processor are identical, the processors’ memory controllers can use channel interleaving to map each
consecutive 8 Bytes of the system’s logical memory map to a different physical memory channel. This increases the probability
that memory accesses will be spread out more evenly across the memory channels and that the potential composite bandwidth
can be achieved.
Populating the system in these ―logical rows‖ and in groups of three or six DIMMs helps create balanced memory
configurations that can help maximize system throughput. When configuring ProLiant G6 Intel-based servers, observing the
following general rules should maximize system performance:
When possible, keep memory balanced across the memory channels of a processor (for example, three of the same DIMMs
in a logical row).
When installing memory on a 2P system, keep memory balanced across CPUs.
Mixing DIMM SKUs within a memory channel should not adversely affect memory throughput. However, mixing DIMM
SKUs across memory channels in a logical row will disrupt channel interleaving and potentially affect overall performance.

ISS Technology Update Volume 8, Number 4
5
Unbuffered DDR-3 DIMMs are a cost-effective solution for ProLiant G6 servers
HP ProLiant G6 servers support Double Data Rate-3 (DDR-3) Synchronous Dynamic Random Access Memory (SDRAM)
technology. DDR-3 SDRAM technology doubles the peak bandwidth of DDR-2 when operating at the same memory bus
frequency. Memory manufacturers are producing two types of DDR-3 SDRAM DIMMs: Registered DIMMs (RDIMMs) and
Unbuffered DIMMs (UDIMMs).
ProLiant G6 servers support single-rank, dual-rank, and quad-rank RDIMM configurations. RDIMMs contain a register that
buffers the address and command signals from the memory controller. The register also reduces the electrical load on the
memory controller, allowing up to 3 DIMMs per channel (DPC), higher capacity per DIMM (2, 4, or 8GB), and faster bus
speeds than UDIMMs. However, the register also increases the power requirements of each RDIMM by almost 1 Watt and adds
1 bus clock of latency. RDIMMs are identified with an R suffix in the module manufacturer’s name (example PC3-8500R).
Select ProLiant G6 servers support both single-rank and dual-rank UDIMM configurations. UDIMMs use less power and cost less
than comparably sized RDIMMs; however, the capacity of UDIMMs is limited to 1GB or 2GB. ProLiant G6 servers that support
UDIMMs, support a maximum of two UDIMMs per channel, which allows a maximum memory capacity of 24 gigabytes using
dual-rank 2-gigabyte DIMMs. UDIMMs are a very cost-effective solution for servers with memory capacity of 24 GB or less.
ProLiant G6 servers ship standard with RDIMMs, which are required to reach the maximum capacity of each server. Select
ProLiant G6 servers support both RDIMMs and UDIMMs; however, RDIMMs and UDIMMs cannot be used together in a system.
Customers should consider using UDIMMs when cost and power consumption are top priorities. HP simplifies memory selection
by providing an on-line ProLiant Memory Configuration Tool (www.hp.com/go/ddr3memory-configurator) to help configure
each server’s memory and provide an orderable parts list.
Table 1. Comparison between UDIMMs and RDIMMs in HP ProLiant G6 servers
Table head
UDIMMs
RDIMMs
Maximum speed supported
DDR3-1333
DDR3-1333*
Ranks per DIMM
1 or 2
1, 2, or 4
Maximum DIMMs per channel
2 dual rank DIMMs
3 dual rank DIMMs
Maximum memory capacity
18-slot servers
12-slot servers
24 GB
24 GB
144 GB
96 GB
*Maximum speed for quad-rank RDIMMs is DDR3-1067
Additional resources
For additional information on the topics discussed in this article, visit:
Resource
URL
Memory technology evolution: an
overview of system memory
technologies, 8th Edition
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c
00256987/c00256987.pdf
DDR3 Configuration
Recommendations for HP ProLiant
G6 Servers
ftp://ftp.hp.com/pub/c-products/servers/options/Memory-Config-
Recommendations-for-Intel-Xeon-5500-Series-Servers-Rev1.pdf

ISS Technology Update Volume 8, Number 4
6
The transition from Megahertz to Gigatransfers
Traditional front side bus architecture
The front side bus (FSB) has been the traditional interconnect between processors and the system chipset. The FSB is
unidirectional and it is shared by all of the devices (processors and chipset) that are attached to it. The FSB in Intel processor-
based systems is clocked at a frequency (speed) of 266 MHz, 333 MHz or 400 MHz. However, the processors and chipsets
deliver a new set of 64 bits onto the bus at four times per clock cycle, referred to as ―quad-pumping.‖ Therefore, a more
accurate way to express the speed of the bus is to use the number of data transfers per second. For example, a quad-pumped
400-MHz bus delivers 4 transfers/clock cycle × 400 million cycles/sec, or 1600 Megatransfers/sec. This number is sometimes
incorrectly quoted as megahertz (MHz).
Point-to-Point link-based architectures
AMD Opteron ™based HP ProLiant servers and Intel®Microarchitecture Nehalem-based HP ProLiant Generation 6 (G6) servers
have eliminated the FSB in favor of bi-directional, point-to-point links that connect the processors to each other and each
processor to the chipset. These links (the AMD HyperTransport™ and the Intel®QuickPath Interconnect) allow each processor to
independently send data to, or receive data from, another processor or the chipset.
Both the AMD HyperTransport and Intel QuickPath Interconnect (QPI) have bi-directional links that are double-pumped (two data
transfers per clock cycle). The AMD HyperTransport 3.1 has scalable links up to 16 bits (2 bytes) wide each, and it supports
data rates up to 6.4 GT/s in each direction. The Intel QPI also uses a maximum of 16 bits per link to transfer data, and it
currently supports data rates up to 6.4 GT/s in each direction.
Bandwidth –the ultimate yardstick
To effectively compare the unidirectional, shared FSB with bi-directional, point-to-point links, we need to use a common metric—
the theoretical maximum bandwidth. Bandwidth is calculated by multiplying the data rate by the width of the bus or link.
For example, the bandwidth of the 400-MHz, quad-pumped FSB mentioned earlier is calculated as follows:
Maximum bandwidth = (1600 MT/s x 8 bytes per transfer) / 1000 = 12.8 GB/s
In the case of the AMD HyperTransport 3.1 and Intel QPI, bandwidth is calculated is as follows:
Maximum bandwidth = 6.4 GT/s x 2 bytes per transfer x 2 links per interconnect
= 25.6 GB/s (12.8 GB/s maximum in each direction)
Additional resources
For additional information on the topics discussed in this article, visit:
Resource
URL
Introduction to Intel QuickPath
Technology
www.intel.com/technology/quickpath
HyperTransport (HT) Consortium
homepage
www.hypertransport.org

ISS Technology Update Volume 8, Number 4
7
Meet the Expert—Robert Elliott
Robert Elliott is a Master Architect with the HP Industry Standard Server Platform
Architecture Team. Over his 15-year career, he has become widely respected
within HP and in the server storage industry. Rob represents HP in the T10
Technical Committee of the InterNational Committee on Information Technology
Standards (INCITS, pronounced "insights‖), and he has served as the editor of
several standards including Serial Attached SCSI (SAS).
Gene Freeman, manager of the ISS Platform Architecture Team, believes that
Rob’s success is largely due to his ability to assimilate and synthesize massive
amounts of information and then recall those elements that are relevant to solving
a particular problem.
An explorer
Rob’s hobbies include hiking and go-karting. His hiking treks have included
Mt. Kilimanjaro in Tanzania, the Daikiretto in the Japanese Alps, the Inca Trail to
Machu Picchu in Peru, Mt. Rainier in Washington, and about a dozen of the
Colorado 14ers (peaks with elevation greater than 14,000 ft.).
A significant contributor to SAS technology
As editor for the T10 Technical Committee, Rob led the industry's development of
SAS since its inception in July 2001. In April 2009, Rob received the INCITS
Service Award, which recognized his contributions to the INCITS/T10 technical
committee in the development of the SAS standards. The award cited Rob’s
―tremendous skills and commitment to standardization efforts.‖
A customer advocate
Rob’s participation in ISS Technology Exchange events and the ISS Technology Advisory Community (TAC) has given him
detailed insight into customer usage models for storage. The knowledge and experience he gains helps engineering and
marketing to develop features that customers desire. He says that the TAC feedback has been very helpful in planning products
that are a few years out, while the ISS Tech Exchanges provide good feedback for short-term planning.
A valuable resource for HP’s storage leadership
Rob states that the technical expertise and customer feedback ISS receives has paid great dividends; HP Smart Array RAID
controllers lead the industry in reliability, performance, and features. SAS technology has enabled HP to build servers and
storage systems that let customers choose whether to deploy enterprise-class SAS drives or entry-level SATA drives. Customers
benefit by not being locked into a single storage solution. SAS technology has enabled the introduction of small form factor
(SFF) 2.5-inch drives into the server market. The SFF design is more versatile and will be a better form factor for solid state
drives (SSDs).
Name: Robert Elliott
Title: Master Architect, ISS Platform
Architecture
Years at HP: 15
University: BS Computer Engineering,
University of Illinois at Urbana-
Champaign
U.S. Patents: 16 issued, 5 active
applications

ISS Technology Update Volume 8, Number 4
Legal Notices
© Copyright 2009Hewlett-Packard Development Company, L.P. The information contained herein is subject to
change without notice. The only warranties for HP products and services are set forth in the express warranty
statements accompanying such products and services. Nothing herein should be construed as constituting an
additional warranty. HP shall not be liable for technical or editorial errors or omissions contained herein.
AMD and AMD Opteron are trademarks of Advanced Micro Devices, Inc.
Intel, Intel Xeon, and Intel Itanium are trademarks of Intel Corporation in the United States and other countries.
TC090610NL
8
Contact us
Send comments about this newsletter to TechC[email protected]m.
To subscribe to the ISS Technology Update newsletter, click mailto:[email protected]?subject=newsletter_subscription