Newsletters   Subscriptions  Forums  Store   Career  Media Kit  About Us  Contact  Search   Home 
tlb
Volume 2, Number 32 -- August 23, 2005

AMD Nabs Chip Hotshot, Challenges Intel to Dual-Core Duel


by Timothy Prickett Morgan


The Intel Developer Forum may be raging in San Francisco this week, but Advanced Micro Devices did its best to try to steal a little thunder away from its chip-making rival by announcing it had hired away yet another hotshot system designer from IBM to help it plan its future chip cores, system architectures, and manufacturing processes. AMD is also talking a tiny bit about its more near-term multicore chip strategy.

AMD has lured Jeff VerHeul, a 25-year IBMer, to be its new vice president of silicon design. VerHeul most recently created IBM's fledgling Engineering and Technology Services business, which sells as a service the kind of chip-making expertise that IBM develops for its various Power processors. In fact, AMD is one of IBM's flagship customers for the ETS unit, which is probably how AMD came to know him. VerHeul will lead the development of all future AMD computing solutions, according to a company statement, including chip roadmaps and design projects across all of AMD's engineering sites. AMD keeps the name of its key technical people out of the limelight, and the company didn't want to discuss who VerHeul replaced, and seemed only to bring it up as a means to show that a very serious systems and chip guy was now in charge of the AMD product lines out beyond 2007. Back in the late 1990s, when IBM was working on the Regatta server platform and the Power4 chip, the first dual-core processor brought to market, VerHeul was vice president of server and workstation development at IBM's Server Group. It doesn't get more serious than that.

VerHeul was undoubtedly a key player in the inking of a technology partnership between IBM and AMD that has allowed the latter to accelerate its adoption of advanced chip-making processes and improve its yields--a key differentiator that has allowed it to fight tooth-and-nail against Intel.

Back in April, AMD announced that Rich Oehler, a former IBM system designer who was working for Newisys on a 32-way chipset for the Opterons code-named "Horus," had been enticed away from that subsidiary of Sanmina-SCI. Oehler also has a pretty strong pedigree in that he was the lead designer for IBM's Power family of chips for many years and was also one of the key designers of the "Summit" family of chipsets Big Blue created to make scalable, SMP-NUMA hybrid servers based on Intel's Xeon and Itanium processors. Oehler has been working on processor and server designs at IBM for decades, and is still at AMD.

While not being specific about what VerHeul and Oehler are up to, Pat Patla, director of server and workstation marketing at AMD, said that VerHeul will be taking over the development of the next generation of AMD cores and system technologies, and by this he did not mean the multicore chips that AMD has in the pipeline for 2007 or so. "These are the things that are still on the white board, things that don't even have a code name and certainly not a part number yet," explained Patla. He said that Oehler has been working on system designs, and reiterated the obvious fact that with the Opteron chip having an embedded memory controller and HyperTransport links for ganging them up together into SMP and NUMA clusters, AMD has had to think in terms of systems.

Patla would not be drawn out too much on AMD's multicore plans, except to say that the Opteron processors were designed to support more than two cores from the get-go and that AMD would ship processors with more than two cores in 2007. When asked if he believed that AMD would get to multicores (he would not say four cores, but this is obviously what everyone expects) ahead of Intel, which is also slated to roll out multicore Itaniums and Xeon in the 2007 timeframe, Patla laughed and said, "We can only hope so." He didn't sound like a man who was laughing because he was worried.


What AMD seems to be mulling right now is the interplay between cores, HyperTransport links, and SMP and NUMA scalability. To boost the scalability of machines further, AMD can double the cores and add enough HyperTransport links to take an Opteron server configuration up quite a bit, even without resorting to the Horus chipset. The current Opteron 8000 chipset can take four two-socket SMPs and lash them together with NUMA clustering over HyperTransport to make an eight-way server. For whatever reason, no Tier One server vendor has done this to date, but a number of smaller ones have. (The NUMA scalability is not great.) But now that AMD has dual-core Opterons, companies with four-socket servers can boost their performance without resorting to four-node NUMA clustering, and get more performance in the same thermal and physical profile. AMD could do something interesting for its far-future Opterons, like making multiple parallel HyperTransport links available between future Opteron chips, each with maybe four cores or eight cores with perhaps two or four links per core, to boost the aggregate bandwidth between the cores and create a switched mess network between all of the cores in the box. Whatever AMD might be planning, no one is saying.

And AMD is not all that worried about getting into the big SMP and NUMA server markets--at least not yet. "We need to make our mark on the two-way and four-way server space, and do it now," explains Patla. "This is the volume market, and we need to compete here. But that said, we do have our eye on the market for larger machines, and when we really expand into eight-way and maybe above, we will do it in a volume-market way."

And because it is IDF this week, and Intel is talking up its dual-core and multicore plans for the second time this year and getting a huge amount of airtime in the IT market, AMD has thrown down the gauntlet and launched an ad campaign that is challenging Intel to a shoot-out in the Dual Core Corral. AMD has taken out big ads in major newspapers saying it would meet Intel any time, any place to test its Pentium and Xeon chips against AMD's own Athlon and Opterons. Intel has thus far dodged the issue, and at IDF, Intel's new CEO said he thought that products are "best judged in the marketplace."

Then again, he would have to say that, being the CEO of a chip monopoly, because in many benchmark tests that might be run in a shoot-out, Intel's dual core chips are going to get whupped by their AMD equivalents. The market would, indeed, decide, all other things being equal. But, of course, all other things are not equal. While Hewlett-Packard offers Opteron across a pretty broad range of servers and in one workstation, IBM offers only tepid support (one blade server and one rack-server), Sun has only two servers and a few workstations (but is readying its "Galaxy" Opteron-based servers for a September 12 launch), and Dell, Fujitsu-Siemens, and NEC have not gone anywhere near Opteron for servers, and get uncomfortable talking about any possibilities.

Sponsored By
LINUX NETWORX

Clusterworx® Whitepaper

High performance Linux clusters can consist of hundreds or thousands of individual components. Knowing the status of each CPU, memory, disk, fan, and other components is critical to ensure the system is running safely and effectively.

Likewise, managing the software components of a cluster can be difficult and time consuming for even the most seasoned administrator. Making sure each host's software stack is up to date and operating efficiently can consume much of an administrator's time. Reducing this time frees up system administrators to perform other tasks.

Though Linux clusters are robust and designed to provide good uptime, occasionally conditions lead to critical, unplanned downtime. Unnecessary downtime of a production cluster can delay a product's time to market or hinder critical research.

    Since most organizations can't afford these delays, it's important that a Linux cluster comes with a robust cluster monitoring tool that:
  • Provides essential monitoring data to make sure the system is operational.
  • Eliminates repetitive installation and configuration tasks to reduce
          periods of downtime.
  • Provides powerful features, but doesn't compromise on usability.
  • Automates problem discovery and recovery on would-be critical events.

This paper discusses the features and functions of Clusterworx® 3.2. It details how Clusterworx® provides the necessary power and flexibility to monitor over 120 system components from a single point of control. The paper also discusses how Clusterworx® reduces the time and resources spent administering the system by improving software maintenance procedures and automating repetitive tasks.

High Performance Monitoring

Each cluster node has its own processor, memory, disk, and network that need to be independently monitored. This means individual cluster systems can consist of hundreds or thousands of different components. The ability to monitor the status and performance of all system components in real time is critical to understanding the health of a system and to ensure it's running as efficiently as possible.

Because so many system components need to be monitored, one of the challenges of cluster management is to efficiently collect data and display system health status in an understandable format. For example, let's say a cluster system has 100 nodes and is running at 97 percent usage. It's very important to know whether 100 nodes are running at 97 percent usage or whether 97 nodes are running at 100 percent usage while three nodes are down.

Clusterworx® provides real-time analysis of over 120 essential system metrics from each node. Data is displayed in easy-to-read graphs, thumbnails, and value tables. Clusterworx® collects data from groups of nodes to spot anomalies, then drills down to single node view to investigate problems. This allows users to determine exactly what the problem is before taking corrective action.

Clusterworx® also tracks the power and health state of each node and displays its status using visual markers in a node tree view throughout the user interface. Power status shows whether the node is on, off, provisioning, or in an unknown state. The health state tracks informational or warning messages and critical errors. Health state messages are displayed in a message queue on the interface.

Clusterworx®'s comprehensive monitoring and easy-to-read charts and graphs allow users to quickly asses the state of each node and the overall system at a glance - while providing the necessary information to make informed decisions about the cluster system.

To read the rest of this whitepaper, please visit www.linuxnetworx.com


Editor: Timothy Prickett Morgan
Contributing Editors: Dan Burger, Joe Hertvik, Kevin Vandever,
Shannon O'Donnell, Victor Rozek, Hesh Wiener, Alex Woodie
Publisher and Advertising Director: Jenny Thomas
Advertising Sales Representative: Kim Reed
Contact the Editors: To contact anyone on the IT Jungle Team
Go to our contacts page and send us a message.


THIS ISSUE
SPONSORED BY:

Egenera
Linux Networx
Roaring Penguin
California Digital
ShaoLin Microsystems


The Linux Beacon

BACK ISSUES

TABLE OF
CONTENTS
Unisys, IBM Further Prove Linux Performance on OLTP

AMD Nabs Chip Hotshot, Challenges Intel to Dual-Core Duel

The Source of All Good Bits

Mad Dog 21/21: The Grinchy Code

But Wait, There's More


The Four Hundred
IBM's Power6 Gets First Silicon as Power5+ Looms

The Many Pros and Few Cons of iSeries Logical Partitioning

ISVs Offer Six-Month Report Card on iSeries Innovation Program

Mad Dog 21/21: The Grinchy Code

The Windows Observer
Microsoft Hangs Tight to Visual Studio Ship Schedule

Intel Moves Paxville MP Chip Ahead into 2005, Adds DP Variant

VMware Goes for Per-Socket Pricing, But Can It Hold?

Sage Updates Software for SMBs and Governments

The Unix Guardian
HP's Sales Up 10 Percent as Repatriation Taxes Whack Profits

VMware, Sun Microsystems Partner on Server Partitioning

3PAR Delivers Server Provisioning on Solaris Boxes

Intel Moves Paxville MP Chip Ahead into 2005, Adds DP Variant


Copyright © 1996-2008 Guild Companies, Inc. All Rights Reserved.
Guild Companies, Inc. (formerly Midrange Server), 50 Park Terrace East, Suite 8F, New York, NY 10034
Privacy Statement