The Triumphant return of Power now = Unlocked CPUs 4 all!

[Sentential]

AMD To Provide On-Demand Power Management For The Rack-Dense Enterprise Market

Addition of AMD PowerNow!™ technology with Optimized Power Management will further enhance the AMD Opteron™ processor for datacenters

SUNNYVALE, CALIF -- December 6, 2004 --AMD (NYSE: AMD) today announced the AMD Opteron™ processor family will include AMD PowerNow!™ technology with Optimized Power Management (OPM) in the first half of 2005. AMD PowerNow! technology for the AMD Opteron processor is designed to allow enterprise IT and workstation customers to decrease overall power consumption by enabling them to optimize performance-on-demand.

The addition of the innovative AMD PowerNow! technology with OPM is expected to strengthen the industry-leading performance-per-watt capabilities of the AMD Opteron processor. Not only will AMD PowerNow! technology with OPM provide enhanced power management capabilities, it can also aid in platform investment protection for today’s demanding enterprise server environments by providing less strain on datacenter cooling and ventilation systems.

The integration of AMD PowerNow! technology with OPM can reduce power consumption in the datacenter by dynamically changing power states based on workload utilization. This flexibility will not only enable lower total datacenter operational costs, but will also provide additional work environment benefits:
Lower Cost of Ownership: Power-on-demand allows systems to run at optimum performance and power levels, reducing electricity costs.
Investment Protection: AMD PowerNow! technology can reduce heat, decreasing strain on facility cooling systems.
Lower Acoustics: AMD PowerNow! technology can reduce heat produced by a system’s processors, allowing cooling fans to run at reduced speeds and keep noise levels down.
“The addition of AMD PowerNow! technology to the AMD Opteron processor will prove to be another proof point in AMD’s philosophy of customer-centric innovation and listening to the needs of enterprise IT managers,” said Marty Seyer, corporate vice president and general manager of the AMD Microprocessor Business Unit, Computation Products Group. “With innovative technologies like AMD PowerNow! technology with Optimized Power Management and our upcoming AMD64 multi-core technology, AMD will continue to lead the industry in performance-per-watt efficiencies.”

Industry Support
Since its introduction nearly four years ago, AMD PowerNow! technology has been widely embraced by AMD’s notebook customers. In the first half of 2005, leading server and workstation OEMs currently offering AMD Opteron processor-based platforms will support this latest enhancement to the industry’s first 32- and 64-bit x86 processor.

“Power efficiency is one of a set of emerging components that is redefining the total cost of ownership for datacenter solutions,” said Paul Miller, vice president of marketing, Industry Standard Servers, HP. “AMD Opteron processor-based HP ProLiant servers coupled with the power management advantages of AMD PowerNow! technology will enable our customers to manage their datacenters, offering the greatest rack density, while at the same time reducing IT operating costs. That combination is tough to beat.”

“IBM is committed to optimizing the datacenter environment and IBM PowerExecutive and Calibrated Vectored Cooling technologies already bring distinct advantages to our customers,” said Alex Yost, director, IBM eServer Product Management. “AMD’s PowerNow! technology adds to these solutions to ensure the performance-per-watt efficiencies, and expands our choice of the best data center solutions for our customers.”

“Our joint customers expect us to deliver on the promise of innovation throughout the datacenter—from OS-level innovation in Solaris 10 to processor-level innovation like AMD PowerNow! technology,” said John Fowler, executive vice president at Sun Microsystems. “This additional technology from AMD will give our Sun Fire servers and Sun Java Workstations based on the AMD Opteron processor even better performance-per-watt capabilities, which is directly in line with Sun's goal of simplifying the datacenter.”

About the AMD Opteron™ Processor
The world’s first 32-bit and 64-bit processor compatible with the x86 architecture, the AMD Opteron processor is based on AMD64 technology with Direct Connect Architecture. Direct Connect Architecture helps eliminate the bottlenecks inherent in a front-side bus by directly connecting the processors, the memory controller and the I/O to the central processor unit to enable improved overall system performance and efficiency. AMD was the first to announce the completion of an x86-based dual-core processor design and the first to demonstrate an x86-based dual-core processor for 32- and 64-bit computing. As more solution providers join the AMD64 ecosystem, the industry is approaching the day when 32-bit-only systems will become obsolete.

About AMD
AMD (NYSE:AMD) designs and produces innovative microprocessors, Flash memory devices and low-power processor solutions for the computer, communications and consumer electronics industries. AMD is dedicated to helping its customers deliver standards-based, customer-focused solutions for technology users, ranging from enterprises and governments to individual consumers. For more information, visit www.amd.com.

Cautionary Statement
This release contains forward-looking statements, which are made pursuant to the safe harbor provisions of the Private Securities Litigation Reform Act of 1995. Investors are cautioned that forward-looking statements in this release involve risks and uncertainties that could cause actual results to differ materially from current expectations, including the possibility that the company may not achieve its current product and technology introduction schedules. We urge investors to review in detail the risks and uncertainties in the company’s Securities and Exchange Commission filings, including but not limited to the Annual Report on Form 10-K for the year ended December 28, 2003, and the Quarterly Report on Form 10-Q for the quarter ended September 26, 2004.

Those of you who are familiar with Powernow! know that it is the sole reason why Mobile Bartons remained unlocked to the bitter end. With its return we may see fully unlocked Opterons and possibly unlocked Mobile S754/S939 CPUs in the near future. w00t!

 

[ps2cho]

=) very interesting.
have i mentioned i love u ?

haha

 

[deception``]

I suspect that we'll see this only on the server side of things, in an effort to manage heat issues. However, such a phenomena could trickle down to us desktop users. Unfrotunately, realization of this will take quite some time. Most likely, we won't even be able to find such chips until after AMD has released it's final s939 revision.

deception``

 

[Sentential]

I suspect that we'll see this only on the server side of things, in an effort to manage heat issues. However, such a phenomena could trickle down to us desktop users. Unfrotunately, realization of this will take quite some time. Most likely, we won't even be able to find such chips until after AMD has released it's final s939 revision.

deception``

But by then we will see S939 in a mobile varient.....which *will* support Powernow!. Hopefully we will see some S754s with this feature before the socket completely dies off

 

[deception``]

But by then we will see S939 in a mobile varient.....which *will* support Powernow!. Hopefully we will see some S754s with this feature before the socket completely dies off

If you recall, mobile AXP chips didn't make it in our hands until the Barton core was released. So I think that AMD will go ahead with their Revision E chips before they begin to leak mobile s939 chips, which may or may not use the PowerNow! feature. I do admit, however, that I am very enthusiastic about K8 cores with even lower power ratings. Nonetheless, I doubt that AMD will go to great lengths to accomodate their once-mighty single channel socket, as s939 will be the mainstream once this technology debuts again.

deception``

 

[Sentential]

Nonetheless, I doubt that AMD will go to great lengths to accomodate their once-mighty single channel socket, as s939 will be the mainstream once this technology debuts again.

deception``

Very true. Perhaps this news is to imply their new CPUs. I forget what its called but it is supposed to be the direct competitor to the Centrino. Some ultra-low power CPU that we have yet to see.

 

[deception``]

Very true. Perhaps this news is to imply their new CPUs. I forget what its called but it is supposed to be the direct competitor to the Centrino. Some ultra-low power CPU that we have yet to see.

Well, the lower the power, the greater the likelihood that chip has of making it into my next DFI.

deception``

 

[cujo]

a64s are unlocked downwards from stock. correct me if i'm wrong but power now would only need to downclock from stock. so why would power now allow upward unlocked multis? am i missing something?

 

[Petr]

All K8 CPUs including Opterons support PowerNow! v1.4 from day #1. What do you think is Cool'n'Quiet? It is PowerNow! plus thermally controlled fans.

Since K8 chips do not have regular hardware multipliers, the only way to set multiplier is through PowerNow!. And only FX series have max. PowerNow! multiplier set to 25x (platform limit).

 

[Gautam]

a64s are unlocked downwards from stock. correct me if i'm wrong but power now would only need to downclock from stock. so why would power now allow upward unlocked multis? am i missing something?

You're dead on.

If everyone just realizes that not having high multipliers doesn't actually get in the way of ocing, we can all stop whining about it.

 

[SomethingClever]

You're dead on.

If everyone just realizes that not having high multipliers doesn't actually get in the way of ocing, we can all stop whining about it.

It's true they aren't strictly necessary; however, I was under the impression that higher multi's yield better performance in distributed computing projects (not to mention overall system speed--more so than FSB or HTT boosts). Given this, I would think more people would be interested in attaining the highest multiplier possible.

PS: Sorry for the thread hijack!

 

[Kuroimaho]

I think you mean highest speed not highest multi. I doubt ppl want 20* multi.

 

[SomethingClever]

They don't!?

What I actually meant was the highest *realistic* multiplier. Just because 25x is the ceiling doesn't mean it's realistic to expect to hit it, even using phase change. I'm not even sure anything above 14 is terribly useful.

 

[Clean_Baldy]

Well, the lower the power, the greater the likelihood that chip has of making it into my next DFI.

deception``

COMPLETELY off topic...

Why do you keep writing your name after you type something?

 

[deception``]

COMPLETELY off topic...

Before you flame me for making one off-topic sentence, perhaps you'd care to refer to the other relevant posts that I have made in this thread?

Why do you keep writing your name after you type something?

How about this: Because I want to? If you have a problem with me, drop me a pm.

deception``

 

[Seabee]

It's true they aren't strictly necessary; however, I was under the impression that higher multi's yield better performance in distributed computing projects (not to mention overall system speed--more so than FSB or HTT boosts). Given this, I would think more people would be interested in attaining the highest multiplier possible.

PS: Sorry for the thread hijack!

I believe you have a few things mixed up here. A high FSB does very little for most DC projects. That doesnt mean a chip at 12.5*200 is going to do better than one at 10*250. Now, that DC project would run better the higher you get the overall clockspeed but it doesnt matter HOW you get that clockspeed. Understand? ex: A 2.2ghz CPU will do better than a 2.0 CPU. A 2.2 CPU at 12.5*200 will do the same as a 2.2 CPU at 10*250. This of course is under the assuption both CPUs are completely identical(ie 2600+/2600+ or P4/P4).

As far as your system goes, the higher FSB is pretty much better than having a high multi. High multi is just that, a high multi and high clockspeed. A high FSB is high clockspeed and more memory bandwidth, therefor increasing your memory performance. So, if you could only get your CPU to run at 2.5ghz, the 10 multi * the 250 FSB is the 'prefered' way to go(assuming you can reach 250 FSB) rather than a multi of 12.5 and FSB of 200.

Consensus is get your FSB as high as it can go, then push your multi. Rather than the other way around.

 

[Tebore]

I doubt we'll see unlocked chips. AMD's PowerNow is the same as CnQ. Which was on all Desktop A64s but not on Opterons. If we didn't have unlocked chips then I doubt we'll have them now.

When I say unlocked I mean adjustments over the default not the half assed unlock like it is now where you still have the lower mults.

 

[SomethingClever]

I believe you have a few things mixed up here. A high FSB does very little for most DC projects.

Sorry if I didn't phrase that correctly. I'm well aware that FSB alone will do nothing for DC. What I meant to say was that given the choice, I would choose a higher mult over a higher FSB. Based on what I've read, a high mult will improve the speed at which every program runs, unlike FSB, which only helps out bandwidth-dependant programs. That's just my take on it, and I could be wrong.

As far as your system goes, the higher FSB is pretty much better than having a high multi... Consensus is get your FSB as high as it can go, then push your multi. Rather than the other way around.

Unfortunatly, I can't force my FSB past 192. I'm not sure, but I suspect my cheap RAM is to blame (shame on me for not doing my research). In this case, the only way I'm going to get a good OC is to pump the mult for all it's worth. It's not even worth messing with at this point, as I'll be upgrading soon anyway.

 

[Seabee]

Ok, so you push your RAM to 192, then push the multi up. That is fine, I didnt mean you HAD to hit 200 or higher. The multi doesn do 'work' so to speak. The multi doesnt do anything for programs, just the CPU.

A DC will run virtually the same(if there is a difference, you probably wouldnt even notice it) on your CPU either at 13*192=2496 or 14*178=2492. But, most other programs will run slightly better with the 13*192 cuz of the bandwidth. Possibily noticable, possibily not.

But, I do believe if you ask around, 90% or more would choose a higher FSB. To me, 90% cant be wrong.

 

[Gautam]

What I meant to say was that given the choice, I would choose a higher mult over a higher FSB. Based on what I've read, a high mult will improve the speed at which every program runs, unlike FSB, which only helps out bandwidth-dependant programs. That's just my take on it, and I could be wrong.

There's no way to put it lightly. I'm afraid you are wrong. The front side bus is pretty much the most important data path in a computer. A high multi will do nothing but reduce the front side bus. Multipliers by themselves do absolutely nothing. On the other hand, the memory flows through the front side bus making it absolutely crucial. Without going into a debate over latencies, its safe to say that the more memory bandwidth, the better. A high multiplier is useless, I repeat.