Any information available about the upcoming AMD Phenom 9950?

[DEredita]

Any information available about the upcoming AMD Phenom 9950? Read on Toms hardware that it is coming out in the next few weeks, along with a new AMD chipset and the R700 GPUs.

Linky: http://www.tomshardware.com/news/AMD-R700-GPU,5785.html

Any more info out there?

 

[SlowSpyder]

AMDZone says that the 9950 is for sale now. It's just a 2.6GHz 140watt version, compared to the 9850 which is 2.5GHz, 125watt. They also have links for the 65watt version of the Phenom.

 

[Phynaz]

Remember when they called Prescott a flame thrower?

 

[SlowSpyder]

Originally posted by: Phynaz
Remember when they called Prescott a flame thrower?

I'm not super familiar with the Intel processors, but from what I can tell their quad cores run from 95 - 136watt... not sure if their numbers are directly comprable to AMD's. Assuming they are measured in a similar method then the AMD stuff is right there with the Intel stuff in terms of power useage. Obviously the Intel processors are better, but it's not like the AMD stuff require twice the power for less performance.

 

[heyheybooboo]

I believe the Phenom 9950BE is the AMD quad (not-so-secretly advertised) to operate at 3GHz at stock AMD cpu 200MHz clock when combined the upcoming 790gx chipset and SB750.

If the rumors are true, the HT speed gets a big bump and the ram divisors work out close to 1066MHz you have a $300 cpu which performs in the Q9450-Q9550 range. Possible multithreading monster? I guess we'll find out.

 

[Accord99]

Originally posted by: SlowSpyder
Obviously the Intel processors are better, but it's not like the AMD stuff require twice the power for less performance.

That's roughly what it is:
http://www.behardware.com/arti...nom-b3s-in-detail.html

 

[Idontcare]

Originally posted by: SlowSpyder

Originally posted by: Phynaz
Remember when they called Prescott a flame thrower?

I'm not super familiar with the Intel processors, but from what I can tell their quad cores run from 95 - 136watt... not sure if their numbers are directly comprable to AMD's. Assuming they are measured in a similar method then the AMD stuff is right there with the Intel stuff in terms of power useage. Obviously the Intel processors are better, but it's not like the AMD stuff require twice the power for less performance.

As you mention the big differences comes down to what "TDP" means to the consumer for each manufacturer.

For the Intel consumer, having a quad with a TDP of 136W simply means under some of the absolute worst case powerbug routines the chip is going to consume (and require a HSF to dissipate) something close to 136W, but not more than that.

For the AMD consumer, having a quad with a TDP of 140W means under a broad range of typical desktop applications that a consumer would use a quad-core processor with, AMD determined the typical (not max) power consumption will likely be 140W for that processor.

They are so far apart in terms of what the numbers mean to system integrators as well as consumers that it's become pretty much a pointless metric of comparison between the two.

 

[Phynaz]

Originally posted by: heyheybooboo
I believe the Phenom 9950BE is the AMD quad (not-so-secretly advertised) to operate at 3GHz at stock AMD cpu 200MHz clock when combined the upcoming 790gx chipset and SB750.

You really believe AMD is coming out with a cpu that operates at different speeds depending upon what chipset it's paired with?

Nvidia and others would sue them into oblivion.
Not that that Hector isn't doing a fine job taking AMD there to begin with.

 

[Idontcare]

Originally posted by: Phynaz
You really believe AMD is coming out with a cpu that operates at different speeds depending upon what chipset it's paired with?

Nvidia and others would sue them into oblivion.
Not that that Hector isn't doing a fine job taking AMD there to begin with.

Hey now, Hector can't be blamed for his creator gifting him with that double-digit IQ.

The blame goes to he who put Hector in power (Sanders) and those who keep him there (shareholders and BOD).

It's not cool to give the neighborhood mentally challenged kid a handgun for the obvious reasons, why folks gifted Hector with managing AMD has always boggled my mind a bit.

It's not like he had no record or a stellar record prior to his AMD stint, he has performed exactly to the expectations set by his resume when he walked into the job.

 

[VirtualLarry]

Originally posted by: Idontcare

Originally posted by: SlowSpyder

Originally posted by: Phynaz
Remember when they called Prescott a flame thrower?

I'm not super familiar with the Intel processors, but from what I can tell their quad cores run from 95 - 136watt... not sure if their numbers are directly comprable to AMD's. Assuming they are measured in a similar method then the AMD stuff is right there with the Intel stuff in terms of power useage. Obviously the Intel processors are better, but it's not like the AMD stuff require twice the power for less performance.

As you mention the big differences comes down to what "TDP" means to the consumer for each manufacturer.

For the Intel consumer, having a quad with a TDP of 136W simply means under some of the absolute worst case powerbug routines the chip is going to consume (and require a HSF to dissipate) something close to 136W, but not more than that.

For the AMD consumer, having a quad with a TDP of 140W means under a broad range of typical desktop applications that a consumer would use a quad-core processor with, AMD determined the typical (not max) power consumption will likely be 140W for that processor.

They are so far apart in terms of what the numbers mean to system integrators as well as consumers that it's become pretty much a pointless metric of comparison between the two.

I thought that it was the other way around. AMD TDP is worst case max, and Intel is "typical". At least that's what I've read in the past.

 

[VirtualLarry]

Originally posted by: Phynaz

Originally posted by: heyheybooboo
I believe the Phenom 9950BE is the AMD quad (not-so-secretly advertised) to operate at 3GHz at stock AMD cpu 200MHz clock when combined the upcoming 790gx chipset and SB750.

You really believe AMD is coming out with a cpu that operates at different speeds depending upon what chipset it's paired with?

Nvidia and others would sue them into oblivion.
Not that that Hector isn't doing a fine job taking AMD there to begin with.

Why would Nvidia sue AMD into oblivion? NVidia already pulled the same stunt with their LinkBoost feature, overclocking the PCI-E bus from 100 to 125Mhz, when using an Nvidia graphics card with an Nvidia chipset.

AMD overclocking their CPUs, when run on their own AMD chipsets, would just be a smart marketing decision. Nvidia couldn't do a thing about it.

 

[harpoon84]

Originally posted by: VirtualLarry
I thought that it was the other way around. AMD TDP is worst case max, and Intel is "typical". At least that's what I've read in the past.

Times change. AMD used to be top dog too...

AMD's 'ACP' is probably close to Intel's P4 era TDP in that they both measure 'average' power consumption whatever that is...

Intel's current TDP rating is generally quite conservative, eg. the Wolfdale chips only consume around 30W under load but have a 65W TDP. Likewise, the Penryn quads have a 95W TDP but in reality use up less than 65W.

 

[heyheybooboo]

Originally posted by: Phynaz

Originally posted by: heyheybooboo
I believe the Phenom 9950BE is the AMD quad (not-so-secretly advertised) to operate at 3GHz at stock AMD cpu 200MHz clock when combined the upcoming 790gx chipset and SB750.

You really believe AMD is coming out with a cpu that operates at different speeds depending upon what chipset it's paired with?

Yes.

As a matter of fact the Phenom pinout seems to be a bit of a corporate secret.

More importantly (not that I expect you to care - much less understand), the issues regarding overclocking stability are directly related to the variable clock timings presented by the platform and the Phenom architecture.

And I find it rather appropriate that you would ask such a silly question simply because the individual cores of the Phenom microprocessor operate with variable timings and voltages across the HT and NB (also variable) in the AM2+arch.

 

[Idontcare]

Originally posted by: VirtualLarry

I thought that it was the other way around. AMD TDP is worst case max, and Intel is "typical". At least that's what I've read in the past.

That was the case for P4 vs. K8. It's reversed with the C2D vs. K10 situation.

Perhaps not too surprisingly, anything that marketing can fudge and spin will be fudged and spun by whichever side is playing the weak hand.

All that is important is that as a consumer (who cares about power consumption) one is aware of the marketing dept's influence on these metrics (as with all "performance ratings") and that nothing is noble or sacred when one company is down and out.

 

[heyheybooboo]

3.6GHz Phenom 9950BE ---> Here


On a MSI 790fx with sb600, 2x1Gb Kingston DDR2 800

Temps @ 3GHz-1.3v-stock cooler & cpu clock 200MHz/15x multi
idle: 36
load: 62

 

[Sylvanas]

Originally posted by: heyheybooboo
3.6GHz Phenom 9950BE ---> Here


On a MSI 790fx with sb600, 2x1Gb Kingston DDR2 800

Temps @ 3GHz-1.3v-stock cooler & cpu clock 200MHz/15x multi
idle: 36
load: 62

3.6ghz is pretty sweet, I have the same board on water I wonder what it would get to. I am tempted to pick one of these up but I think I might just hold out that little bit longer for Deneb.

 

[the kernel]

Originally posted by: SlowSpyder

Originally posted by: Phynaz
Remember when they called Prescott a flame thrower?

I'm not super familiar with the Intel processors, but from what I can tell their quad cores run from 95 - 136watt... not sure if their numbers are directly comprable to AMD's. Assuming they are measured in a similar method then the AMD stuff is right there with the Intel stuff in terms of power useage. Obviously the Intel processors are better, but it's not like the AMD stuff require twice the power for less performance.

You thought wrong.

ArsTechnica

Basically the 9950 consumes more than 100 watts more than the old Q6600 at load, which is comparable in performance and based on the old 65nm architecture. When you switch to the newer 45nm chips this gets even WORSE.

To top it off the Q6600 is cheaper and overclocks much better. This doesn't even take into account the stability and performance issues with Cool'n'Quiet that Anandtech uncovered. Is there really any argument that anyone can make about this chip being a decent value proposition?

 

[the kernel]

Originally posted by: heyheybooboo
Yes.

As a matter of fact the Phenom pinout seems to be a bit of a corporate secret.

Doesn't matter, this improved chipset = 200Mhz free is still just a rumor at BEST.

Originally posted by: heyheyboobooMore importantly (not that I expect you to care - much less understand), the issues regarding overclocking stability are directly related to the variable clock timings presented by the platform and the Phenom architecture.

And I find it rather appropriate that you would ask such a silly question simply because the individual cores of the Phenom microprocessor operate with variable timings and voltages across the HT and NB (also variable) in the AM2+arch.

The proof is still in the pudding as they say. Regardless of the Phenom's architectural differences, it still can't provide the performance when overclocked of an overclocked Intel core, even when compared to the year and a half old Q6600. When compared to the 45nm parts, the equation is even MORE in Intel's favor.

Totally unsubstantiated rumors are interesting, but no one should base buying decisions on them. Besides, there is a lot more evidence that Nehalem (and by evidence I mean actual benchmarks of actual hardware) will increase the performance gulf even MORE in Intel's favor.

 

[Idontcare]

Originally posted by: the kernel
Totally unsubstantiated rumors are interesting, but no one should base buying decisions on them. Besides, there is a lot more evidence that Nehalem (and by evidence I mean actual benchmarks of actual hardware) will increase the performance gulf even MORE in Intel's favor.

I remember when AMD got to that certain inflection point on the rise in their dominance cycle (before cresting) where they just sorta stopped pushing the leading edge of the performance envelope and all the clock increases just about flatlined at 2.2GHz or thereabouts for quite some time.

Looks like Intel is going to get there chance here. Do they start coasting like AMD did, or do they double-down to really drive home the point and push onwards to 90% market dominance?

 

[bradley]

Heyheybooboo's theory is likely the correct one, and goes beyond mere motherboard design. Just about any electronic design needs pure signals and proper timing for stability. The less synchronization and phase of the signal, the more jitter, ringing, noise etc. is introduced to create decreased stability and efficiency, among a host of other things. Lately Heyheybooboo is like a lone voice in the wilderness, as he seems very well-versed in AMD -- much needed around here.

Right now AMD is using a kludge analog PLL. They also have completely bypassed the SB710's release and are shooting toward the SB750, with its integrated PLL. The PLL is what is responsible for predicting accurate phase. So.. in theory the SB750 should allow not only the Phenom, but also the X2, to clock much higher. Otherwise, test even those 5-phased 780G/SB700 motherboards (with their cheaper parts) on an oscilloscope, I bet it all starts falling apart before ever reaching the 125W threshold. The voltage spikes alone are probably what's killing the mosfets on lower-end 780G motherboards.

Perhaps a high quality overcurrent protect would be needed as a stopgap. But again, these things cost money. I slightly blame motherboard manufacturers for throwing more phases at these designs, instead of simply using higher quality parts: mosfets, pwm etc. But these marketing departments are just as good at selling unnecessary multi-phases designs as they are selling monsterous 1000W powersupplies.

Migrating to a digital PWMs also might solve many of the problems with transient response we currently face. They would be a huge long-term money savers, while also increase efficiency and reliability. Except again, the start-up costs are initially high. Meanwhile just imagine all those hulking capacitors replaced by tiny, relatively inexpensive, SMT resistors.

We must also consider that low-voltage multi-cores running on teeny nm processes are some of the worlds most super-complex designs, with far slimmer margins of error. A lot of people here seem to give this very little consideration. Running below 2 volts has already created a whole new set of rules needing to be resolved within the motherboard itself, mostly through inefficient analog stages. Add 64-bit registers, sloppy software developers, and a lackluster performer in Vista, and there are even more setbacks with which to contend. That's why coordinating a good roadmap, along with a great partnership between AMD/Intel/Via and mobo manufacturers/software designers is more imperative than ever.

And I'm sorry, there's also a huge lack of general understanding about the major differences in design between the K8 and K10. I'm actually disgusted pretty much by the whole enthusiast movement. There seems a willingness to quote just about any source, and a lack of overall intellectual curiosity. There are also a dearth of disingenuous sites that just repeat marketing with very little verification. The worst seem to be those extolling the the virtues of a certain design, while accentuating the negatives of the competitor, classic marketing 101. And unfortunately very few, if any, of these review sites have an understanding beyond simple electronic theory to break the vicious cycle.

We (as consumers) are being conditioned to not demand the correct things: digital pwms, solid caps, more efficient mosfets, a move away from the ancient FSB and sequential computing, better fusion between GPU and CPU under superior platforms, greater power efficiency, better 64-bit compatibilty etc. Instead we're stuck on the same broken rhetoric and power struggle that pits AMD vs. Intel. Meanwhile, missteps withstanding, what's killing AMD is not so much their parts, but this almost persistent viral marketing campaign, sad to say. Of course, AMD did an extremely poor job containing the damage.

What is needed is a reviewer with an advanced electronic engineering degree who is also knowledgeable about AMD parts, otherwise we just have more of the same hoodwinking. At least we're now starting to see more talk about the importance of HT and an IMC with Nehalem fast approaching. For me, it's better late than never.

 

[the kernel]

Originally posted by: bradley
We must also consider that low-voltage multi-cores running on teeny nm processes are some of the worlds most super-complex designs, with far slimmer margins of error. A lot of people here seem to give this very little consideration. Running below 2 volts has already created a whole new set of rules needing to be resolved within the motherboard itself, mostly through inefficient analog stages. Add 64-bit registers, sloppy software developers, and a lackluster performer in Vista, and there are even more setbacks with which to contend. That's why coordinating a good roadmap, along with a great partnership between AMD/Intel/Via and mobo manufacturers/software designers is more imperative than ever.

What point exactly are you trying to make here? That making high performance CPUs is hard? I think everyone is well aware of that.

Originally posted by: bradley
And I'm sorry, there's also a huge lack of general understanding about the major differences in design between the K8 and K10. I'm actually disgusted pretty much by the whole enthusiast movement. There seems a willingness to quote just about any source, and a lack of overall intellectual curiosity. There are also a dearth of disingenuous sites that just repeat marketing with very little verification. The worst seem to be those extolling the the virtues of a certain design, while accentuating the negatives of the competitor, classic marketing 101. And unfortunately very few, if any, of these review sites have an understanding beyond simple electronic theory to break the vicious cycle.

Those who are praising the virtues of Core 2 over K10 are doing so based on certain metrics including application performance in a wide variety of benchmarks, performance-per-watt, overclockability, scalability at increased clock speeds, cost, and more. If you have a different methodology then feel free to present it, but you are going to have to work hard to try and justify to the enthusiast community that the factors I have outlined above should be ignored.

Originally posted by: bradley
We (as consumers) are being conditioned to not demand the correct things: digital pwms, solid caps, more efficient mosfets, a move away from the ancient FSB and sequential computing, better fusion between GPU and CPU under superior platforms, greater power efficiency, better 64-bit compatibilty etc. Instead we're stuck on the same broken rhetoric and power struggle that pits AMD vs. Intel. Meanwhile, missteps withstanding, what's killing AMD is not so much their parts, but this almost persistent viral marketing campaign, sad to say. Of course, AMD did an extremely poor job containing the damage.

What is killing AMD is that they do not have competitive parts in the areas I have outlined above except in perhaps some extremely narrow price/performance segments. The fact that their roadmap is inferior to Intel's doesn't help either.

You seem to be suggesting that technology for the sake of technology is better, which is exactly the wrong way to look at it (and it's not the way a proper engineer nor product manager would look at it either). Products are designed to meet certain goals, and technology is only used if it is necessary to reach those goals. As any engineer will tell you, the best designs are the ones that use the LEAST amount of cutting edge technology.

If I may quote an acquaintance of mine who is a mechanical engineer:

Michael Wong wrote......In fact, the most elegant engineering solutions are those that require the least technology, not the most. A good example is a machine gun; it uses a simple, elegant and robust mechanical system to eject each cartridge and load the next, based on gas pressure, springs, rods, and other low-tech principles. The simpler, the better. With modern technology, we could design a machine gun that uses miniaturized robotics instead, but why? The resulting weapon would be far more expensive, and far less reliable. It would require a power source, and software. It would be far more difficult to maintain...

I think the point of the above speaks for itself. You can go on and on about checklist features, but it's the PERFORMANCE of the part that matters.

Originally posted by: bradley
What is needed is a reviewer with an advanced electronic engineering degree who is also knowledgeable about AMD parts, otherwise we just have more of the same hoodwinking. At least we're now starting to see more talk about the importance of HT and an IMC with Nehalem fast approaching. For me, it's better late than never.

Now you're being just completely intellectually dishonest. Everyone in the enthusiast community is well aware of the advantages of HT and the IMC, the reason they aren't talked about much anymore is because they are five years old in AMD designs and already well understood. In addition, HT doesn't mean much for systems that aren't 4p or greater, so most enthusiasts don't think about it much.

IMC is important to the performance of K8/K10, but it's already factored into the overall performance of the processor. Nothing can be said about the IMC that isn't reflected by the right selection of benchmarks.

 

[Idontcare]

Originally posted by: bradley
What is needed is a reviewer with an advanced electronic engineering degree who is also knowledgeable about AMD parts, otherwise we just have more of the same hoodwinking. At least we're now starting to see more talk about the importance of HT and an IMC with Nehalem fast approaching. For me, it's better late than never.

I liked what "the kernel" had to say but I'd also add that you seem to be intent on blaming the reviewers for what in all fairness is an absolute failure on AMD's behalf. Be it their own lackluster engineering of robust eco-system of integrators (mobo's, chipsets, CPU's) or lacking marketing/sales dept which ought to be doing something to effect dissemination of the very technical information you are seeking.

It is not the reviewer's job or objective to debug and solve AMD's platform level issues for AMD or their partners (mobo makers). Presumably AMD ought to have an active fielded team of talented engineers who are waaaay out ahead of the mainstream journalist in identifiying, isolating, and resolving the very issues that crop up in the reviewer's articles.

That is continues to be the reviewers who are finding these seemingly obvious weaknesses in the platforms is a harbinger of what lies below the surface. How many snafus must one witness before one realizes there is something rotten in denmark?

 

[Cookie Monster]

Also adding with what IDC said, CPUs and also GPUs nowadays are very much too complex for a single person or even a group with an EE degree to keep a track of and actually know what the chip is doing at a circuit level. Sure they can assume and predict with more accuracy and credible ideas than most average people, but thats about it.

 

[Phynaz]

Sounds like the fanboys are upset.

 

[RallyMaster]

Originally posted by: Accord99

Originally posted by: SlowSpyder
Obviously the Intel processors are better, but it's not like the AMD stuff require twice the power for less performance.

That's roughly what it is:
http://www.behardware.com/arti...nom-b3s-in-detail.html

Why does that say my E7200 only consumes 20.8 watts??