Rumor: AMD "Piledriver" FX CPU production to begin Q3 2012

[blastingcap]

So what did Rory learn about, or come to realize, in the past 6 months that he is now singing a different toon?

Toon? Tunes? Both?

 

[Arzachel]

That's a good point. Figure the design team for this chip was given project directives circa 2008, it is very likely to be what you are saying. Project was too far along to justify aborting when Rory came onboard.

Still though, the Rory of today hardly reflects the "we must be the predator" Rory that wanted the world to think of AMD back in Sept.



So what did Rory learn about, or come to realize, in the past 6 months that he is now singing a different toon? Is it Piledriver not doing what marketing had him thinking it was going to do?

Something changed, the pipeline didn't, but the CEO's rhetoric sure did. What caused the change is very much of interest to me.

I don't think anything has changed. If anything, that statement means that AMD has finally understood the futility of chasing Intel. They simply can't beat Intel at single thread performance, they're short in-house fabs and a hundred billion or so. But there are a lot of things they can do that Intel can't, which is why the whole HSA deal is so fascinating.

 

[Idontcare]

Toon? Tunes? Both?

Oh you, you're such a character

I don't think anything has changed. If anything, that statement means that AMD has finally understood the futility of chasing Intel. They simply can't beat Intel at single thread performance, they're short in-house fabs and a hundred billion or so. But there are a lot of things they can do that Intel can't, which is why the whole HSA deal is so fascinating.

Well that's depressing

I thought Rory was jazzing up the employees to get ready to go kick butt and take some names with his whole "we be predators" speech.

To give up just a mere 6 months later, when all that has changed publicly is IB's release, is rather unsettling.

But blastingcap brought up a crucial point above when he mentioned the pipeline. Nothing coming out of AMD now or in the near future has anything to do with Rory, its all due to Dirk.

We won't know what Rory is up to inside AMD today for about 3 more years yet.

So in that interim period it is, of course, going to look like the CEO is saying one thing while the company appears to be doing another thing. The pipeline makes that an unavoidable reality.

 

[SickBeast]

So what did Rory learn about, or come to realize, in the past 6 months that he is now singing a different toon? Is it Piledriver not doing what marketing had him thinking it was going to do?

Something changed, the pipeline didn't, but the CEO's rhetoric sure did. What caused the change is very much of interest to me.

A 26% improvement is not enough for AMD. Intel will still have the performance crown.

They're not even talking about power consumption so I assume it's still terrible.

 

[Arzachel]

Well that's depressing

I thought Rory was jazzing up the employees to get ready to go kick butt and take some names with his whole "we be predators" speech.

To give up just a mere 6 months later, when all that has changed publicly is IB's release, is rather unsettling.

But blastingcap brought up a crucial point above when he mentioned the pipeline. Nothing coming out of AMD now or in the near future has anything to do with Rory, its all due to Dirk.

We won't know what Rory is up to inside AMD today for about 3 more years yet.

So in that interim period it is, of course, going to look like the CEO is saying one thing while the company appears to be doing another thing. The pipeline makes that an unavoidable reality.

I think the "predator" speech was more about improved execution and playing to their strengths, which is kinda happening on the APU side. Predators know how to pick their fights I guess

A 26% improvement is not enough for AMD. Intel will still have the performance crown.

They're not even talking about power consumption so I assume it's still terrible.

Trinity is already out and uses Piledriver cores, you know?

[Spoilers]One of your statements is true, one is false and one can't be answered [/Spoilers]

 

[ShintaiDK]

A 26% improvement is not enough for AMD. Intel will still have the performance crown.

They're not even talking about power consumption so I assume it's still terrible.

I would note that the claimed improvement is over Husky. aka Llano.

 

[mikeymikec]

A 26% improvement is not enough for AMD. Intel will still have the performance crown.

Also consider that this was written by their marketing department, so "26% at best" is probably the way to read it.

 

[Vesku]

Even if what you have today is "good enough" you can't stand still. Ideally you would pair Intel CPU with a discrete GPU but economic reality means that this isn't 100% of all purchases.

When you peel back the corporate speak what AMD is saying is: don't dismiss our products simply because one portion isn't as good as Intel's. Look at the total package.

Regarding Piledriver specifically, after Bulldozer we can't accept even toned down assessments of AMD info. Confirmed retail silicon or bust.

 

[ShintaiDK]

Regarding Piledriver specifically, after Bulldozer we can't accept even toned down assessments of AMD info. Confirmed retail silicon or bust.

Im not sure of the "Phenom is 40-50% faster than Core 2" aint worse actually.

 

[pelov]

So in that interim period it is, of course, going to look like the CEO is saying one thing while the company appears to be doing another thing. The pipeline makes that an unavoidable reality.

That's true but he can also have a direct impact as well. For the first time in... well, first time ever really, AMD has launched a product on time and it looks like Vishera will be around the same 1 year mark as well assuming an early Q4 release (and it probably will. They don't have any OEM supply issues as OEMs generally jump on their APUs and not their desktop chips). From a company that delayed Bulldozer for 2+ years and is known for delaying everything that's a pretty solid improvement.

I also think the direction was already away from the race with Intel on the desktop front. AMD knew they'd never catch Intel with respect to IPC and efficiency due to the fab advantage back with Meyer at the helm and that's how Bulldozer was born. If we can't match them with either of those then why don't we go with clock speeds and moar coars? Regardless of the architecture being crap it at least showed that they accepted reality. They would never catch Intel in IPC. Now that the fab advantage has only widened they've got no choice but to look for other markets. Intel makes money hand over fist with their small and efficient chips at higher prices while AMD is forced to sell bigger hungrier chips at lower cost. Unlike Meyer, Read is no idiot when it comes to business. He looks at the bottom line and if that line is red and shows no sign of improving he won't hesitate to cut their losses and try something else.

The current pipeline will be released in a timely manner and they'll still try to make whatever $$ they can, whether from server or desktop but I highly doubt that whatever they have currently brewing in their labs has anything to do with the desktop or even a discrete GPU as neither of those have made them any money the past few years.

 

[guskline]

I sure hope the marketing dept at AMD doesn't pull one of those "Guinness Book of Records" stunt like they did with the Bulldozer and OC on only 2 cores in some secret meeting with nitogen floating around. That flamed the false hope of an Uber OC chip to a fever pitch only to be dashed when Bulldozer was released.

I hope the AMD engineers throw the marketing gang out of the room and seriously improves the Piledriver.

As I said in another post, a lot of us bought the 990 FX chipset MBs in hopes of a killer CPU only to be let down.

For myself, I kept the 990FX mb (asus Sabertooth) and was lucky enough to snag a 1100 Thuban before they disappeared. I have to see a reason to go to Piledriver - and that will come from independent reviews, NOT AMD's marketing slides.

 

[frozentundra123456]

Also consider that this was written by their marketing department, so "26% at best" is probably the way to read it.

Plus if you look closely, both performance improvement statements have a superscript by them apparently directing the reader to some qualifying statement, which has been conveniently deleted from the slide.

So I am sure that it is the familiar AMD "up to" improvement metric that is qualified somehow or related to some cherry picked scenario.

I really could cut AMD a lot more slack with their CPU problems if they didnt make so many marketing statements that they cant back up or at least that are true only in a very limited scenario.

 

[Vic Vega]

...

 

[wlee15]

I don't understand why AMD is doing this.

I thought their current CPU's were already "good enough" for the masses, that is why no one needs Intel's even faster chips that are available today, right?

Aren't all the laptops out there now already over-powered? Why is AMD improving on something they already have stated does no one any good if it becomes faster?



It's almost like they say one thing, but don't really mean it or believe it themselves given that they are spending oodles of money directed towards the exact opposite message :hmm:

In fairness Rory Read was talking about laptops while FX processors is desktops. After all when Intel believes that Medfield offers excellent performance I would assume that they don't mean for every computing activity out there. Obviously Read would choose to emphasize graphics and battery life.

 

[Magic Carpet]

If that's 10-20% faster per clock & 20-30% more power efficient, unlocked multiplier. At the right price, I am in

AM3+ owners, get ready

 

[podspi]

Honestly, I don't think anything has changed at AMD. Essentially what Read is saying is "Unlike Intel, we are no longer going to focus on performance at any cost".

Ignoring the fact that Intel hasn't done that since the P4 (that part of his statement is all marketing), all he's saying is that AMD's internal strategy is now focused on perf/watt (instead of just performance), and that they aren't going to pursue the performance crown (which is unnecessary).

Given the time it takes for products to move through the pipeline, it seems to me that performance/watt has been AMD's focus for quite a while. It is also pretty clear to me that something went very, very wrong with the FX-series. Ignoring absolute performance levels, perf/watt for AMD products has increased dramatically since the release of Bobcat and Llano, and perf/watt is what most people are buying these days.


That being said, I really hope AMD doesn't leave us enthusiasts behind, even though it is beginning to look like that is what is going to happen. How come there aren't any threads bashing Texas Instruments for not having an i7 buster? I'll tell you why, because you don't have to produce the fastest chip in the world to be a successful semi-company.

Oh and, does anybody have any idea how much ST performance might be helped by the inclusion of L3 cache? It'll be interesting to see how much faster per clock Vishera is than Trinity... (assuming there are no other changes...)

 

[Ajay]

If that's 10-20% faster per clock & 20-30% more power efficient, unlocked multiplier. At the right price, I am in

AM3+ owners, get ready

TDP will be the same, IPC is supposed to be up a solid 15% - don't know what that will translate into for real world apps. Since the GloFo process is more mature ultimate overclocks should be higher. Those are the things I've read that I believe. This chip was originally slated to be out 3Q12, so something has slipped again somehow.

If it slipped because AMD put more time into optimizations and we see even larger performance increases - that would be cool. If AMD pulled a rabbit out of it's hat and it's 25% faster and overclocks easily to 5GHz on high end air - that would catch my attention. I'd really like to seem them starting to compete again (despite what Reed has said) - all though I'm not expecting to see really big changes till 22nm (when Intel will be on 14nm).

 

[Xpage]

actually I think GloFlo is skipping 22 and going to 20nm from what I read last

 

[Atreidin]

I'm excited to see how the Bulldozer architecture ideas pan out. Sure it is lackluster in pretty much all but specially selected workloads, but it will be fun to see how things improve. As processor design is an interest and hobby of mine I think it is interesting to analyze different design approaches.

It sucks that bad management can screw up technology development but it is what it is. History is littered with examples of companies that screwed up horribly with a design that sounded good on paper but sucked in production (Intel included), because they didn't account for one or more vital aspects. Not every company has the resources that Intel has to stay in the game after bad mismanagement and terrible design choices. I'm glad AMD is still around to employ engineers that can try out different ideas to advance the field and keep Intel on its toes.

Who knows, maybe there are some adjustments that can turn the BD architecture into something a lot better. They are stuck with it for now, so I think it will be interesting on an intellectual level to see where it goes.

If certain people would stop acting like AMD killed their dog and focused on the technology and engineering aspects of things they might get more enjoyment out of all of this. It is a lot more fun than taking personal offense to every little thing.

 

[sequoia464]

If certain people would stop acting like AMD killed their dog

Not going to happen.

 

[exar333]

Honestly, I don't think anything has changed at AMD. Essentially what Read is saying is "Unlike Intel, we are no longer going to focus on performance at any cost".

Ignoring the fact that Intel hasn't done that since the P4 (that part of his statement is all marketing), all he's saying is that AMD's internal strategy is now focused on perf/watt (instead of just performance), and that they aren't going to pursue the performance crown (which is unnecessary).

Given the time it takes for products to move through the pipeline, it seems to me that performance/watt has been AMD's focus for quite a while. It is also pretty clear to me that something went very, very wrong with the FX-series. Ignoring absolute performance levels, perf/watt for AMD products has increased dramatically since the release of Bobcat and Llano, and perf/watt is what most people are buying these days.


That being said, I really hope AMD doesn't leave us enthusiasts behind, even though it is beginning to look like that is what is going to happen. How come there aren't any threads bashing Texas Instruments for not having an i7 buster? I'll tell you why, because you don't have to produce the fastest chip in the world to be a successful semi-company.

Oh and, does anybody have any idea how much ST performance might be helped by the inclusion of L3 cache? It'll be interesting to see how much faster per clock Vishera is than Trinity... (assuming there are no other changes...)

Great post.

Intel's strategy since Nehalem has paid-off very well for them regarding increasing efficiency. The bonus to having a very efficient high-end CPU is that you can usually scale that down to many different slower CPUs that are still fast, but equally efficient. We saw that with recent Pentiums and Celerons that really sipped power in 2C setups.

Hopefully AMD can right the ship and get something similar out the door. Even if they cannot scale them up to the fastest, they can still be competitive and have a architecture to scale-down with.

 

[guskline]

For me it's simple. Will Piledriver be a noticeable improvement over my 1100 Tuban.

 

[Arzachel]

For me it's simple. Will Piledriver be a noticeable improvement over my 1100 Tuban.

Depends on what you call "noticeable".

 

[ShintaiDK]

TDP will be the same, IPC is supposed to be up a solid 15%

You are very wrong there. AMD used 15% in performance/watt. Never IPC to direct performance.

 

[ShintaiDK]

For me it's simple. Will Piledriver be a noticeable improvement over my 1100 Tuban.

The answer will be no, even tho alot of people might be hopeful and try refuse that claim. But the uarch is just broken. AMD themselves says pilediver will be a 15% performance/watt improvement. To compare in that metric. IB is like 30% or so over SB. But not really a revolution performance wise. And remember Bulldozer is a massive powerhog, "easy" to improve on that.