Piledriver not coming until 2013 now?

[Ferzerp]

I guess what I'm saying is the current video card scape has appeared stagnant, and it's really a shame. We know AMD is holding back on the 79xx series' capability. It *could* be a big leap, if they wanted it to, but it appears that it's being priced and released at the speeds it is to milk every bit of profit possible out of it.

Anyway, this is a CPU thread, I just got off topic with the whole competition thing. I hope you get what I'm trying to say. I could almost run at 2560x1600 with a 480 with most games when it came out (have to turn down some settings), and here we are in 2012, and I still can almost run at 2560x1600 (have to turn down a few settings (less than with the 480, but still I have to turn things down) if I buy a 7970.

 

[janas19]

Can someone explain to me, in a nutshell, why Bulldozer's performance was admittedly lackluster? I'm not talking about corporate culture, or Intel vs AMD, I mean strictly from a technical point of view. Was it the transistors per core, or the instruction set, or something else?

/thread hijack

 

[Ferzerp]

Slow cache, and trying to bring back the Ghz race but with core count this time made them have some pretty big negatives that they weren't expecting? The two cores mated together caused a pretty massive performance penalty. The module thing just doesn't appear to work well in practice as implemented. I still feel it would be much better received if they'd have decided to market a module as an SMT capable core.

But I am no processor design expert at all. That's just my layman guess. The reason I suspect it is the case, is because they already have a better core as far as IPC goes, and a die shrink + tacking 2 more cores on (from x6 up to x8) would have outperformed BD, and would have costed much less to develop, I am sure. It is kind of like when intel abandoned an architecture for something that was slower per clock, in the hope of making a big number (Ghz in that case). AMD has fallen for the same thing, but cores.

 

[janas19]

Ah, I see. Two cores sharing the same resources is what it sounds like.

I guess it has potential, just AMD couldn't reveal that full potential from it's inception. Which is understandable.

 

[Ferzerp]

Well, we know they can make a traditional core that performs better than a BD "core". So, that IPC loss had to come from somewhere? What was the major change? The module concept.

Yes, the chance does exist for them to work out why the cost is so high *if* it is not just something inherent with the whole design philosophy.

The odd thing is if they made an 8 core thuban it would have performed better and had a similar transistor count.

 

[janas19]

I think they could work out the module concept. Pretty much any design concept will have certain flaws, but sustained development will keep improving it... If AMD decides to move forward on it, I'd be excited about it.

 

[Ferzerp]

I suggest caution. Save your excitement until after any results are had. Latching on to tech blindly makes for some bad situations once it is released. Ignore hype, wait for results.

 

[janas19]

I suggest caution. Save your excitement until after any results are had. Latching on to tech blindly makes for some bad situations once it is released. Ignore hype, wait for results.

True that.

 

[eternalone]

Is it confirmed not coming out this year????

 

[ShadowVVL]

What about bobcat, doesn't it use the same setup 2 cores on 1 module sharing resources?

It seemed to do well imo.

 

[AtenRa]

Lets face it, AMD no longer has any interest in us. All they're going to be offering us is a low end server CPU that can't run games for dog crap.

http://www.anandtech.com/show/5448/the-bulldozer-scheduling-patch-tested/3

I believe you will find that DX-11 games (Today and in the future) will need more GPU performance than CPU IPC

Slow cache, and trying to bring back the Ghz race but with core count this time made them have some pretty big negatives that they weren't expecting? The two cores mated together caused a pretty massive performance penalty. The module thing just doesn't appear to work well in practice as implemented. I still feel it would be much better received if they'd have decided to market a module as an SMT capable core.

I believe you should check Bulldozer's CMT Scaling vs CMP and against Intel's SMT. CMT works fine, it is the low single thread performance that hurts Bulldozer.

But you cannot have two things at the same time, AMD went for high Multithreading scaling and higher core count with Bulldozer than high IPC design like Intel.
When the IPC difference is at or bellow 35% between Bulldozer and SandBridge, AMDs Module CMT design equals Intel's Core + HT SMT design both in performance and die space.
When the difference is bellow 25% the CMT architecture is faster

What about bobcat, doesn't it use the same setup 2 cores on 1 module sharing resources?

It seemed to do well imo.

BobCat is different architecture than Bulldozer, more like a traditional single core design.

 

[IntelUser2000]

let's see...core 2 duo realesed in 2006
in 2006 ati showed X1900

whitch one more than triple the performance?

There's a difference between the two. 3x increase in CPU performance increase performance in a LOT of applications. 3x increase in GPU performance only impacts gaming performance. The more specialized you get, the easier it is to get performance out of it.

 

[Medu]

Why would anyone care when they came out? A 10-15% improvement in performance per watt over the current cores when they need closer to a 100% improvement. AMD are finished in the standalone CPU department.

 

[BD231]

We all really just need to stop talking about BD, it was bad when it came out, it's bad now and it'll take a miracle to fix with all those slow moving parts it has. There's like three guys on this entire form who even care to discuss the positive in it all and they're all biased due to ownership. If you're foolish enough to believe piledriver will fix the IPC issues or even make BD a viable upgrade from the phenom II series you're all in for a major disappointment.

The type of performance AMD is aiming for with BD is of no use or benefit to a typical desktop user/gamer.

 

[AtenRa]

There's like three guys on this entire form who even care to discuss the positive in it all and they're all biased due to ownership.

So, owners of Intel CPUs are biased too when they highlighting the positives of their CPUs right ??

 

[Sweepr]

Yes, lets completely ignore this..

...the fact that Intel x86 cores have a ~50% IPC advantage right now and the fact that they'll increase IPC even more + cut TDP by 20% in about 2 months...

Seriously, AMD really thinks Vishera (10-15% performance boost over Zambezi) will be enough to face Haswell in 2013 or they're slowly losing interest in the high-end non-APU desktop market?

 

[toolbag]

Don't waste time in your life always waiting for the next cpu. Life has so much more to offer.

 

[Olikan]

I guess what I'm saying is the current video card scape has appeared stagnant, and it's really a shame. We know AMD is holding back on the 79xx series' capability. It *could* be a big leap, if they wanted it to, but it appears that it's being priced and released at the speeds it is to milk every bit of profit possible out of it.

but what i am saying is that since 2006 cpus are stagnant...we only see some 10% to 25% increase in each generation, yet the TPD is decreasing

while gpus are stagnant since 2009, but the limit is more about the TPD and bandwidth, than nvidia\amd desire

There's a difference between the two. 3x increase in CPU performance increase performance in a LOT of applications. 3x increase in GPU performance only impacts gaming performance. The more specialized you get, the easier it is to get performance out of it

you know, with some bios and kernel updates, and a motherboard that support IOMMU, the 7970 could actually boot up linux

 

[BD231]

So, owners of Intel CPUs are biased too when they highlighting the positives of their CPUs right ??

Only if their perspective goes completely against the general consensus

 

[Nemesis 1]

Well that works out fine, i've been waiting for them to axe it completely and focus on APU's where the real performance growth exists. Maybe time for an APU section here on Anandtech?

LOL I have said that for years. That WE need seperate Intel/AMD and NV/AMD CPU and GPU sectionns .

APU = AMD PROGRAMMING UNIT. Intel Isn't ever going to call their IGPU an APU not Ever. IPU yes APU not ever. I thought I was a fanboy . NO no I am not . YOU are a fanboy. I going to love what IB Igpu does to your smallish world.

 

[Ferzerp]

To the guy who got latched on to "CMT scaling". That's absolutely not what I am talking about. I am talking about the penalty for that system, not how it scales.

We know AMD can do better as far as IPC, their last gen was better, at a similar transistor per core cost. I'm saying, when they sandwich them together, the compromises they have to make, makes for crappy single-threaded capability with their current implementation.

 

[Nemesis 1]

Vishera v.s. Ivy Bridge...interesting.

NO no its not at all . Vishera Vs. Haswell . What happened to trinity vs. IVB ? or is IB have to go against 2 generations of amd APUs . LOL.

 

[Rvenger]

APU = AMD PROGRAMMING UNIT.

Source = http://www.amd.com/us/products/technologies/fusion/Pages/fusion.aspx

"The AMD Fusion APU (Accelerated Processing Unit) allows you to configure small-form-factor, long-battery-life PCs that consumers crave, while enabling the powerful, next-generation visual experiences that can set you apart and help create a competitive advantage in a hyper-competitive market."

YOU are a fanboy.

Oh and quit crapping in my thread.

 

[Nemesis 1]

Not crapping Accelerated Processing Unit was coined by AMD . So AMD PrOGRAMMING Unit is the way INTEL see it . Intel will never use APU , Ever . and I not crapping on your thread . Just tring to keep people honest here . I was one of the few who went against The AMD marketing hero hereat AT . I even took ban points . But in the end it was all covered over. AMD has no fusion process. AMD switched to to what a new method for APUs more akin to what intel has been working on since 2004 when Intel bought Elbrus. Intel is way ahead in this area as the compiler came befor the chip.

 

[Olikan]

Not crapping Accelerated Processing Unit was coined by AMD . So AMD PrOGRAMMING Unit is the way INTEL see it . Intel will never use APU , Ever . and I not crapping on your thread . Just tring to keep people honest here . I was one of the few who went against The AMD marketing hero hereat AT . I even took ban points . But in the end it was all covered over. AMD has no fusion process. AMD switched to to what a new method for APUs more akin to what intel has been working on since 2004 when Intel bought Elbrus. Intel is way ahead in this area as the compiler came befor the chip.

HSA vs AVX-2 is going to be a hell of a fight