AMD Q414 results

[Abwx]

You're missing what I'm saying.

Im saying Zen won't perform as well as Excavator since it's not aimed at even getting close to Intel's flagship.

When I say "jack of all trades, master of none", I really do mean master of none. It won't perform as well as Steamroller, nor scale as low as Jag, it'll perform and scale somewhere in between.

I selected your sentences to point where i do not agree.

Zen is supposed to replace Excavator, so why would they design it as less performing, why all the efforts since it would be much more cost efficient to use what they have in hand and wich would be supposedly better.

The real question is rather how much better it will be compared to Excavator and Steamroller, the low power cores dont even enter the equation for comparison since AMD clearly stated that Zen will be a high performance and high frequency design.

Since we're talking about a replacement for Bulldozer, this is an entirely different class of beast from the "Puma+" and Cortex-A57 cores in the first SkyBridge parts.

AMD's execs noted that these are high-frequency, high-performance CPU cores that will span the range from laptops to desktops to servers—and not just "microservers."

http://techreport.com/review/26418/amd-reveals-k12-new-arm-and-x86-cores-are-coming

Why did they need all those years to release Bulldozer? Time elapsed does not guarantee a good product.

You forgot that Bulldozer was devellopped in parrallel with Phenom and Bobcats, that s 3 different designs.

Also despite the ST perfs BD did reach its targeted throughput wich was to be equal to a 4C/8T Sandy bridge, that the software side of things didnt catch up as expected is another matter.

 

[monstercameron]

To me, Zen sounds a lot like a 14nm shrink for Jaguar. And you know Sony and MS are both going to be interested in a 14nm version of the 28nm chip they are already using in their consoles.

hmm, just a die shrunken jaguar? what about ipc improvements and feature additions like new instructions? sounds like a step back...

 

[elemein]

e AMD clearly stated that Zen will be a high performance and high frequency design.

Can I get a source to this and the context?

 

[Abwx]

Can I get a source to this and the context?

You didnt click on the link i provided as source.?.

All this is coming straight from Jim Keller and AMD s CTO mouths.

 

[Phynaz]

You didnt click on the link i provided as source.?.

All this is coming straight from Jim Keller and AMD s CTO mouths.

Yeah, AMD has never misrepresented the performance of one of their CPUs prior to launch.

 

[elemein]

You didnt click on the link i provided as source.?.

All this is coming straight from Jim Keller and AMD s CTO mouths.

You mean this?

http://translate.google.com/transla...-zen-forst-ut-i-summit-ridge-for-14-nanometer

Because... I don't see where it mentions anything about it being a high-performance arch...

Could be blind though.

 

[Abwx]

Yeah, AMD has never misrepresented the performance of one of their CPUs prior to launch.

They werent condemned in any class action, contrary to Intel, so your accusations are deffamation and nothing else.

Besides we re talking of future products, the absolute perfs doesnt matter for this discussion since we dont know where they did set the bar but for sure we can discuss the relative perfs, that is, better or worse than Steamroller/Excavator, i say Zen will be significantly better.

 

[monstercameron]

Yeah, AMD has never misrepresented the performance of one of their CPUs prior to launch.

maybe but what does the statement abwx have made have to do with the past. like stirring the pot I see....:sneaky:

 

[Abwx]

You mean this?

http://translate.google.com/transla...-zen-forst-ut-i-summit-ridge-for-14-nanometer

Because... I don't see where it mentions anything about it being a high-performance arch...

Could be blind though.

Here :

http://techreport.com/review/26418/amd-reveals-k12-new-arm-and-x86-cores-are-coming

 

[frozentundra123456]

I dont see any direct quotes. All I see is paraphrasing by the author of what the execs supposedly said, subject the the author's interpretation. It also says the ARM core and the x86 core will have similar performance. Does that make any sense? If true, there is no way it will be comparable to even quad core haswell, much less Skylake and hex/octa core intel chips.

 

[elemein]

Here :

http://techreport.com/review/26418/amd-reveals-k12-new-arm-and-x86-cores-are-coming

Hrm. That does skew my expectations now. I still don't see it going at Intel's offerings off a pure resource deficiency viewpoint, but maybe it is aimed to perform higher than what I previously expected. Thanks.

Though AMD's definition of "high performance" and "high frequency" have been a bit... underwhelming in the past. I know, that was then and this is now but can't help it.

 

[Abwx]

I dont see any direct quotes. All I see is paraphrasing by the author of what the execs supposedly said, subject the the author's interpretation. It also says the ARM core and the x86 core will have similar performance. Does that make any sense? If true, there is no way it will be comparable to even quad core haswell, much less Skylake and hex/octa core intel chips.

There s direct quotes in the article, and other sentences where the journalist is repeating their words since he says that he dont know how to interpret exactly thoses sayings.

A few exemples perhaps.?.

 

[Abwx]

Hrm. That does skew my expectations now. I still don't see it going at Intel's offerings off a pure resource deficiency viewpoint, but maybe it is aimed to perform higher than what I previously expected. Thanks.

Though AMD's definition of "high performance" and "high frequency" have been a bit... underwhelming in the past. I know, that was then and this is now but can't help it.

Dont know if AMD s "high frequencies" claims has been underwhelming, what about Bulldozer.?.

For the rest it s obvious from the article that they are cautious to not give too much infos, i dont pay attention to the journalist comments and interpretations, what matters is the enginers sayings.

As for the exact level of perfs i have no doubt that they can release a high performing CPU, we are no more in 2003 and designs tools have made huge progress that more than compensate for the increasing complexity of the designs, just check the Synopsys and other Cadence sites.

http://www.synopsys.com/home.aspx

 

[elemein]

Dont know if AMD s "high frequencies" claims has been underwhelming, what about Bulldozer.?.

For the rest it s obvious from the article that they are cautious to not give too much infos, i dont pay attention to the journalist comments and interpretations, what matters is the enginers sayings.

As for the exact level of perfs i have no doubt that they can release a high performing CPU, we are no more in 2003 and designs tools have made huge progress that more than compensate for the increasing complexity of the designs, just check the Synopsys and other Cadence sites.

http://www.synopsys.com/home.aspx

:| Obviously Bulldozer was high frequency, but it certainly wasn't high performance in the grand scheme of things. My opinion in that is concrete.

Also, yes, other designers, tool makers and foundries have come a long way, but it's more or less undisputed that Intel is ahead. In that, my opinion (as well as many others) is concrete.

Yes, my opinion on the matter has changed, but I don't think they'll actually be a performance threat to whatever Intel's flagship at the time is without having some gleaming deficiency (high TDP, crap iGP, etc. etc.) Though if they somehow surprise myself and everyone else, it'll be the biggest development since the K8.

 

[monstercameron]

:| Obviously Bulldozer was high frequency, but it certainly wasn't high performance in the grand scheme of things. My opinion in that is concrete.

Also, yes, other designers, tool makers and foundries have come a long way, but it's more or less undisputed that Intel is ahead. In that, my opinion (as well as many others) is concrete.

Yes, my opinion on the matter has changed, but I don't think they'll actually be a performance threat to whatever Intel's flagship at the time is without having some gleaming deficiency (high TDP, crap iGP, etc. etc.) Though if they somehow surprise myself and everyone else, it'll be the biggest development since the K8.

hmmm relative to what other cpus? amd is pretty much only beat by intel and maybe those monster IBM chips.

 

[mrmt]

So what market segment do you suggest Zen and K12 should target?

I think Zen/K12 are two huge engineering challenges. AMD should be targeting slightly above A53 raw performance *and* performance per watt but comparable costs. This is the only way I see for AMD to develop the micro server market and still be able to get embedded business. Fail to reach the performance goal and AMD won't be able to get into embedded business with any advantage over the drove of ARM MPU designers, fail to reach the cost target and they won't be able to leverage on the chip to expand the embedded business.

The only chips that made money for AMD in the last 4 years weren't the big ones, but the small, IP friendly, cost controlled cat family. I don't see why AMD would throw this solid and profitable track record for the sake of daydreaming about a performance race they are ill-equipped to stay, let alone win. Whatever they are designing, it will have the business features and parameters of the cat family, not of the derp or K8 family. The design will be IP friendly, highly synthesis-able, efficient and with the smallest possible die.

 

[Abwx]

:| Obviously Bulldozer was high frequency, but it certainly wasn't high performance in the grand scheme of things. My opinion in that is concrete.

It is a high performance CPU since it had as much throughput, and even more actualy, than a Sandy Bridge, but i guess that you dont hold it as a high perf CPU because of the ST perf.?.

If that s the case then you ll had to brand Intel s recent 12 cores Haswell as not being high perf CPUs at all since due to frequency they have not better ST perf than an AMD FX83xx, yet i did read nowhere in the press, or in this forum, that thoses 2650L, or 2697, were unworthy and overpriced CPUs...

Also, yes, other designers, tool makers and foundries have come a long way, but it's more or less undisputed that Intel is ahead. In that, my opinion (as well as many others) is concrete.

Yes, my opinion on the matter has changed, but I don't think they'll actually be a performance threat to whatever Intel's flagship at the time is without having some gleaming deficiency (high TDP, crap iGP, etc. etc.) Though if they somehow surprise myself and everyone else, it'll be the biggest development since the K8.

You cant expect to have a bad GPU on the AMD side, as for CPUs even their current designs have much more potentials than one could think, it s just that they are at a node disadvantage, Intel s better perf/watt is related to process and about nothing else, at equal nodes a 8C Steamroller would be more efficient than a i7, they have 15% voltage disadvantge wich induce, what, only 32% higher TDP...

 

[mrmt]

Though if they somehow surprise myself and everyone else, it'll be the biggest development since the K8.

That honor falls to Bulldozer. AMD spent more money with it, far more than K7 and K8, and they can't muster the same amount of resources to develop K12 and Zen now. Bulldozer will be the biggest AMD project for the foreseeable future, if not for the entire life of the company.

maybe but what does the statement abwx have made have to do with the past. like stirring the pot I see....:sneaky:

I think AMD executive team has a credibility problem, especially the management team that was doing tricks in order to reach sales forecasts, and misrepresenting the personnel cuts they were making to their investors. If they are doing that kind of things, to misrepresent the performance of their products would be a really small step.

 

[elemein]

It is a high performance CPU since it had as much throughput, and even more actualy, than a Sandy Bridge, but i guess that you dont hold it as a high perf CPU because of the ST perf.?.

If that s the case then you ll had to brand Intel s recent 12 cores Haswell as not being high perf CPUs at all since due to frequency they have not better ST perf than an AMD FX83xx, yet i did read nowhere in the press, or in this forum, that thoses 2650L, or 2697, were unworthy and overpriced CPUs...

You cant expect to have a bad GPU on the AMD side, as for CPUs even their current designs have much more potentials than one could think, it s just that they are at a node disadvantage, Intel s better perf/watt is related to process and about nothing else, at equal nodes a 8C Steamroller would be more efficient than a i7, they have 15% voltage disadvantge wich induce, what, only 32% higher TDP...

"Intel s better perf/watt is related to process and about nothing else,"

I hate when people say this. People are acting as if Intel using their process advantage which their engineers worked tirelessly over is some cheap, cheating, conniving and illegal trick that should be looked down upon. No.

Intel has that advantage. They worked for it. They deserve it. Whether "oh if PD had that advantage it'd be better!!" is true or not, that isn't how it is. So that argument is irrelevant.

And yes, I am saying it's not high performing because of the ST performance. As a consumer, myself and many (like many many... Like the overwhelming vast majority of computer users) use software that rely heavily to an incredibly high degree on singlethread performance. You can have an arch that has fourteen billion cores and have multithreaded performance the likes the Earth has never seen, but if no software takes advantage of it. It's useless. Irrelevant point.

The 2697 is Ivy Bridge and the 2650L is a 65W TDP CPU. What does that have to do with anything?

"Intel s recent 12 cores Haswell as not being high perf CPUs at all since due to frequency they have not better ST perf than an AMD FX83xx"

I don't understand what you're trying to say here. Are you trying to say that the FX83xx have higher ST performance than Haswell, while saying right before that they have less ST performance?

"You cant expect to have a bad GPU on the AMD side,"

Or I can expect it to have the iGP of the FX-8350. None.

"induce, what, only 32% higher TDP..."

What do you mean "only 32% higher TDP"? That's a whole 1/3. That's significant.

I'm sorry, your arguments previously were all explained and logical, but the ones detailed in your past post either don't make sense, conflict, don't have any sense of scale, or even, dare I say, a little desperate.

 

[Abwx]

"Intel s better perf/watt is related to process and about nothing else,"

I hate when people say this. People are acting as if Intel using their process advantage which their engineers worked tirelessly over is some cheap, cheating, conniving and illegal trick that should be looked down upon. No.

Intel has that advantage. They worked for it. They deserve it. Whether "oh if PD had that advantage it'd be better!!" is true or not, that isn't how it is. So that argument is irrelevant.

You are in full straw mode here, did i said that having a better process is cheating.?

Or is it the fact that i pointed that this is their main advantage, anyway you reaction is telling since you re litteraly doing deffamation out of things invented by you and that you re unshamlessly putting in mly mouth.

My point is that without process advantages their designs wouldnt be as good compared to the competition, that s a fact based by the known differences in processes.

And yes, I am saying it's not high performing because of the ST performance.

Here you are of the worst double standard one can imagine, do a rendering on Povray and compare this CPU to say a Pentium and then claim here and there that the Pentium has higher perfs because it does better in ST in a softs that is used in Mt anyway, you do rendering in ST..?.

According to you multicore haswell is not of high performance, and dont tell me again that it s 65W because you re just using the metrics that suit your views, so in a case you define high perf by perf/watt and in another case it s just single thread perf, talk of changing the rules once they dont yield the expected outcomes.

As a consumer, myself and many (like many many... Like the overwhelming vast majority of computer users) use software that rely heavily to an incredibly high degree on singlethread performance.

That s wrong, software that rely on ST dont need high ST perf otherwise they would be multithreaded but it just doesnt worth the effort, but tell us what are those tasks demanding ST perfs, i m quite curious about them.

What do you mean "only 32% higher TDP"? That's a whole 1/3. That's significant.

1V at 3.5GHz in a case and 1.15V in the other case, the difference will be 1.15 x 1.15 = 1.325.

I'm sorry, your arguments previously were all explained and logical, but the ones detailed in your past post either don't make sense, conflict, don't have any sense of scale, or even, dare I say, a little desperate.

What is deseperate is to change metrics once the single thread argument is exposed as moot, you re saying that a CPU with lower ST perf is of high perfs if it s an Intel CPU, that s you discourse in your post, besides you re saying that you re just a consumer, that s possible but a dont believe that it s actualy exactly the case.

 

[CHADBOGA]

That s wrong, software that rely on ST dont need high ST perf otherwise they would be multithreaded but it just doesnt worth the effort, but tell us what are those tasks demanding ST perfs, i m quite curious about them.



What is deseperate is to change metrics once the single thread argument is exposed as moot, you re saying that a CPU with lower ST perf is of high perfs if it s an Intel CPU, that s you discourse in your post, besides you re saying that you re just a consumer, that s possible but a dont believe that it s actualy exactly the case.

For as long as Intel is offering Quads and above in core count, then ST matters incredibly.

That is why we have seen AMD decimated in the desktop market, despite offering Hex & Octa cores.

 

[Abwx]

For as long as Intel is offering Quads and above in core count, then ST matters incredibly.

That is why we have seen AMD decimated in the desktop market, despite offering Hex & Octa cores.

Fair point of yours but the discussion was about ST being THE metrics that define a CPU perf, i think it s not stretching things that to state that it s one metric, although quite important, among others, no amount of ST perf will compensate if there s an application that just use 1.5-2 cores, yet most scale at this level including browsers like Firefox that fire 4 cores.

 

[elemein]

-snip-

"You are in full straw mode here, did i said that having a better process is cheating.? "

It's not strawmaning. You're making some hypothetical situation that isn't real, making it seem like Intel having their process advantage is somehow less legitimate than their arch. Intel has the process advantage. Anything that says "intel only has x because of their process advantage" is pretty pointless. They have the advantage and that's it. It's that simple.

"Here you are of the worst double standard one can imagine, do a rendering on Povray and compare this CPU to say a Pentium and then claim here and there that the Pentium has higher perfs because it does better in ST in a softs that is used in Mt anyway, you do rendering in ST..?."

No idea what you're referring to. What is "this CPU", do you mean an 8350? Compare it to a Pentium? ... What? Compare it to a Pentium. What? You want me to compare a CPU to another that has fewer cores, lower clock rate, less cost, much lower on the product line and then say "Yes this is a fair comparison?" No. Just no. Not happening. Sorry.

Also, since when does the vast majority of computer users render? It's pretty much concrete that most dont.

"according to you multicore haswell is not of high performance,"

...What? When? Where did I say this? What?

"and dont tell me again that it s 65W because you re just using the metrics that suit your views,"

You're comparing a processor to another processor that has a completely different TDP and aim, and spot in the product line then trying to say they're comparable. What? Okay, by that logic, let's compare The i7-5960X to the A4-4000 and go from there. What's that? Completely different TDP and not even near the same in product line? Oh don't worry about that, you're just trying to make your argument better by using a metric that suits your view better. An extreme example? Yes. But when A 65W TDP CPU goes up against a CPU with a 125 W TDP, you cannot possibly say "thats a fair comparison" with a straight face.

" talk of changing the rules once they dont yield the expected outcomes."

What rules? The rules are you have to be high performing. Period. Most hardware that is used by the vast majority of computers runs better on Intel hardware. Period. The rules are that simple.

"That s wrong, software that rely on ST dont need high ST perf otherwise they would be multithreaded, but it just doesnt worth the effort"

That's right. (/sarcasm) That must be why AMD is performing so well in the desktop market.

Oh wait. If it's not worth the effort, then it's not worth the effort, and if that's so, then how come Intel hardware performs better.

C-could it be?! Single threaded performance... matters?! What an epiphany!

"1V at 3.5GHz in a case and 1.15V in the other case, the difference will be 1.15 x 1.15 = 1.325."

Not sure where this math is coming from, but either way, 1.15V vs 1.325V is a pretty big difference... Wherever this math is coming from... I really don't know where you're trying to put this...

"What is deseperate is to change metrics once the single thread argument is exposed as moot"

Yep, ST performance is moot. Must be why AMD is doing so well.

Oh wait.

"you re saying that a CPU with lower ST perf is of high perfs if it s an Intel CPU, that s you discourse in your post"

What? I dont understand what you're trying to say. You're saying that I'm saying that a CPU with lower ST performance is high performance? Is it opposite day?

"besides you re saying that you re just a consumer,"

I am. Im a college student. Pretty consumer.

 

[Abwx]

No idea what you're referring to. What is "this CPU", do you mean an 8350? Compare it to a Pentium? ... What? Compare it to a Pentium. What? You want me to compare a CPU to another that has fewer cores, lower clock rate, less cost, much lower on the product line and then say "Yes this is a fair comparison?" No. Just no. Not happening. Sorry.

We re talking of the definition of high perf and you said that the metric is single thread performance, but now seems that you re protesting that core count, TDP and so on have also a say in this definition....

You're comparing a processor to another processor that has a completely different TDP and aim, and spot in the product line then trying to say they're comparable.

Why the need of all thoses metrics since ST perf is THE metric, at least according to your sliding rule former position....

"1V at 3.5GHz in a case and 1.15V in the other case, the difference will be 1.15 x 1.15 = 1.325."
Not sure where this math is coming from, but either way, 1.15V vs 1.325V is a pretty big difference... Wherever this math is coming from... I really don't know where you're trying to put this...

Dont know either from where it comes but i thought that a purely artificial argument would be left unoticed...

More seriously power increase as a square of voltage, take a 2GHz CPU, increase voltage by a 1.414 factor without changing the frequency, the dissipated power at 100% load will be double.

I am. Im a college student. Pretty consumer.

Ok, then i can only advise you to better follow the physics courses...

 

[VirtualLarry]

no amount of ST perf will compensate if there s an application that just use 1.5-2 cores, yet most scale at this level including browsers like Firefox that fire 4 cores.

Firefox is single-threaded, isn't it? I use Waterfox, a 64-bit port, and it uses a rough max of 50% of my dual-core 1007U CPU, when a thread gets "stuck" and starts chewing up CPU time.