AMD Q415 results

[Fjodor2001]

Maybe because intel never said that it was "their most significant processor". Some random Wiki contributer did that.

Multiple reliable sources has reported that Intel said it. The contributors to the Wikipedia article provided references to some of those sources, so it was not something the contributors made up themselves.

In fact, intel said.. "Introducing Our Best Processor ever"
http://www.intel.com/content/www/us/en/pc-upgrades/6th-gen-core-processor-family.html?cid=sem118p87779g-c&intel_term=intel%20skylake&gclid=CjwKEAiAiZK1BRD509nPsYiUk2YSJAAMoAwCNsMxA6atnd8NHplvSQUgk-RIaa6bN4ZNP9TCQZw47xoCQvPw_wcB

Intel says lots of things. Just because they have said what you referred to above does not mean they haven't made the statement referenced in the Wikipedia article too.

 

[ShintaiDK]

The quote you refer to was made up by Agam Shah. The author behind all the articles you used as reference.

It was never said by Intel. Then you can keep on derailing this thread as much as you want. But it doesn't make it right.

 

[ShintaiDK]

Perhaps this has been misquoted somewhere.

How about this quote, straight from the best source ever:

http://newsroom.intel.com/community...eration-intel-core-intels-best-processor-ever

Even that one is nowhere near the claimed one by Agam Shah that Fjodor keeps defending

And unless you go backwards, the next CPU is always the best.

 

[Fjodor2001]

The quote you refer to was made up by Agam Shah. The author behind all the articles you used as reference.

It was never said by Intel.

What evidence do you have of that? Do you have the sources that all referenced article authors used when writing their articles?

Oh, and by the way here are two more articles with a very similar quotes (not written by Agam Shah):

http://www.usatoday.com/story/tech/2015/09/02/5-things-know-intels-skylake-processors/71584764/
Says: "A “once in a decade” experience: That’s how Intel is describing the launch of its 6th generation “Skylake” Core processors".

http://www.channelweb.co.uk/crn-uk/...-2015-a-once-in-decade-high-for-pc-innovation
Says: "2015 will see a level of innovation experienced only once or twice each decade, thanks to the launches of Windows 10 and Intel's Skylake chip, a top Intel executive told delegates at the Intel Solutions Summit (ISS)."

So it seems quite clear that Intel has said something along those lines. Different authors have just phrased it a bit differently in their articles.

Now please stop derailing this thread further with your false denial.

 

[Dresdenboy]

Even that one is nowhere near the claimed one by Agam Shah that Fjodor keeps defending

And unless you go backwards, the next CPU is always the best.

Independently of any discussions going on here, this statement was not about having a better CPU, but having one of the most significant advancements.

6th Gen Intel Core processors deliver some of the most significant advancements in computing that we've ever seen

Maybe he was already referring to AVX512?

 

[Arachnotronic]

Perhaps this has been misquoted somewhere.

How about this quote, straight from the best source ever:

http://newsroom.intel.com/community...eration-intel-core-intels-best-processor-ever

Good post, Dresdenboy!

 

[Arachnotronic]

What evidence do you have of that? Do you have the sources that all referenced article authors used when writing their articles?

Oh, and by the way here are two more articles with a very similar quotes (not written by Agam Shah):

http://www.usatoday.com/story/tech/2015/09/02/5-things-know-intels-skylake-processors/71584764/
Says: "A “once in a decade” experience: That’s how Intel is describing the launch of its 6th generation “Skylake” Core processors".

http://www.channelweb.co.uk/crn-uk/...-2015-a-once-in-decade-high-for-pc-innovation
Says: "2015 will see a level of innovation experienced only once or twice each decade, thanks to the launches of Windows 10 and Intel's Skylake chip, a top Intel executive told delegates at the Intel Solutions Summit (ISS)."

So it seems quite clear that Intel has said something along those lines. Different authors have just phrased it a bit differently in their articles.

Now please stop derailing this thread further with your false denial.

Dude, are you serious? The "once in a decade" that they are referring to is to the fact that Intel is putting out a brand new platform around the same time that Microsoft is putting out the OS that is intended to fix Windows in the eyes of consumers.

This is a "once in a decade" experience.

They're not saying, "ONCE IN A DECADE IPC JUMP 4 j00!"

 

[Sweepr]

Says: "2015 will see a level of innovation experienced only once or twice each decade, thanks to the launches of Windows 10 and Intel's Skylake chip, a top Intel executive told delegates at the Intel Solutions Summit (ISS)."

Thanks for clarifying Fjodor. So it' all about Skylake + Windows 10 integration, not exactly performance like you've been quoting.

And unless you go backwards, the next CPU is always the best.

Indeed, and Skylake is better than its predecessors (especially mobile).

 

[frozentundra123456]

Does it really matter what some executive said or didn't say? We are arguing about a non-specific statement anyway, which could be interpreted in a variety of ways, if it was even made.

Even then, the usual suspects that accuse others of derailing AMD threads are doing it themselves to try to say skylake was overhyped.

 

[ShintaiDK]

Independently of any discussions going on here, this statement was not about having a better CPU, but having one of the most significant advancements.



Maybe he was already referring to AVX512?

My mistake

Talking about computing. Have you heard anything if AMD gets 256bit paths for (fast) AVX2/FMA3 loads?

 

[Fjodor2001]

Thanks for clarifying Fjodor. So it' all about Skylake + Windows 10 integration, not exactly performance like you've been quoting.

First, the claim has never been about performance specifically, despite that you repeatedly have been lying about me saying so. The claim from the articles have been about Skylake being the "most important architecture/chip/whatever in a decade". Not performance specifically. So to begin with, can you please stop lying about that?

Then you're only referring to one of the articles, which also mentioned Win 10. The other ones don't.

To sum it up we have multiple independent and credible sources all saying basically the same thing:

http://www.pcworld.com/article/2893112/intels-skylake-chips-to-appear-in-tablets-pcs-servers.html
Says: "Intel has hailed Skylake as its most important chip architecture in a decade"

http://www.channelweb.co.uk/crn-uk/n...-pc-innovation
Says: "2015 will see a level of innovation experienced only once or twice each decade, thanks to the launches of Windows 10 and Intel's Skylake chip, a top Intel executive told delegates at the Intel Solutions Summit (ISS)."

http://www.usatoday.com/story/tech/2...sors/71584764/
Says: "A “once in a decade” experience: That’s how Intel is describing the launch of its 6th generation “Skylake” Core processors".

The wording is a bit different, likely because it's something that has been spoken by the Intel representative at a speech or in individual interviews with the journalists. So the articles phrase it a bit differently, but the intention is the same.

Merge them together and we get something like:

"Intel claims that Skylake is its most important chip/experience/architecture in a decade"

The point is obvious regardless. But unfortunately Skylake fails to fulfill that promise for most users.

 

[VirtualLarry]

Thanks for clarifying Fjodor. So it' all about Skylake + Windows 10 integration, not exactly performance like you've been quoting.

Indeed, and Skylake is better than its predecessors (especially mobile).

I believe that comment, if it were ever made, was referencing the SGX extensions in SKL.

Which I mentioned before in this thread, but I guess you'all have me on ignore?

 

[Sweepr]

First, the claim has never been about performance specifically, despite that you repeatedly have been lying about me saying so. The claim from the articles have been about Skylake being the "most important architecture/chip/whatever in a decade". Not performance specifically. So to begin with, can you please stop lying about that?

Cut down on your personal attacks. There's evidence all over this and other threads that you've been interpreting that statement as performance improvement.

There was no limit on how Skylake performed.... until release.

Most important architecture in a decade and what not.

Carefull selection of words won't change it.
And you still haven't provided the quote where someone at Intel said it.

The wording is a bit different, likely because it's something that has been spoken by the Intel representative at a speech or in individual interviews with the journalists. So the articles phrase it a bit differently, but the intention is the same.

Can you prove it? Please show us Intel representative's word about it Fjodor.
Are you hiding it from us?

You should save you energy for the damage control regarding another product launching later this year. Bad mouthing a great chip like Skylake unfortunately won't improve AMD's competitiveness.

I believe that comment, if it were ever made, was referencing the SGX extensions in SKL.

Seriously could be many different things. Since we have no transcript of what intel actually said, it's up to our own interpretation. Edit: assuming it is real.

 

[Dresdenboy]

My mistake

Talking about computing. Have you heard anything if AMD gets 256bit paths for (fast) AVX2/FMA3 loads?

For Zen no sign. I think that's not AMD's strategy to scale SIMD in a single core (still relying on HSA for that stuff -> Exascale). Instead at fixed frequency and power they might just throw more cores at it. Not so dense FP code (e.g. CB) should be fine, dense FP code (Linpack, Prime95, BLAS) will be significantly slower than HSW/SKL with just a single thread. MT should be fine (moar coars).

I think even Zen+ won't change that.

 

[ShintaiDK]

For Zen no sign. I think that's not AMD's strategy to scale SIMD in a single core (still relying on HSA for that stuff -> Exascale). Instead at fixed frequency and power they might just throw more cores at it. Not so dense FP code (e.g. CB) should be fine, dense FP code (Linpack, Prime95, BLAS) will be significantly slower than HSW/SKL with just a single thread. MT should be fine (moar coars).

I think even Zen+ won't change that.

That's a smart move from a consumer segment perspective. My 6700K could have been labelled with something like 65W TDP without. Same reason AVX512 wont hit client space. So that will give them a slight edge in perceived TDP. Tho that wont reflect actual usage. But its something they can promote marketing wise.

 

[Dresdenboy]

That's a smart move from a consumer segment perspective. My 6700K could have been labelled with something like 65W TDP without. Same reason AVX512 wont hit client space. So that will give them a slight edge in perceived TDP. Tho that wont reflect actual usage. But its something they can promote marketing wise.

A lot of server applications with little need of FP throughput would benefit, too. TDP has to be seen, but also measured TDP would be lower than with a 256b SIMD block - and probably increased again due to the additional cores. That's a case for power management.

I'll cover that topic soon.

 

[itsmydamnation]

Also remember FP and SIMD aren't always the same thing. Zen could be very good at FP with many low latency FP execution resources. It looks like FMA (looks like 1 a clock) and 256bit ops that will suffer vs intel. SSE should also be a strength vs intel but we will have to wait and see.

Lots of Server software is FP heavy without being SIMD.

 

[Dresdenboy]

Also remember FP and SIMD aren't always the same thing. Zen could be very good at FP with many low latency FP execution resources. It looks like FMA (looks like 1 a clock) and 256bit ops that will suffer vs intel. SSE should also be a strength vs intel but we will have to wait and see.

Lots of Server software is FP heavy without being SIMD.

FMA is also 2 issue for Zen. They also seem to use an rectangular multiplier (like the Cat cores). I'm currently checking whether they opted to include a so called High Throughput FP unit, which can overlap instructions, getting about the same latencies for MUL+ADD as for FMA. ADD+ADD would be faster, too.

 

[NTMBK]

For Zen no sign. I think that's not AMD's strategy to scale SIMD in a single core (still relying on HSA for that stuff -> Exascale). Instead at fixed frequency and power they might just throw more cores at it. Not so dense FP code (e.g. CB) should be fine, dense FP code (Linpack, Prime95, BLAS) will be significantly slower than HSW/SKL with just a single thread. MT should be fine (moar coars).

I think even Zen+ won't change that.

Makes sense. If you have a workload which benefits from ultra-wide FP vectors with FMA, buy the HPC APU.

 

[Dresdenboy]

Makes sense. If you have a workload which benefits from ultra-wide FP vectors with FMA, buy the HPC APU.

In ST (no SMT) the total FP t'put with overlapping (lower effective latency) and "auto-FMA-conversion" might catch up to a 8DP/cycle FMA FPU.

 

[Fjodor2001]

Cut down on your personal attacks. There's evidence all over this and other threads that you've been interpreting that statement as performance improvement.

When you repeatedly claim I have said things I haven't, that's you making a personal attack on me. What you quoted from me is nothing at all like what you accused be of having said. And there is no such evidence, or you would have posted it long ago. So again, please just stop lying about that.

And you still haven't provided the quote where someone at Intel said it.

Can you prove it? Please show us Intel representative's word about it Fjodor.
Are you hiding it from us?

All the referenced articles specifically say it's Intel that has said what they write in the article, so the source of the statements is well proven. And the statements from Intel in the articles say basically the same thing.

So we've got multiple credible sources all confirming it, yet you're still in denial.

But it's quite obvious you already made up your mind long ago to stay in denial about this no matter what evidence is put forward. You'll just keep asking for even more evidence. It wouldn't matter if someone posted a video from a court of law where Intel's CEO Brian Krzanich under oath confessed to it. You would still not accept it. Maybe his tie looked a little weird, it must have been a fake video.

Then at the same time we have IST members who on a daily basis fill the forum with controversial fanboy claims that they say they are dead certain about, but provide no sources at all to back it up. You can check my sig for a small selection of such ones. However then I don't hear you complaining or asking for extremely well documented sources, or even any sources at all. Talk about double standards.

Anyway, I don't intend to argue about this any further with you. If you want to stay in denial that's fine by me.

 

[Fjodor2001]

Regarding Zen server CPUs, IIRC when I first heard about Zen, AMD was supposed to release Zen for servers first, and then later for desktop. But now that has changed to the opposite.

First: Do I remember this correctly? If I do, are there any reasonable conclusions to draw from that? Server CPUs taking longer time to design due to being more complex? Zen being more competitive in HEDT segment than server segment...?

 

[Arachnotronic]

So we've got multiple credible sources all confirming it

Echo chamber.

 

[mrmt]

First: Do I remember this correctly? If I do, are there any reasonable conclusions to draw from that? Server CPUs taking longer time to design due to being more complex? Zen being more competitive in HEDT segment than server segment...?

AMD will sell the same server die as a desktop chip, so the validation process will be the same.

You have a point when you say that Zen will be more competitive as a HEDT chip, if it can reach the performance levels for this segment. Servers is all about efficiency, because low efficiency kills TCO, and platform features, I don't think AMD will bring anything close to the Purley platform and desktop users are more than willing to throw away efficiency in the name of raw performance, and even if the performance is inferior AMD can win consumers on price. They have a lot more of latitude on the desktop market.

 

[svenge]

Here's an interesting question: Why does it seem that Eastern Europe and Russia are so heavily over-represented in the ranks of the ADF? It's been 25 years since the Soviet Union collapsed, so you'd think they'd get over their fondness of crappy electronics.

Or is it that AMD's lousy performance reminds them of the old Elektronika computers from their childhood and AMD's consistent party line (which always stays the same regardless of actual conditions in reality) is comforting somehow?


Racist comments are not allowed in a technical forum
Markfw900