AMD CPU speculation... and expert conjecture

[jdwii]

Kirby and Intel are probably two of the most unethical yet rich companies, I am guessing its alright since they make you fast CPU's we can just turn a blind eye.

You can bet any company that has made over a billion dollars has done something unethical. That's just the nature of capitalism.

Its not like Amd's perfect, with basically lying in marketing.

 

[noob2222]

+1

The upvote/downvote functionality clearly needs to be returned to the forums.

http://www.youtube.com/watch?v=KE6pIMqMQ2E He says "In our new product next year, codenamed Kaveri." So it's 2014 now?

that would be a bit comical considering they released the new "roadmap" just yesterday with kaveri on there for 2013. http://techreport.com/news/24883/amd-roadmap-suggests-kaveri-apus-will-use-new-fm2-socket

he probably was talking "next year" meaning thats all thats coming in 2014. I don't expect to see any excavator APU till at the earliest, late late 2014, so next year is the year of kaveri APU. Im expecting a slow to limited release in 2013 due to 28nm transition, it always is buggy or low yield at first, esp if its GF.

I don't expect FM2+ to be much more than gddr5 support option.

BTW hcl123, don't forget the ever lasting ... "your just a conspiracy theorist, there is no proof, therefore its a lie." That one always cracks me up.

Personally im not sure who is worse tho, Apple or Intel. Apple is abusing the patent system to block competition, Intel just used their own money instead of the governments. I can at least have a tiny bit if respect for Intel for that aspect.

 

[mayankleoboy1]

Could it be that the secret is a fully working RCM technology, that apparently wasnt working in PD. That could explain the large clock increase, in the sam TDP.

 

[mayankleoboy1]

Yes S|A is in the business of selling *rumors*...

Atleastfor the last 2-3 years, his rumours have been spot on, and usually several months before anyone else has a clue about it. He bases his rumours on inside leaks (moles) and technical + financial prediction.

I'm not judging, but i stand in complete awe ! ... who in their right mind would buy a rumor! ... and i though IT was about *science* not rumors.

He is already making large amount of money from this model, and most paying members are part of a tech company....

If that proves to be a consistent business model, then everything is lost, because the parties with the most cloth and manipulative power, by rumor, intrigue and deception will end up winning *EVERYTHING*...

This already happens now.

 

[noob2222]

A bit off topic, but funny news. MS lost the Start battle.

PCs today are evolving for a world of mobile computing where people interact with their devices through touch, and we designed Windows 8 for this. But we also recognize there are many non-touch devices in use today – especially in the commercial setting. As such we’ve focused on a number of improvements to ensure easier navigation for people using a mouse and keyboard.

http://blogs.windows.com/windows/b/bloggingwindows/archive/2013/05/30/continuing-the-windows-8-vision-with-windows-8-1.aspx

Return of the start button may convince OEMs to start adding more windows 8.1 machines, end users ... will still be skeptical at best.

 

[Cazalan]

Could it be that the secret is a fully working RCM technology, that apparently wasnt working in PD. That could explain the large clock increase, in the sam TDP.

Possibly, but PD saw a sizable TDP change from BD as well. CPUs have like 12 layers of metal now. They're small 3D jigsaw puzzles. There's always ways to tweak the layout.

Maybe some P states were broken or more were added like for 7790. That would have to be listed in the errata for the part.

 

[hcl123]

Could it be that the secret is a fully working RCM technology, that apparently wasnt working in PD. That could explain the large clock increase, in the sam TDP.

No.

Its the SOI process alright, that was revamped. Actually attending that die shot of SR, i'm convinced that BDver1 was not truly 32nm SOI, but an hybrid of 45nm, since SR packs a lot of double structures by that die shot, yet the "module" manages to be even a little smaller(making good faith on the images).

Just consider that 32nm SOI, which by this proves is better than 22 nm bulk finfet of Intel for high performance, is the process used by IBM to put a "commercial" 600mm² chip at 5.5Ghz ( no not typos 600mm² and 5.5Ghz).

That is something that Charlei of S|A will never admit

http://semiaccurate.com/forums/showpost.php?p=151814&postcount=24

Let me repeat myself one more time.

THERE IS NO SOI UNDER 28nm FROM ANYONE, PERIOD.

If anyone is unclear, please ask again, so I can get annoyed.

-Charlie

FD-SOI ARM-based SmartPhone Chip Hitting 3GHz in Barcelona – But Wait: It’s the Low Active Standby Power (0.6V for 1GHz) That’s Really Amazing!
http://www.advancedsubstratenews.com/2013/02/fd-soi-arm-based-smartphone-chip-hitting-3ghz-in-barcelona-but-wait-its-the-low-active-standby-power-0-6v-for-1ghz-thats-really-amazing/

LOL

About the so called "temperature" oriented turbo, all it takes is a serious look at official AMD slides, Trinity already had that
http://www.notebookcheck.net/AMD-announces-new-A-Series-APU-codenamed-Richland.89691.0.html

image
http://www.notebookcheck.net/fileadmin/_migrated/pics/8_01.png

Perhaps there has been a lot of money for the take to whitewash intel finfet process, and now everybody parrots that intel has the best process, when its AMAZINGLY EVIDENT that is not truth, neither for high performance where 32 PD-SOI rules, neither for low power where STMicro 28nm FD-SOI is absolute king...

A REAL SECRET lost in the UNMEANING benchmarks is ;
http://www.guru3d.com/articles_pages/core_i5_3570k_review.html

Performance wise Ivy Bridge series never did stun and shock like the Sandy Bridge series did once they where introduced. In fact clock for clock it is all roughly the very same, with a +0.8% performance offset in favor of Ivy Bridge.

And is worst than that, because IB has many issues with "hot spots" so tends to OC worst.

So much for the great finfet mantra!...

[EDIT: a great conclusion that we can take of all this, is that intel never fails, even when it fails in the most miserable way ]

[EDIT2: how did i come out with all this!?... just look the links, i'm not in a "campaign", just started to avoid the sites that many now suspect of biasing... far better than entering in pointless bickering]

 

[hcl123]

Yes S|A is in the business of selling *rumors*...

Atleastfor the last 2-3 years, his rumours have been spot on, and usually several months before anyone else has a clue about it. He bases his rumours on inside leaks (moles) and technical + financial prediction.

I hope my previous post just shatters that myth about long before time infallibility of S|A... and people now more and more in several places take issues with possible(and evident) biasing on charts and numbers, but want to force themselves in the same mistakes over and over... if so why wouldn't they even try to be worst and flat out deceive ?

 

[truegenius]

i liked the last ubuntu i tried. imo it's good for newbies. but i liked puppy linux more even though i am a linux newbie myself. it helped me to resuscitate a couple of dead pcs. and often helps me with pc maintenance. :lol: i think the latest ubuntu puts a bit too much pressure on old gfx.

linux everytime !

ok, which linux or ubunto (if they are seperate things) will you recommend for a beginer (with no experience of any other os except windows) ?????
i will give it a try

Personally I use Scientific Linux. That's more or less the same as CentOS except it sounds cooler and has a better logo.

If you're just starting out I suppose you would do well to try Mageia.

http://www.mageia.org/en/

Go for the KDE version.

I personally like Mint much more and think its better than Ubuntu when it comes to noobs.

I personally like Mint much more and think its better than Ubuntu when it comes to noobs.

+1

ubunto works fine and gives pretty much a windows like experience.

linux everytime !

ok, which linux or ubunto (if they are seperate things) will you recommend for a beginer (with no experience of any other os except windows) ?????
i will give it a try

Ubuntu 13.04 is the most stable, user friendly release available. You can get it from Canonical for something like $5 (If you want a disk, if you have the freespace, it's a free download and you can make your own OS disk or even flash drive). It's a very user friendly design with an easy to navigate UI, you'll need to get familiar with some of the commands in the OS, but they're not terribly complicated.

linux everytime !

ok, which linux or ubunto (if they are seperate things) will you recommend for a beginer (with no experience of any other os except windows) ?????
i will give it a try

Ubuntu is a distribution of Debian Linux, one of the first of its kind. I would suggest trying out Linux Mint, a distribution of Ubuntu with amazing visuals and said to be one of the best distributions of the former, you can install Ubuntu and Mint as a Windows Application as well for east deletion.

Someone with no knowledge of the OS trying Mint 14 out. http://www.youtube.com/watch?v=I-PTQVeCF9k

http://www.linuxmint.com/

linux everytime !

ok, which linux or ubunto (if they are seperate things) will you recommend for a beginer (with no experience of any other os except windows) ?????
i will give it a try

Ubunto is pretty much the "go to" starter linux. After your comfortable working inside the environment and know how to make a proper build chain you can get a bit more creative. Personally I prefer CentOS though I tend to heavily customize it after install. Debian has by far the largest software base but you'll run into problems with drivers and such due to their licensing model. Debian requires all code to be opensource, hardware venders don't always like to release their IP (they spend a lot of money developing optimized drivers) as OS for their competitors to copy.

thanks everyine for your suggtestions
i will try them all (puppy linux, ubuntu 13.04, mint, centos, mageia)

 

[gamerk316]

That is THE ONLY THING that users can do, fight for immutable benchmark software, anything else you can't force no one to nothing (its not legal even).

Question: Wouldn't you then be complaining that AMD is being hurt because they decided to shift their focus away from X87 processing?

 

[gamerk316]

Just consider that 32nm SOI, which by this proves is better than 22 nm bulk finfet of Intel for high performance, is the process used by IBM to put a "commercial" 600mm² chip at 5.5Ghz ( no not typos 600mm² and 5.5Ghz).

Power7 based, which is an architecture that is designed to scale in clock. Can't compare architectures on clock alone.

 

[esrever]

+1

The upvote/downvote functionality clearly needs to be returned to the forums.

http://www.youtube.com/watch?v=KE6pIMqMQ2E He says "In our new product next year, codenamed Kaveri." So it's 2014 now?

It also says that richland has gcn. Which I really doubt.

 

[8350rocks]

+1

The upvote/downvote functionality clearly needs to be returned to the forums.

http://www.youtube.com/watch?v=KE6pIMqMQ2E He says "In our new product next year, codenamed Kaveri." So it's 2014 now?

It also says that richland has gcn. Which I really doubt.

Originally, there was speculation Richland would have GCN, it has since been shown that it does not...however, it does boast dramatically improved graphics.

Kaveri, on the other hand, will be GCN.

 

[anxiousinfusion]

+1

The upvote/downvote functionality clearly needs to be returned to the forums.

http://www.youtube.com/watch?v=KE6pIMqMQ2E He says "In our new product next year, codenamed Kaveri." So it's 2014 now?

It also says that richland has gcn. Which I really doubt.

He only says "8000 series".

 

[esrever]

+1

The upvote/downvote functionality clearly needs to be returned to the forums.

http://www.youtube.com/watch?v=KE6pIMqMQ2E He says "In our new product next year, codenamed Kaveri." So it's 2014 now?

It also says that richland has gcn. Which I really doubt.

He only says "8000 series".

the interview guy asked if its GCN and he agreed.

 

[esrever]

http://www.computerbase.de/news/2013-05/amds-kaveri-apu-bekommt-einen-neuen-sockel/
looks like AMD roadmap still shows kaveri for 2013

 

[hcl123]

That is THE ONLY THING that users can do, fight for immutable benchmark software, anything else you can't force no one to nothing (its not legal even).

Question: Wouldn't you then be complaining that AMD is being hurt because they decided to shift their focus away from X87 processing?

That is a side issue... x87 has very few uses now, lest of all in the context of "client/desktop" systems.

I think you ask about superpi right ?... meaningless on top of meaningless. Yet the worst meaningless is that superpi blob has no md5 signature, you can't even have any assurance its the superpi real superpi (very easy to fake what it does), so 2 tests of superpi, can be 2 tests of 2 different things.

This is the most generalized sad part of this story.

Just consider that 32nm SOI, which by this proves is better than 22 nm bulk finfet of Intel for high performance, is the process used by IBM to put a "commercial" 600mm² chip at 5.5Ghz ( no not typos 600mm² and 5.5Ghz).

Power7 based, which is an architecture that is designed to scale in clock. Can't compare architectures on clock alone.

Right. But you can have a very serious perspective about the worth of a fab process with the clock. And in here we are talking about high performance high clocks were power is a secondary issue... like with GPUs.

And its not Power uarch, is Z uarch, which is a speed daemon uarch resembling intel P4 in some aspects ( what for some is no good for others can be marvelous)

http://www.hotchips.org/wp-content/uploads/hc_archives/hc24/HC24-9-Big-Iron/HC24.29.930-zNext-Shum-IBM.pdf

Which leads to suspect... again IT is full of propaganda and myths... the worst failure of P4 was not the design, was the fab process, and the design felt strongly those restraints (Prescott would never had to be, or a very different Prescott would had been possible, if intel had chosen SOI for fab).

But this already lost context in time alright, you can say its a supposition of mine, but i feel very positively about it.(edit)

 

[de5_Roy]

http://www.computerbase.de/news/2013-05/amds-kaveri-apu-bekommt-einen-neuen-sockel/
looks like AMD roadmap still shows kaveri for 2013

heh, seems like xbitlabs link that i read and posted 'creatively modified' the leaked slides. others posted uncropped images. if i am to believe that, then kaveri is on track for 2013 release while steamroller opterons are gone (i.e. zero number of sr cpu). +NaCl

@richland: i can remember as far back as early trinity-rumors and roadmaps where 'richland' was shown as the codename for low end trinity a4 apus. >_> that was before multiple speculations (no 'leaks') of richland having pd cores with gcn igpu.... which settled on rl being trinity 1.05.

 

[hcl123]

No Richland is not GCN is VLIW4

C'mon guys if the link were AT, it would be dissected by now... is that so hard ?.. do they use some kind of subliminal mind programming or what ? ... and then complain about biasing !?

http://www.notebookcheck.net/AMD-announces-new-A-Series-APU-codenamed-Richland.89691.0.html

 

[8350rocks]

http://www.computerbase.de/news/2013-05/amds-kaveri-apu-bekommt-einen-neuen-sockel/
looks like AMD roadmap still shows kaveri for 2013

heh, seems like xbitlabs link that i read and posted 'creatively modified' the leaked slides. others posted uncropped images. if i am to believe that, then kaveri is on track for 2013 release while steamroller opterons are gone (i.e. zero number of sr cpu). +NaCl

@richland: i can remember as far back as early trinity-rumors and roadmaps where 'richland' was shown as the codename for low end trinity a4 apus. >_> that was before multiple speculations (no 'leaks') of richland having pd cores with gcn igpu.... which settled on rl being trinity 1.05.

Opterons were pushed back in favor of getting the Opteron-X project on the ground as quickly as possible is what occurred as near as I can tell from the information coming out of AMD these days...(i.e. not much). Though, bright side of news stated that they suspected that opterons based on steamroller were not going to be this year when the slide was first leaked. The presentation itself was about opterons, but their take on it was that the technology of steamroller cores would be out in 2013, not necessarily opterons with them. Additionally, they commented that Kaveri may have some server use, and speculated that may be the other part of it that could be coming into play. SoC Opteron SKUs.

 

[juanrga]

No.
because the first move(i.e. dropping cinebench and intel-biased benchmarks, especially synthetic ones) has to be from amd. and it has to be public and with irrefutable proof.

AMD, VIA, and Nvidia already abandoned BAPCO because of Sysmark bias. Cinebench is propietary.

Again, that doesn't solve the root issue, since internet is free speech, it only serves to polarize the discussions. AnandT is biased (many know), but on discussions pum! charts and numbers of AT are shoved in everybody faces, and ppl waste an enormous amount of time discussing it no matter what, meaning they use "willingful eager users" to do their fight for them, and you can't convince those otherwise so easily ; " AT bad ? you are trolling, and an AMD fanboy"... " no you are blind and an intel fanboy" ... " "no you are this... and this and that sucks" ... and worst and worst...

See what i mean ? ... leads to nowhere but very aggressive posts.

Politely exposed that those are not prove of evidence of any argument, don't click on them... and its not only AT, its the all review "business" that with current methodologies and instruments has a tremendous lack of meaning.

SOLUTION

Attack the message not the messengers...

if is *HARDWARE* that is to be measured than software must not change, or you'll have no reference points .

Which is *exactly* what happens now, benchmark software have no guaranties of anything not even origin, and since is "software" and prone to bugs and errors, you don't even know if the build x last month is the same of the build y this month(how many times did a bench failed on some test machine ? is it hardware fault or software ?), that is, it can change a lot to, voiding the all deal of consistency and meaning.

Ask for md5 signatures, at least you would know if there are any changes in those binaries.

That is THE ONLY THING that users can do, fight for immutable benchmark software, anything else you can't force no one to nothing (its not legal even).

Precisely, my main argument was not to trust review sites that use biased benchmarks.

 

[8350rocks]

PARSEC is a really good alternative...it uses GCC and it's really under utilized outside of some private sector circles. The public would do well to learn more about it.

 

[de5_Roy]

AMD, VIA, and Nvidia already abandoned BAPCO because of Sysmark bias. Cinebench is propietary.

Spoiler

Again, that doesn't solve the root issue, since internet is free speech, it only serves to polarize the discussions. AnandT is biased (many know), but on discussions pum! charts and numbers of AT are shoved in everybody faces, and ppl waste an enormous amount of time discussing it no matter what, meaning they use "willingful eager users" to do their fight for them, and you can't convince those otherwise so easily ; " AT bad ? you are trolling, and an AMD fanboy"... " no you are blind and an intel fanboy" ... " "no you are this... and this and that sucks" ... and worst and worst...

See what i mean ? ... leads to nowhere but very aggressive posts.

Politely exposed that those are not prove of evidence of any argument, don't click on them... and its not only AT, its the all review "business" that with current methodologies and instruments has a tremendous lack of meaning.

SOLUTION

Attack the message not the messengers...

if is *HARDWARE* that is to be measured than software must not change, or you'll have no reference points .

Which is *exactly* what happens now, benchmark software have no guaranties of anything not even origin, and since is "software" and prone to bugs and errors, you don't even know if the build x last month is the same of the build y this month(how many times did a bench failed on some test machine ? is it hardware fault or software ?), that is, it can change a lot to, voiding the all deal of consistency and meaning.

Ask for md5 signatures, at least you would know if there are any changes in those binaries.

That is THE ONLY THING that users can do, fight for immutable benchmark software, anything else you can't force no one to nothing (its not legal even).

Precisely, my main argument was not to trust review sites that use biased benchmarks.

see, it's not that hard to do. what i was proposing was that amd should do something similar with cinebench. so what if it's proprietary? that's not the issue here. amd should tell people how cinebench skews bench results (tell in a discreet way), not ask maxon to change code or compiling process or anything else.

 

[mayankleoboy1]

but its easier to play the silent victim..

 

[Ranth]

Off topic:
For all those who wish they could overclock the haswell i3's with bclk... (if intel do the same with th i3 as the i5-i7 that is...) :

"That’d all be well and good if Intel was enabling this new level of flexibility for the folks who don’t spend extra on K-series parts, giving them the ability to pick higher BCLK settings without access to the ratio multiplier. However, the company instead chooses to restrict the ratios to the Core i7-4770K and i5-4670K—the same ones you can already overclock in 100 MHz increments. Anyone buying one of the 11 other SKUs in Intel’s new Core i7 and i5 line-up is out of luck."

http://www.tomshardware.com/reviews/core-i7-4770k-haswell-review,3521-10.html