AMD Piledriver rumours ... and expert conjecture

[Reynod]

We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...

 

[Chad Boga]

http://semiaccurate.com/2012/03/21/ibm-power-7-spotted-and-it-is-a-monster/

intel can't do something like this yet maybe not for more than a year. not to mention the 14nm waffer from both IBM and samsung which intel can probably do but has not publicly shown yet.

That is an irrelevant to desktop, niche product and there is about as much evidence to suggest IBM and Samsung are ahead of Intel on 14nm as there was that they were ahead(or as far along as Intel, which I believe was their claim) on HKMG, when it was at least 3 years AFTER Intel first shipped a HKMG product, that they finally had something to show.

 

[Cazalan]

Intel already demo'd a 14nm Haswell chip running live in a system last year. That's better than just a wafer which could be a dud.

 

[jimmysmitty]

Intel already demo'd a 14nm Haswell chip running live in a system last year. That's better than just a wafer which could be a dud.

It wouldn't be Haswell but this is about the same:

http://www.pclaunches.com/processors/intel-has-a-14nm-chip-in-its-lab.php

Haswell is the new arch on 22nm set for 2013 release. Bridgewell (so far as I know) is 14nm.

Most new processes start in SRAM form then once they have it down, they move to putting a CPU onto it. I would imagine since Intel is already building the 14nm FAB, they have already done SRAM and are able to do a CPU and most of 2013 will be leaks/demos of it running. Probably half way through 2013 we will see someone like Coolaler getting a early demo chip, like he did with Ivy Bridge mid last year, and it will go from there.

 

[Blandge]

This article actually clears a lot of this up.

http://semiaccurate.com/2011/03/31/after-intels-haswell-comes-broadwell-sk/

Haswell is the new arch on 22nm set for 2013 release. Bridgewell (so far as I know) is 14nm.

Either way, we are pretty sure Broadwell is the 16nm (14nm now) Haswell successor.

From there we get to yet another new architecture, the family is called SkyLake

He also goes on to talk about packaging

Haswell is where Intel starts getting serious about die stacking, interposers, and related advanced packaging technologies.

 

[king smp]

oh, you must have forgot about me...?
I know that I only throw darts at my certifications on the wall now, but I did earn them at one time ya know..
I feel insulted.

dont be insulted
it is the whole testing procedure process
what is involved with finding places that proctor,how much they cost and doing the tests
It wasnt the tech questions I needed help with
You know I always bug you with those

 

[jimmysmitty]

This article actually clears a lot of this up.

http://semiaccurate.com/2011/03/31/after-intels-haswell-comes-broadwell-sk/

Either way, we are pretty sure Broadwell is the 16nm (14nm now) Haswell successor.

From there we get to yet another new architecture, the family is called SkyLake

He also goes on to talk about packaging

Haswell is where Intel starts getting serious about die stacking, interposers, and related advanced packaging technologies.

Stacked RAM would be what I see with Haswell or the "L4" will make its first apperance (much like L3 did in the Pentium 4 EE 3.4GHz) but I have heard that L4 will be MCM like, or maybe use "interposers" instead.

I guess we will see as once Ivy hits, we will start hearing all about Haswell.

My bet is we have more info on Haswell before PD is even release. Maybe even before Trinity is released.

 

[Cazalan]

Haswell is the new arch on 22nm set for 2013 release. Bridgewell (so far as I know) is 14nm.

Ah right I got those confused. They were showing a near threshold (0.7v) test chip. I thought they needed 14nm for that but they didn't say specifically. Intel spending 8Bil or so to make a brand new 14nm fab and retool another for 14nm by 2013.

 

[dogman_1234]

Ivy Bridge>Haswell>Boradwell>Skylane>Skoymount>then we hit atoms...lovely, lovely atoms...

 

[truegenius]

^and then what

 

[noob2222]

You mean MCM? Where you have two or more chips on the same package? Intel already did that with Pentium D and Core 2 Quad.

As for 14nm, I didn't read anything about IBM/Samsung building a 14nm FAB that costs $5 billion yet.

As for the tri-gates, I would assume Intel already has something up their sleeve for 14nm. They probably have it already done just need the FAB for mass production.

BTW this is interesting:

http://www.semiwiki.com/forum/content/1112-intel%92s-first-14nm-chip-will-not-x86-processor.html

A non-x86 to be the first 14nm chip and it is said to be sometime in early 2013, even before we will see it with Bridgewell.

Samsungs announcements are not very common and usually not brought up very often, but they are much bigger than most people realize. They aren't building a 5B fab because thats too small.

Samsung, for example, has a $33-billion Korean expansion underway that will create the world's largest semiconductor complex in the cities of Giheung and Hwaseong. Scheduled for completion in 2012, the multi-phase project south of Seoul will add eight production lines and an R&D center. All told, the expansion will cover a whopping 230 acres (92 hectares) and create 14,000 new Samsung jobs.

found in this article http://www.siteselection.com/ssinsider/bbdeal/bd060504.htm covering the 4B for upping the Austin fabrication process that took 1 year to complete and only 5 months to reach full operation. Here was another article I read where as soon as that was done, they are looking to drop another 1B into austin.

http://www.statesman.com/business/samsung-may-spend-1-billion-to-upgrade-austin-2108072.html

 

[bawchicawawa]

^and then what

Then we build computers made up of JUST ELECTRONS.

 

[Cazalan]

^and then what

You start stacking. Houses used to be 1 floor, now we got skyscrapers.

 

[king smp]

Then we build computers made up of JUST ELECTRONS.

miniature fiber optic light chips
instead of electron or electrical pulses
it will be optical based CPUs
they are already working on it
http://en.wikipedia.org/wiki/Optical_computing

 

[king smp]

I want a cold fusion or fission reactor powered liquid N20 cooled quad Xeon with a quad HD 7990 or quad GT 690
with a self destruct button......

 

[king smp]

and oh yeah the nice female voice saying "self destruct sequence intiated-T-10 and counting"

 

[bawchicawawa]

I want a cold fusion or fission reactor powered liquid N20 cooled quad Xeon with a quad HD 7990 or quad GT 690
with a self destruct button......

octo-crossfire and octo-sli? Do wants, too bad all games would probably run like utter crap on that kind of setup, lol.

 

[jimmysmitty]

octo-crossfire and octo-sli? Do wants, too bad all games would probably run like utter crap on that kind of setup, lol.

It would be physically possible with a Asus Rampage IV Extreme:

http://www.newegg.com/Product/Product.aspx?Item=N82E16813131802

it has 5 PCIe x16 slots, 4 of which run at full x16 and one at x8 (the black one). So you could feesibly put 4 HD7990s into it, but I doubt there is driver support to push that setup and you would probably need at least dual 1200W 80+ Platinum PSUs.....

 

[fazers_on_stun]

^and then what

Erm, quarks?? 😗

 

[fazers_on_stun]

Samsungs announcements are not very common and usually not brought up very often, but they are much bigger than most people realize. They aren't building a 5B fab because thats too small.

Samsung, for example, has a $33-billion Korean expansion underway that will create the world's largest semiconductor complex in the cities of Giheung and Hwaseong. Scheduled for completion in 2012, the multi-phase project south of Seoul will add eight production lines and an R&D center. All told, the expansion will cover a whopping 230 acres (92 hectares) and create 14,000 new Samsung jobs.

found in this article http://www.siteselection.com/ssinsider/bbdeal/bd060504.htm covering the 4B for upping the Austin fabrication process that took 1 year to complete and only 5 months to reach full operation. Here was another article I read where as soon as that was done, they are looking to drop another 1B into austin.

http://www.statesman.com/business/samsung-may-spend-1-billion-to-upgrade-austin-2108072.html

Hmm, still doesn't answer the question - is Samsung building a 14nm production fab, or nutz??

 

[fazers_on_stun]

light does have mass, its just taken to be massless because its negligible.

Experimental checks on photon mass

The photon is currently understood to be strictly massless, but this is an experimental question. If the photon is not a strictly massless particle, it would not move at the exact speed of light in vacuum, c. Its speed would be lower and depend on its frequency. Relativity would be unaffected by this; the so-called speed of light, c, would then not be the actual speed at which light moves, but a constant of nature which is the maximum speed that any object could theoretically attain in space-time.[21] Thus, it would still be the speed of space-time ripples (gravitational waves and gravitons), but it would not be the speed of photons.
...
These sharp limits from the non-observation of the effects caused by the galactic vector potential have been shown to be model dependent.[27] If the photon mass is generated via the Higgs mechanism then the upper limit of m≲10−14 eV/c2 from the test of Coulomb's law is valid.

Photons inside superconductors do develop a nonzero effective rest mass; as a result, electromagnetic forces become short-range inside superconductors.
See also: Supernova/Acceleration Probe

http://en.wikipedia.org/wiki/Photon

I'd agree with that since a 'light sail' depends on changing solar photon momentum for thrust. If a photon had zero relativistic mass then how could light from the sun accelerate solar sailships?

 

[fazers_on_stun]

Energy=Mass times the speed of light.

He did not. It is not a strictly mass-less particle,( meaning it can have mass in a vacuum). Even if it was to slow down, it would lose energy not mass.

Even with high propagation of other matter, light can retain mass in it form or in packets called photons. With the presence of other matter, there is still a vacuum. When an atom takes the photon, it energizes the electron and excites it. Basically, the electron took on the photon and added more energy....which requires mass.

Actually Einstein's special theory of relativity says E = MC^2.

IOW, the square of the speed of light.

 

[king smp]

you know if a thread is going to derail
then a discussion of photons,quarks and matter is a pretty good way to go LOL

realistically (sort of)
a dual SB based Xeon rig with a dual HD 7990 or GTX 690 would just have to do
i mean it would be alright for facebook and angry birds anyway

 

[fazers_on_stun]

Then...once we spit quarks....we shall control whatever they are made of.

First, quarks. Tomorrow...The World!

So far no hint of anything inside a quark has been found, despite the best efforts of the LHC and Fermilab.

That's not to say that there isn't interior structure; just that our current machinery is too weak at 7 TEV to find any.

 

[g4114rd0]

A Host of Choices..."
http://blogs.amd.com/work/2012/03/23/a-host-of-choices/

Welcome to OpenCL 1.2..."
http://blogs.amd.com/developer/2012/03/23/welcome-to-opencl%E2%84%A2-1-2-check-out-these-beta-drivers-that-contain-a-complete-opencl%E2%84%A2-1-2-solution/

 

[jimmysmitty]

erm..
Corsair H60 Hydro vs KUHLER H2O = who wins.
going to cool the AM3+ unit..

Antec. The Khuler is great and only the H100 beats it while the H80 is about the same.

As for light, how about we just agree that that will be the next phase at some point. We already know Intel is going for it:

http://techresearch.intel.com/ResearchAreaDetails.aspx?Id=26

And they will also be pushing out their fiber based Thunderbolt this year.

Other than that, light is just light and thats all irrelevant.