[AMD] AMD Financial Analyst Day 2015 - Page 27 - Overclock.net - An Overclocking Community

Forum Jump: 

[AMD] AMD Financial Analyst Day 2015

Reply
 
Thread Tools
post #261 of 283 (permalink) Old 05-08-2015, 06:03 AM
Bite my shiny metal GPU!
 
svenge's Avatar
 
Join Date: Dec 2012
Location: USA
Posts: 4,037
Rep: 406 (Unique: 283)
Quote:
Originally Posted by The Stilt View Post

Unless Bulldozer was designed by complete morons the design was definitely targeted for extremely high frequencies, most likely past 6GHz.

That's a true 50/50 if I ever saw one...

svenge is offline  
Sponsored Links
Advertisement
 
post #262 of 283 (permalink) Old 05-08-2015, 06:21 AM
New to Overclock.net
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,529
Rep: 853 (Unique: 325)
Quote:
Originally Posted by svenge View Post

That's a true 50/50 if I ever saw one...

biggrin.gif
The Stilt is offline  
post #263 of 283 (permalink) Old 05-08-2015, 04:56 PM
New to Overclock.net
 
EniGma1987's Avatar
 
Join Date: Sep 2011
Posts: 6,307
Rep: 338 (Unique: 248)
Quote:
Originally Posted by Serios View Post

Also Bulldozer was not designed to be a speed demon just to reach higher clocks that Stars

Yes it was... AMD targeted base speeds at launch of Bulldozer for 5GHz and moving up from there. It was entirely designed to be a speed demon design. Of course, it failed completely for various reasons including process node and launched far slower than the target of 5GHz launch speeds. AMD designed the pipelines to be "deep" so as to reach higher clock speeds. Go look it up, AMD did say all this prior to BD launch

EniGma1987 is offline  
Sponsored Links
Advertisement
 
post #264 of 283 (permalink) Old 05-08-2015, 11:44 PM
New to Overclock.net
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,529
Rep: 853 (Unique: 325)
More I look into the chosen process the more worried I get mad.gif

BEML comparison between the previously used and current processes:
Code:
GlobalFoundries 32nm SHP PD-SOI = 11 (STR, BD, PD)
GlobalFoundries 28nm HPP = 9 - 15 (XV - SR)
GlobalFoundries (SEC) 14nm LPP FinFET = 11
Intel 14nm Gen. 2 FinFET = 13
Intel 22nm Gen. 1 TG HKMG FinFET = 9
TSMC 28nm HP = 12
TSMC 28nm HPL = 11
ST 28nm FD-SOI = 10

Intel said they had to increase the BEML count from 9 to 13 when moving from 22nm to 14nm in order to tackle the process issues.
The Samsung 14nm LPP has up to 11 layers so I wonder how it manages with a design with: roughly five times bigger in size, 40-60x higher power consumption and (hopefully) with twice as high operating frequency.

Also something good to read http://www.soitec.com/pdf/WP_handel-jones.pdf

Never thought I said it, but I am starting to agree with Hector (HR) in the matter of having own fabs ph34r-smiley.gif
The Stilt is offline  
post #265 of 283 (permalink) Old 05-09-2015, 01:20 AM
New to Overclock.net
 
Ultracarpet's Avatar
 
Join Date: Feb 2008
Posts: 3,642
Rep: 216 (Unique: 139)
Quote:
Originally Posted by The Stilt View Post

More I look into the chosen process the more worried I get mad.gif

BEML comparison between the previously used and current processes:
Code:
GlobalFoundries 32nm SHP PD-SOI = 11 (STR, BD, PD)
GlobalFoundries 28nm HPP = 9 - 15 (XV - SR)
GlobalFoundries (SEC) 14nm LPP FinFET = 11
Intel 14nm Gen. 2 FinFET = 13
Intel 22nm Gen. 1 TG HKMG FinFET = 9
TSMC 28nm HP = 12
TSMC 28nm HPL = 11
ST 28nm FD-SOI = 10

Intel said they had to increase the BEML count from 9 to 13 when moving from 22nm to 14nm in order to tackle the process issues.
The Samsung 14nm LPP has up to 11 layers so I wonder how it manages with a design with: roughly five times bigger in size, 40-60x higher power consumption and (hopefully) with twice as high operating frequency.

Also something good to read http://www.soitec.com/pdf/WP_handel-jones.pdf

Never thought I said it, but I am starting to agree with Hector (HR) in the matter of having own fabs ph34r-smiley.gif

I enjoy you being here. wubsmiley.gif

Ultracarpet is offline  
post #266 of 283 (permalink) Old 05-09-2015, 04:37 AM
M'aiq the Liar
 
BoredErica's Avatar
 
Join Date: Jul 2012
Location: Bay Area, California, USA
Posts: 7,401
Rep: 417 (Unique: 271)
Quote:
Originally Posted by The Stilt View Post

More I look into the chosen process the more worried I get mad.gif

BEML comparison between the previously used and current processes:
Code:
GlobalFoundries 32nm SHP PD-SOI = 11 (STR, BD, PD)
GlobalFoundries 28nm HPP = 9 - 15 (XV - SR)
GlobalFoundries (SEC) 14nm LPP FinFET = 11
Intel 14nm Gen. 2 FinFET = 13
Intel 22nm Gen. 1 TG HKMG FinFET = 9
TSMC 28nm HP = 12
TSMC 28nm HPL = 11
ST 28nm FD-SOI = 10

Intel said they had to increase the BEML count from 9 to 13 when moving from 22nm to 14nm in order to tackle the process issues.
The Samsung 14nm LPP has up to 11 layers so I wonder how it manages with a design with: roughly five times bigger in size, 40-60x higher power consumption and (hopefully) with twice as high operating frequency.

Also something good to read http://www.soitec.com/pdf/WP_handel-jones.pdf

Never thought I said it, but I am starting to agree with Hector (HR) in the matter of having own fabs ph34r-smiley.gif

Please...translate...into...English. frown.gif

BoredErica is offline  
post #267 of 283 (permalink) Old 05-09-2015, 05:21 AM
Linux Lobbyist
 
Hattifnatten's Avatar
 
Join Date: Mar 2012
Location: Bergen, Norway
Posts: 1,641
Rep: 104 (Unique: 92)
Quote:
Originally Posted by The Stilt View Post

More I look into the chosen process the more worried I get mad.gif

BEML comparison between the previously used and current processes:
Code:
GlobalFoundries 32nm SHP PD-SOI = 11 (STR, BD, PD)
GlobalFoundries 28nm HPP = 9 - 15 (XV - SR)
GlobalFoundries (SEC) 14nm LPP FinFET = 11
Intel 14nm Gen. 2 FinFET = 13
Intel 22nm Gen. 1 TG HKMG FinFET = 9
TSMC 28nm HP = 12
TSMC 28nm HPL = 11
ST 28nm FD-SOI = 10

Intel said they had to increase the BEML count from 9 to 13 when moving from 22nm to 14nm in order to tackle the process issues.
The Samsung 14nm LPP has up to 11 layers so I wonder how it manages with a design with: roughly five times bigger in size, 40-60x higher power consumption and (hopefully) with twice as high operating frequency.

Also something good to read http://www.soitec.com/pdf/WP_handel-jones.pdf

Never thought I said it, but I am starting to agree with Hector (HR) in the matter of having own fabs ph34r-smiley.gif

Ummm, I'm pretty sure Kaveri was 28nm SHP, not HPP. Much higher density, better efficency, at the cost of frequency. From the launch-reviews:
Quote:
Kaveri moves from Global Foundries' 32nm High-K Metal Gate SOI process to its bulk 28nm SHP (Super High Performance) process. The process node shift actually explains a lot about Kaveri's targeting. While the 32nm SOI process was optimized for CPU designs at high frequency, GF's bulk 28nm SHP process is more optimized for density with a frequency tradeoff.

OCN quotes (Click to show)
Quote:
Originally Posted by Cyclops go_quote.gif

Let's sacrifice a goat!!!

Quote:
Originally Posted by Sean Webster go_quote.gif

Not using an SSD is a computer building sin

Quote:
Originally Posted by CynicalUnicorn go_quote.gif

Didn't you know? G-Sync is powered by the souls of dead babies! rolleyes.gif

Quote:
Originally Posted by CynicalUnicorn go_quote.gif

Quote:
Originally Posted by CasualCat go_quote.gif

Why does it translate Radeon to Geforce?
SweClockers -> Sweden -> Scandinavia -> Finland -> Alatar. It all makes sense! eek.gif


Hattifnatten is offline  
post #268 of 283 (permalink) Old 05-09-2015, 05:35 AM
New to Overclock.net
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,529
Rep: 853 (Unique: 325)
Quote:
Originally Posted by Hattifnatten View Post

Ummm, I'm pretty sure Kaveri was 28nm SHP, not HPP. Much higher density, better efficency, at the cost of frequency. From the launch-reviews:

Unless AMD has their own production lines for the exclusive process variant I have to assume the process is in fact HPP and not SHP.
GlobalFoundries only offers two different variants; Super Low Power (SLP) and High Performance Plus (HPP).

http://www.globalfoundries.com/technology-solutions/leading-edge-technology/28-slp

http://www.globalfoundries.com/technology-solutions/leading-edge-technology/28-hpp

Only AMD has mentioned 28nm SHP process variant lately (during last few years).
The last time it was mentioned in the GlobalFoundries official documentation was in 2009 I believe.
The 28nm SHP was supposed to be a FD-SOI process (according to the old plans).

Unless some clear evidence on the 28nm SHP variant existence emerges I would call it as a mistake or misinterpretion by AMD marketing.
Until Carrizo AMD had only one design made on the alledged 28nm SHP process. I don´t think the volumes of it would justify the exclusive or customized manufacturing process version.
The Stilt is offline  
post #269 of 283 (permalink) Old 05-09-2015, 04:38 PM
New to Overclock.net
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,529
Rep: 853 (Unique: 325)
Quote:
Originally Posted by Darkwizzie View Post

Please...translate...into...English. frown.gif

Here you can find quite simplified explanation: http://www.design-reuse.com/articles/31011/power-grid-design.html

Basically the smaller the linewidth is the higher the BEML count should ideally be in order to avoid the additional potential and likely issues. Usually less complex, lower power consumption and frequency targeted designs or processes have lower BEML count at the same linewidth.

The process specific layer numbers I posted earlier are either the number of layers used (in previous designs) or the maximum number of layers supported by the process (unreleased products).

AMD drastically reduced the number of BEML in Carrizo (from 15 in Kaveri to 9 in Carrizo).
Thats because Carrizo is much more efficiency oriented and less frequency targeted design than Kaveri.
Reducing the number of BEML will hurt in the maximum frequency capabilities but also reduce the manufacturing cost significantly, which appears to be the main reason for the change.
The Stilt is offline  
post #270 of 283 (permalink) Old 05-09-2015, 05:06 PM
Linux Lobbyist
 
Nnimrod's Avatar
 
Join Date: Jul 2011
Posts: 1,626
Rep: 52 (Unique: 48)
So the tldr of all this is no 8ghz from Zen? Do you think it will scale to -190?

wheee.gif Kpop! Yay! wheee.gif
"Lock horns, I push, and I strive. Somehow I feel more alive"


Nnimrod is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off