AMD Carrizo Pre-release thread

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Paul98

Diamond Member
Jan 31, 2010
3,732
199
106
What I am most interested to see is the improvement in the GPU from compression.
 
Aug 11, 2008
10,451
642
126
Yes because the 95W Kaveris are actualy 67W TDP, this is the reason there s not much difference with the 65W parts.

About this i did find it funny that the first reflex of some people was to assume that the 65W parts were of a higher TDP without even thinking that it could be the higher power part that had its official TDP greatly exagerated.

That does not make sense. If everyone is trying to get the best performance per watt, why would AMD deliberately exaggerate the TDP of Kaveri?
 

Abwx

Lifer
Apr 2, 2011
11,751
4,674
136
That does not make sense. If everyone is trying to get the best performance per watt, why would AMD deliberately exaggerate the TDP of Kaveri?

All chips that do not pass the 65W rating at projected frequencies are relegated as 95W parts even if failing the binning by only a few watts, the 95W moniker is convenient as it allow usage of existing 95W cooling solutions, this is what they have also done with their AM1 APUs.

The measured power drain CPU wise of a 7850K at stock is 54W with Fritzchess and 67W with Prime 95, the latter number is an absolute maximum as there s no heavy soft that manage to get Prime 95 power numbers, stressing both the CPU and GPU with Luxmark get you 55W, that s a long shot from the official TDP, these are measurements made by Hardware.fr on early retail SKUs, they selected the less performing one from the two they bought.

What is interesting is that they pointed that voltage was 12.6% higher than what was necessary for perfect stability and that lowering it by this number reduced TDP to 50W , 95W was reached at about 4.4GHz, these numbers are with Prime 95.
 

Lepton87

Platinum Member
Jul 28, 2009
2,544
9
81
All chips that do not pass the 65W rating at projected frequencies are relegated as 95W parts even if failing the binning by only a few watts, the 95W moniker is convenient as it allow usage of existing 95W cooling solutions, this is what they have also done with their AM1 APUs.

The measured power drain CPU wise of a 7850K at stock is 54W with Fritzchess and 67W with Prime 95, the latter number is an absolute maximum as there s no heavy soft that manage to get Prime 95 power numbers, stressing both the CPU and GPU with Luxmark get you 55W, that s a long shot from the official TDP, these are measurements made by Hardware.fr on early retail SKUs, they selected the less performing one from the two they bought.

What is interesting is that they pointed that voltage was 12.6% higher than what was necessary for perfect stability and that lowering it by this number reduced TDP to 50W , 95W was reached at about 4.4GHz, these numbers are with Prime 95.

It's been my experience that AMD almost always give significantly more volts than its needed for its CPU probably with the exception of FX9xxx series as that CPU are pushed to the max.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,801
1,284
136
It's been my experience that AMD almost always give significantly more volts than its needed for its CPU probably with the exception of FX9xxx series as that CPU are pushed to the max.
http://images.anandtech.com/doci/7974/Screen-Shot-2014-04-29-at-1.08.08-AM.jpg

The Integrated Voltage Regulation and Per-part Adaptive Voltage, should get rid of overmargining for the voltage.

This explains the thermal and power calculator that are in some Carrizo documents. It's for the AVS IP that is connected with the IVR IP.
 
Last edited:

Abwx

Lifer
Apr 2, 2011
11,751
4,674
136
It's been my experience that AMD almost always give significantly more volts than its needed for its CPU probably with the exception of FX9xxx series as that CPU are pushed to the max.

Indeed, no chip of their current lines, except the ones you mentioned, does make exception, even the AM1 are overvolted, this is due to stability being the number one concern in electronic design, you ll never see this criterion being traded for any other caracteristic, the higher resulting TDP is somewhat beneficial for the manufacturers since higher voltage will increase yields.
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,998
441
126
Excavator in 2011-2012 @ Sunnyvale teams put it on 20-nm LPM.
Excavator in 2012-2013 @ Boston teams put it on 28-nm SLP. (LP node shared with Samsung; SLP = LPP or ET-SOI = FD-SOI)((It is required to be shared with Samsung for AMD to use it))

~2012 -> Steamroller and Excavator was moved from Sunnyvale teams to Boston teams. So, Sunnyvale could design from scratch a new high performance x86 core. Like, they did in 2005/2006 to 2010 called Bulldozer.

Do you mean the new AMD x86 uArch named Zen? So that will be on 20 nm, and not 14 nm as some have speculated?

AMD Zen is expected to arrive in 2016, so Samsung/GF 14 nm should be available by then.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,801
1,284
136
Do you mean the new AMD x86 uArch named Zen? So that will be on 20 nm, and not 14 nm as some have speculated?

AMD Zen is expected to arrive in 2016, so Samsung/GF 14 nm should be available by then.
Zen will be on the 10-nm(48-nm M1 pitch) node, K12 will be on the 14-nm(64-nm M1 Pitch) node. There is one more Excavator called Excavator+ which will be the 14-nm(64-nm M1 Pitch) x86.

Steamroller and Excavator -> Boston
Sunnyvale places full attention on -> K12 and Zen
 
Last edited:

inf64

Diamond Member
Mar 11, 2011
3,884
4,691
136
Zen will be on the 10-nm(48-nm M1 pitch) node, K12 will be on the 14-nm(64-nm M1 Pitch) node. There is one more Excavator called Excavator+ which will be the 14-nm(64-nm M1 Pitch) x86.

Steamroller and Excavator -> Boston
Sunnyvale places full attention on -> K12 and Zen
Where did you get this information from? Doesn't seem accurate at all.
 

NTMBK

Lifer
Nov 14, 2011
10,398
5,630
136
Zen will be on the 10-nm(48-nm M1 pitch) node, K12 will be on the 14-nm(64-nm M1 Pitch) node. There is one more Excavator called Excavator+ which will be the 14-nm(64-nm M1 Pitch) x86.

Steamroller and Excavator -> Boston
Sunnyvale places full attention on -> K12 and Zen

Zen is coming out in 2016. The only fab anywhere near having 10nm ready for 2016 is Intel, and they ain't gonna fab for AMD. It isn't going to be 10nm.
 

Abwx

Lifer
Apr 2, 2011
11,751
4,674
136
Zen is coming out in 2016. The only fab anywhere near having 10nm ready for 2016 is Intel, and they ain't gonna fab for AMD. It isn't going to be 10nm.

Their 14/16nm wasnt ready this summer with core M being given a new stepping for that exact reason, i dont see how their 10nm would be ready in 18 months, unless you take Seronx s post at face value, of course.
 

NTMBK

Lifer
Nov 14, 2011
10,398
5,630
136
Their 14/16nm wasnt ready this summer with core M being given a new stepping for that exact reason, i dont see how their 10nm would be ready in 18 months, unless you take Seronx s post at face value, of course.

This is why I said 'anywhere near'. Also the end of 2016 is 25 months away, not 18. Anyway, the point is that GloFo, Samsung and TSMC just aren't going to have 10nm in 2016.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,801
1,284
136
Yeah, and 3nm by 2018.
You have the oddest fantasies mrmt.

7-nm -> 2017
5-nm -> 2020
4-nm -> 2022
3-nm -> 2024
2-nm -> 2026
1-nm -> 2028

Is the current updated roadmap.
It is a secret project developed by CEA-LETI, UTBETUBERSOI.
Mrmt, how many times do we have to tell you. FinFETs is what the industry has selected.

10-nm RMG FinFETs or 7-nm RMG Strained FinFETs or go home scrub. Gate capacitance 100 times worse than planar who gives a damn! It can always be designed around. No one in high performance cares for the lower dynamic power efficiency.

FinFETs are where it is at and the industry doesn't care about any negatives. Don't get started on the power density or heat issues, we all heard about it.
 
Last edited:

NostaSeronx

Diamond Member
Sep 18, 2011
3,801
1,284
136
Does the universe implode when we go past 1nm?
Nope. The scientific universe is safe till Planck's constant(length).

After 1nm, it becomes increasingly difficult to find what to use.
---
Also, the Carrizo Eng Sample_9874(15W)

CPU wise is equivalent to the FX-7500(19W/20W).
GPU wise is in between the FX-7600P(35W) and the A10-7850K(95W).
 
Last edited:

Ajay

Lifer
Jan 8, 2001
16,094
8,111
136
Nope. The scientific universe is safe till Planck's constant(length).

After 1nm, it becomes increasingly difficult to find what to use.
---
Also, the Carrizo Eng Sample_9874(15W)

CPU wise is equivalent to the FX-7500(19W/20W).
GPU wise is in between the FX-7600P(35W) and the A10-7850K(95W).

First h doesn't have a spacial only representation: http://en.wikipedia.org/wiki/Planck_constant

Secondly, quantum mechanical effects are already changing the structures and materials needed to create a reliable xtor. As far as I can tell, it seems most likely that quantum mechanical behavior will be the guiding principle in semicon design by the end of the decade (over classical E&M). Having to design xtors using QM rules with some sub nano-meter scale elements in an xtor is what becomes and stays hard.

Feynman said that there was plenty of room at the bottom. He didn't say it would be easy getting there. But this is going way OT.
 

monstercameron

Diamond Member
Feb 12, 2013
3,818
1
0
First h doesn't have a spacial only representation: http://en.wikipedia.org/wiki/Planck_constant

Secondly, quantum mechanical effects are already changing the structures and materials needed to create a reliable xtor. As far as I can tell, it seems most likely that quantum mechanical behavior will be the guiding principle in semicon design by the end of the decade (over classical E&M). Having to design xtors using QM rules with some sub nano-meter scale elements in an xtor is what becomes and stays hard.

Feynman said that there was plenty of room at the bottom. He didn't say it would be easy getting there. But this is going way OT.

That's cool guys, interesting conversation. hey why not open a new thread to discuss it?
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,801
1,284
136
That's cool guys, interesting conversation. hey why not open a new thread to discuss it?
No, everyone rather take a dump in your thread.

Poll time;
- 28LPS
- 28SLP
- 28LPH
- 28HP
- 28HPP
- 28SHP
- 28FD
http://strawpoll.me/2918385

Poll time #2;
- Scalable Coherent Hypertransport (4HNC) Ring-Mesh Interconnect
- Garlic, Onion, Onion+ Interconnect
http://strawpoll.me/2918387

Poll time #3;
- 0.8 volts to 1.3 volts (DVFS)
- 0.5 volts to 1.4 volts (Wide DVFS)
http://strawpoll.me/2918390

Poll time #4;
- DVFS + AFS - No AVS or IVR
- AVFS + IVR
http://strawpoll.me/2918394
 
Last edited:
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |