Breaking News
News & Analysis

Intel Xeon Boosts Workstation & Servers

22nm 3D FinFETs Accelerate & Modernize Datacenter
9/9/2014 02:00 PM EDT
5 comments
NO RATINGS
Page 1 / 2 Next >
More Related Links
View Comments: Newest First | Oldest First | Threaded View
TarraTarra!
User Rank
Author
Power and frequency not scaling
TarraTarra!   9/10/2014 2:02:16 PM
NO RATINGS
What was interesting here was that the turbo feature has two frequencies - one with AVX instructions and the other without. When running AVX instructions the frequency drops to 1.9GHz. Without AVX instructions it is 2.3GHz. This goes to say how carefully Intel has had to pack these cores into that power budget. In certain applications the previous IvyBridge 10core and 12core parts have higher performances as they can clock higher.

 

This has been underwhelming to say the least.

TarraTarra!
User Rank
Author
Re: 18?
TarraTarra!   9/10/2014 1:54:37 PM
NO RATINGS
@colin is right. Intel went with a ring based interconnect from sandybridge and that makes it easier to come up with odd number of cores. I am sure 18 was as far as they could push the core count without blowng up the socket power budget. The part is rated at 145W but apparently burns close to 165W in reality.

krisi
User Rank
Author
Re: 18?
krisi   9/10/2014 1:21:33 PM
NO RATINGS
thank you Colin

R_Colin_Johnson
User Rank
Author
Re: 18?
R_Colin_Johnson   9/10/2014 1:18:25 PM
NO RATINGS
It seems that 18 was chosen because the layout allowed it. Here is what Intel's chief architect says: "The internal micro-architecture is not constrained to a power-of two. Implementing the layout as four columns (three with four cores and the fourth with six cores) allowed us to efficiently use the available silicon area and provide the maximum number of cores. "

krisi
User Rank
Author
18?
krisi   9/9/2014 2:59:17 PM
NO RATINGS
18 cores? such a strange number...I thought it should be multiple of 2, say 16...any explanation?

Most Recent Comments
michigan0
 
SteveHarris0
 
realjjj
 
SteveHarris0
 
SteveHarris0
 
VicVat
 
Les_Slater
 
SSDWEM
 
witeken
Most Recent Messages
9/25/2016
4:48:30 PM
michigan0 Sang Kim First, 28nm bulk is in volume manufacturing for several years by the major semiconductor companies but not 28nm FDSOI today yet. Why not? Simply because unlike 28nm bulk the LDD(Lightly Doped Drain) to minimize hot carrier generation can't be implemented in 28nm FDSOI. Furthermore, hot carrier reliability becomes worse with scaling, That is the major reason why 28nm FDSOI is not manufacturable today and will not be. Second, how can you suppress the leakage currents from such ultra short 7nm due to the short channel effects? How thin SOI thickness is required to prevent punch-through of un-dopped 7nm FDSOI? Possibly less than 4nm. Depositing such an ultra thin film less then 4nm filum uniformly and reliably over 12" wafers at the manufacturing line is extremely difficult or not even manufacturable. If not manufacturable, the 7nm FDSOI debate is over!Third, what happens when hot carriers are generated near the drain at normal operation of 7nm FDSOI? Electrons go to the positively biased drain with no harm but where the holes to go? The holes can't go to the substrate because of the thin BOX layer. Some holes may become trapped at the BOX layer causing Vt shift. However, the vast majority of holes drift through the the un-dopped SOI channel toward the N+Source,...

Datasheets.com Parts Search

185 million searchable parts
(please enter a part number or hit search to begin)
Like Us on Facebook
EE Times on Twitter
EE Times Twitter Feed