Intel's chief chipman: '22nm better than expected, 14nm on track' The low-voltage performance of Intel's 22-nanometer chip-baking process turned out better that the company had predicted, and the development of next year's 14-nanometer process technology is proceeding swimmingly, thankyouverymuch. So said Chipzilla's head of process technology, Mark Bohr, speaking at a tech session during …
22nm was so easy it resulted in Ivy Bridge being 3 months late, running hot and OC'ing poorly with a net gain of ~5% in performance over Sandy Bridge. While each die shrink yields less and less advantage, Intel simply rushed Ivy Bridge out (late) in hopes that naive consumers would not know or care about the technical issues in the hot running chippies. If Intel's node change from 32nm to 22nm and trigate can only deliver an over-heating 5% gain CPU, they are in trouble. 14nm ain't gonna compensate for the other technical shortcomings.
This post has been deleted by its author
or maybe since intel is not living up to the performance gains YOU demand for every die shrinkage they pioneer, perhaps YOU can come up with the solution at home? Personally I am not aware of a terrible core i7 (4 years in the making) But please, enlighten me, I'm awaiting your blueprints. While you are at it, plot me the course for the next interstellar science laboratory satellite.
It's been well demonstrated it was the poor choice in thermal compound under the heat spreader that made Ivy Bridge "run hot." Although, I must say that my IB i7 OCed to a modest 4.4Ghz has a 53.3*C (128*F) package temp, 37.2*C (99*F) core temp (according to AIDA64). On air (and no, not the turbine/jet engine kind). The cores don't generate much heat comparatively, it just gets bottle-necked getting through to the heat-spreader.
Oh, and 5% is actually closer to "3 to 15%" depending on your application.
>It's been well demonstrated it was the poor choice in thermal compound under the heat spreader that made Ivy Bridge "run hot."
That's my understanding of what happened, too. Recently built a passively-cooled, completely silent PC around a i7 3770 S. The 1 Kilogram cooler looks like it fell out of the hood of a 1970s American muscle car, but has kept the chip around 45º.
Because adding cores is only useful if you are running multi-threaded applications (many that you would think to use for casually judging performance aren't) or seriously multi-tasking.
Beyond 2 cores is really only useful to a group that while large in absolutely terms, is relatively quite small.
It's just the continuation of the GHz wars at some level.
As the number of cores per socket climb above 16, there are a few issues:
1) These things get more-and-more like SPARC
2) Revenues for core-count software climbs (SQL, etc.)
3) Customers get pushed to enterprise-level Windows, SQL, etc
4) Cache per core drops
If you are not-yet getting above your MPL (multiprogramming max level), then this is great. If not, then you need to be wary of the above four issues with high-core sockets.
INTC need to get their cache-per-core numbers up for databases OR they need to get their single-path GHz up. IBM is staking out the higher-profit markets with POWER7+ with both of these technologies.
Intel's own data shows IB not selling all that well and most hardware review sights are recommending going with Sandy Bridge except for ULV laptop use. IB can hardly be viewed as a step forward in performance other than in lower power consumption. Intel simply did not get the gains they had hoped for with the move from 32nm to 22nm. It's unlikely the drop to 14nm will offer much either so they had better try to improve their architecture a lot more.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
