[GTALUG] processor progress sure has slowed down

Lennart Sorensen lsorense at csclub.uwaterloo.ca
Mon Jan 9 10:53:05 EST 2017


On Fri, Jan 06, 2017 at 08:31:57PM -0500, D. Hugh Redelmeier via talk wrote:
> I got an ad from NCIX for systems with Intel's new Kaby Lake processors.  
> Here's the headline:
> 
> 	UP TO
> 	[LARGER FONT] 28% BETTER PERFORMANCE* [/LARGER FONT]
> 	versus a 3-year-old-desktop
> 
> The footnote shows that a i7-7700K processor system was being compared 
> with a i7-4770K.  Some of the improvement would be in DDR4 vs DDR3 and so 
> on.  OS and video card were the same.  The power requirement of the CPU 
> went up slightly.
> 
> You'd normally expect a 7xxx processor to be compared with a 6xxx 
> procecessor, but they've reached back an extra two generations to show 
> even a modest 28% improvement.

Well at the same clock the 7xxx and 6xxx are apparently the same speed
in almost all cases.  7xxx is a bit more power efficient though, and
can hence overclock a bit more.

But yes, not a big leap this time at all.  There is a new chipset
generation to go with it, although apparently some of the existing 100
series boards can already support the new chip at least with a bios update.

> There are useful features added to the processors since the Haswell 
> generation.  Like a working TSX-NI: Haswell had it, with bugs, so Intel 
> suppressed it with a microcode update.  Not completely fair since the K 
> models never had it.
> 
> Intel has put a lot of extra transistors into on-chip video, something 
> useful at least sometimes.  Perhaps this shows where Intel feels it needs 
> to compete -- AMD has a better on-chip GPU but an uncompetitive CPU.  One 
> can only hope that AMD's Zen/Ryzen lives up to AMD's talk.

Well at least on some models they are improving graphics, but I am not
sure that is the case on all models.

I still have a happy Core2 Q6600 running, which was released 10 years
ago, and I would not consider it a slow machine.  Sure it doesn't match
my i7-3960X machine, but that would not be a fair comparison anyhow.

-- 
Len Sorensen


More information about the talk mailing list