Stag­nat­ing De­vel­op­ment

Have we reached a new Dark Age in tech­no­log­i­cal progress?


AS SOON AS THE FIRST RU­MORS started cir­cu­lat­ing con­cern­ing In­tel’s top-end 9-se­ries chip, hous­ing eight cores and 16 threads, I knew that it was go­ing to have a sol­dered IHS. I re­mem­ber the con­ver­sa­tion I had with Alan at the time: There was sim­ply no way that ar­chi­tec­ture could take an­other two cores in a sin­gle die, with­out im­prov­ing the putty-like TIM ( Ther­mal In­ter­face Ma­te­rial) In­tel was us­ing be­tween the die and the heat spreader. Lo and be­hold, we re­ceive a sam­ple, and that’s ex­actly what the com­pany has done.

Con­sid­er­ing that this TIM and core count up­grade is the only new ad­di­tion to the Can­non Lake se­ries (apart from even more BS when it comes to In­tel’s now leg­en­dar­ily con­vo­luted nam­ing schemes, of course), it’s re­ally dif­fi­cult for me to get ex­cited about it. I’ve driven the point home enough by now, but to re­it­er­ate: Kaby Lake was re­al­is­ti­cally the only op­ti­miza­tion that Sky Lake re­ceived. Cof­fee Lake was sim­ply a slightly higher clocked Kaby Lake with more cores, while Can­non Lake is noth­ing more than Cof­fee Lake with ex­tra cores and a bet­ter TIM.

In­tel’s big­gest road­block right now is get­ting to 10nm. Which is why we’ve seen a four-year de­lay on this die shrink alone. But the fear, for me at least, is what can fol­low 10nm? Man­u­fac­tur­ers are al­ready strug­gling to get to 10, with even Glo Fo pulling out of the race. What’s next for In­tel’s long-term strat­egy? Are we go­ing to be in a sit­u­a­tion where it’s an­other five years of the same ar­chi­tec­ture, with diminu­tive in­creases in per­for­mance, and ex­ces­sive hikes in ther­mal out­put? I cer­tainly hope not….

The only thing I’m vaguely ex­cited about is the pack­ag­ing.

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.