InvestorsHub Logo

This Causes an Error

03/03/17 11:10 AM

#147958 RE: morrowinder #147957

They won't be able to clock as high as Intel's 4 core parts and their supposed desktop TDP advantage will not be as apparent(I would predict its worse than Intel by a fair amount) in lower voltage parts.



TDP advantage is primarily a marketing advantage.



90.3W for 6900K vs 117W for 1800X.



Ryzen is 9% faster than the 6900K in this bench, but it consumes 30% more power to get that edge (as an aside: 6950X consumes less power than 6900K does and outperforms the 1800X in this workload).

Perf/watt, not the perf/TDP, seems to be in favor of 6900K over 1800X. That's old Broadwell core, 1st generation Intel 14nm.


Also, here is the V/F curve for Ryzen from Anand forums:




Requires 1.4V to get 4GHz. Intel's old 22nm 5960X could hit 4GHz on all cores with just 1.075v. ASUS' overclocking guide said that 1.3V for 4.5GHz was an "average" result for old 22nm 5960X (https://rog.asus.com/articles/overclocking/rog-overclocking-guide-core-for-5960x-5930k-5820k/).

Ryzen mobile will have to go up against Kaby Lake Refresh-U and Coffee Lake, which should be built on another iteration of 14nm past 14nm+, which should further increase frequency capability for a given power consumption.

borusa

03/03/17 11:17 AM

#147959 RE: morrowinder #147957

and their supposed desktop TDP advantage will not be as apparent(I would predict its worse than Intel by a fair amount) in lower voltage parts.

So what information do you have to predict this ...? AMD parts run at higher V. and use less power, I don't see anything wrong there.

mas

03/03/17 7:05 PM

#147967 RE: morrowinder #147957

Server customers are also conservative in that they take their time even upgrading the next iteration of their current favored architecture. Where I only see Rizen making an immediate impact is in hpc orientated sites but that is not a huge market. To me it looks like a 2017 version of a chip like the last Mac IBM PowerPC or even Itanium, generally competent and even great in hpc niches but not great all-rounders.

AMD did not go from 2 to 15 on that basis but on the thesis that AMD spread that they had a cooler Core killer which they clearly don't. Intel may have to cut prices to accommodate the stiffer competition but they have room to do so as they have not had any serious x86 competition in a decade and have made serious fat margin hay during that time with scope for cuts which trade margin for more sales. AMD stock will fall further as all those big institutional boys who bought shares/notes at 7 will want to take profit while they can now that retail momentum has turned south.