InvestorsHub Logo
icon url

This Causes an Error

02/22/16 9:39 PM

#144245 RE: mas #144244

As for the foundry 14/16FF competition they just can't go where Intel can go in pricing and the Bay Trail experience proved that Intel will pick up major mobile business if priced ultra-competitively. The trick now is to do it without subsidies which is where Intel's dense 14nm will come in



I think you misunderstand what the subsides were for. The problem on the cost side was that the motherboards themselves, not really the SoC, were just too expensive. Remember that these SoC vendors don't just sell chips, but they do the motherboard design for pretty much all but the most premium customers.

This is why Intel was able to do a BYT-CR which cut the subsidies in half.

Intel could have had cost-appropriate products at 22nm if its platform engineering product definition teams did their jobs correctly, AFAICT.

As far as foundry 14/16nm competition, don't be so sure of that. There are ARM vendors who successfully used 3rd party foundries to put out quite cheap & capable products. TSMC is also doing a cost-optimized 16nm FFC that should be extremely good for getting FinFETs on the cheap.

Intel at this moment has record PC/Server market-share so where does it go from there except potentially down ? The trick then will be to keep the overall market growing even if m/s falls back. Worst case Zen danger is if it beats the 'Lakes in throughput and especially throughput/power in which case AMD will have a major Server/high-end PC renaissance and Intel's GMs will fall.



I agree, AMD is to be closely watched. Of course, no analyst is going to ask BK about what he thinks about AMD's potential resurgence and even if they bothered, he'd probably say something like "We're confident we have a lead, trust us, 50 year history Moore's Law...something..."
icon url

This Causes an Error

02/22/16 11:31 PM

#144246 RE: mas #144244

To my point about foundry 16nm...

BARCELONA, Spain, Feb. 22, 2016 /PRNewswire/ -- Spreadtrum Communications ("Spreadtrum"), a leading fabless semiconductor provider in China with advanced technology in 2G, 3G and 4G wireless communications standards, today announced its 64-bit Octa-Core LTE SoC platform, SC9860 chipset at mass production stage in 2016 Mobile World Congress (MWC) , Barcelona. As Spreadtrum's single chip solution for medium and premium smart phones, SC9860 possesses excellent computing ability and features premium multi-media applications, which bring about ultimate user experience for global handset consumers.



http://www.prnewswire.com/news-releases/spreadtrum-announces-its-first-16nm-5-mode-octa-core-lte-soc-platform-300222878.html

Even Spreadtrum is able to get a FinFET SoC into production. Intel is done in mobile.
icon url

This Causes an Error

03/16/16 7:33 PM

#144587 RE: mas #144244

mas, apparently some engineer at an OEM got ahold of a Zen ES:

It seems that unlike with Bulldozer, AMD has created separate dies for server and consumer parts. The server version of the die has twice the cores, L3 cache and additional I/O controllers per die. I haven´t been able to disassemble one yet, however judging from the package size it is a MCM part. 14nm LPP process.

The relative power consumption is roughly the same as on Intel 14nm parts with similar configuration, but the clocks are quite low :/



http://semiaccurate.com/forums/showpost.php?p=257682&postcount=2112

and...

For a long time I actually like what I see. I´d say as long as the consumer Zen parts can reach high enough clocks (min. 3.5GHz), everything will be pretty good



http://semiaccurate.com/forums/showpost.php?p=257718&postcount=2120

The threat to INTC that you pointed out may indeed materialize with Zen.

Looks like the rumor of a 32C/64T chip was true. We know the consumer part goes up to 8C/16T, which means that the dies on the Zen server chip are 16C/32T. Two of them stuck on a package means 32C/64T per chip.

SKL-EP/EX will top out at 28 cores/56 threads, though that will be a single monolithic die rather than an MCM.