>This is a key milestone in the process and means that the brunt of the work required for the Ice Lake processors has been done. It also implies that the first generation 10nm products are much further down the assembly line and on-track for their 2017 release. The company still holds the industry advantage in terms of process that they have always enjoyed (assuming you don’t fall for the marketing names that other foundries employ which are not indicative of the physical process) and appears to be confident in its promised delivery of 10nm products.
Ringbus has scaling issues due to the size of the bus, number of buses and all the cores are on one die. Doesn't matter how small you make them if they're still all packed in one place on a monolithic die with shit yields. AMD figured out a better way.
Nolan Lopez
No The advantages of smaller nodes are diminishing.
Jason Cooper
GloFo's 7nm FinFET at the end of this year or early next year should be similar to 10-12nm by Intel's standards. So really Intel isn't that far ahead anymore. Also, according to Intel's own data, 10nm will perform worse on all fronts than 14nm++
Gavin Cox
b-but Intel said they were the best ..
Aiden Watson
They were, but they got lazy.
Justin Roberts
>10nm will perform worse on all fronts than 14nm++ whats the point then?
Kevin Davis
More chips per wafer.
Christopher Thompson
>whats the point then? memes
Isaiah Long
higher yield rate, cheaper to make but Jewtel still wont lower price of course :^)
Ethan Turner
Smaller dies, higher yields, cheaper price, and after some maturing 10nm++ should improve on 14nm++
Christian Taylor
>using the smiley with a carat nose
Owen Rogers
>higher yields so... this seems like an inefficient strategy compared to AMD's in the long run
Evan Murphy
If memes aren't real, how can 10nm be real?
Nathaniel Butler
I'm really not sure how AMD is getting 80% yields on their Ryzen dies.
Robert Murphy
>TSMC I want to see the same graph comparing it with GloFo instead.
Ethan Sanders
If they're taping out 10nm right now, they absolutely are not getting their new processor out in 2017.
This is pure marketing for inteldrones too stupid to question the bullshit Intel is spewing.
Expect 10nm Intel CPU's at least 15 months after tapeout and more like 18 months. Which means it should be mid 2018 to early 2019 at best.
Worse yet, this is still Intel's "we don't give a fuck just add +5% performance from last gen" level of architectural development. For them to actually become competitive again they need to actually add some serious improvements which is going to take years to design.
I don't expect Intel to be competitive again until 2020-21
Jordan Moore
I mean, even if intel gets better yields on 10nm, the problem is their architecture requires all cores to be good on their high core count chips. Whereas AMD will just stitch together octocores with infinity fabric and have great yields. IF/when AMD goes to a smaller process, they'll just have even more good octocores to stitch together.
Nathaniel Butler
I'm talking about the individual zeppelins or whatever they're called.
Cameron Richardson
>taping out Made a mistake, it's taping in. Or maybe one is taping out and one is taping in? I don't know.
Zachary Smith
And the brilliance on top of this is we're now entering a point in time when people actually want higher core counts. So not only can AMD produce higher core count CPU's for a fraction of the price that Intel can, but they've also pushed the market towards higher core counts on top of it.
In two years the standard CPU is going to be either a 6-core or an 8-core. People with 4-cores are going to be looked at like people with single or dual-core today. This completely destroys intel's minor singlethread performance advantage. The ass-raping that zen is giving Intel is unreal.
AMD has overnight become the defacto market leader. It's just not apparent to everyone yet.
Fuck I really should buy some of their stock.
Joseph Thompson
>in b4 this
Elijah Thompson
You should probably add labels to indicate this is trying to show single and multithread performance at the same time
Aaron Sanchez
...
Colton Roberts
>tfw other synthetic benches are so shit, Cinebench, which was designed solely for testing video editing and visual effects design capabilities, is considered the gold standard
How long before Intel buys out Cinebench and the metrics are completely changed?
Nolan Martinez
It'd be pretty hard for them to fuck with benchmarks based on real workloads.
Elijah Collins
Intel might pay Adobe to use the gimped algos if they detect AMD CPU.
Angel Price
Adobe's garbage code is pretty fucked even without Intel's intervention.
Matthew Diaz
>delid kit sold separately
Zachary Turner
meanwhile amd is streaming down on 7nm with zen2.... oh man i just wanna see the tdp of the zen2 compared to 10nm ringbus
Austin Bell
>10nm enjoy no OC and paying $2k for K edition
Brayden Bennett
Intel used to recommend cinnebench to benchmark performance, obviously doesn't anymore since AMD's FPUs (non SIMD) are notably more powerful
Kayden Kelly
How long until Intel will start selling their own delid kits?