My thoughts are that AMD chose TSMC to do the lower clock server chips as the clocks would be acceptable. That gives AMD & GloFo the freedom to tweak their own 7nm to squeeze out clock speed without the pressure of delivering server chips.

D3M1G0D:

If AMD can start shipping 7nm EPYCs by next year then that'll be a major win. Intel won't have their 10nm stuff ready for servers by at least 2020 (and that's assuming there are no more delays). 48-core EPYC on 7nm vs 28-core Xeon on 14nm = no contest.

It matters not, Intel will still reign supreme as far as gaming performance goes. And that's all that matters to a lot of us.

Irenicus:

It matters not, Intel will still reign supreme as far as gaming performance goes. And that's all that matters to a lot of us.

How do you figure? Do you think Intel's 14nm++ parts will still reign supreme over AMD's 7nm parts? Matisse will arrive in Q2 of next year and Intel will have nothing to challenge it until the holidays (assuming that 10nm isn't delayed again, which is most likely will!). Not only will Rome give Intel a serious run for their money in the server market, but their long-time dominance of the consumer market will also come under threat. Seriously now, give me a realistic scenario of how Intel will counter AMD's upcoming 7nm Zen 2 products?

intel will counter that most likely with an upgraded version of the nice industrial chiller with RGB leds + more fans on it, as well as a new motherboard+socket combo offer

I know what you did here "AMD Zen 2 nm architecture", I read 2nm 😀

D3M1G0D:

How do you figure? Do you think Intel's 14nm++ parts will still reign supreme over AMD's 7nm parts? Matisse will arrive in Q2 of next year and Intel will have nothing to challenge it until the holidays (assuming that 10nm isn't delayed again, which is most likely will!). Not only will Rome give Intel a serious run for their money in the server market, but their long-time dominance of the consumer market will also come under threat. Seriously now, give me a realistic scenario of how Intel will counter AMD's upcoming 7nm Zen 2 products?

D3miG0d is exactly right. the other shoe that needs to be dropped is pricing. it is physically impossible for a 10nm cpu to meet yields and pricing of a 7nm cpu. and while that 10nm is having problems, the 7nm is about to be shipped, leaving the reality of a 14nm cpu vs. a 7nm cpu in the market this year. while it is not a true doubling of yield, it comes close. so TSMC's process produces nearly twice as much at nearly half the cost from the same sized silicon wafers. and the true reality is the vast majority of "gamers" are young and cannot afford much. if anyone thinks that an average gamer will plonk down over 50% more for a less efficient, hotter cpu when there is an equivalent or better for less money i have a bridge to sell them. fanboy-ism is one thing, a complete lapse of reasoning another. just look at AMD's numbers for the last quarter, and Intel's...if you do not count Intel's data center income (over 50%) AMD has grabbed a huge share of the enthusiast (non-oem) market and that was on a similar process...not 7nm - which is more profitable.

I just want to point out that the nm numbers don't matter for direct comparisons - you need to look at feature size. Intel's 10nm is still overall slightly better than Samsung/TSMC/GF's 7nm nodes. I think the gap has been closed though and it will only be a matter of time before the other three are shipping better manufacturing giving AMD an advantage. Intel will either have to sell off it's foundry or compete in some other way.

Guys i see something else my self, intel was since ...20 years or so on a superior lithography 1 step ahead , this is the first time ever amd will have the advantage on it! Amd still is on the backfoot in this both 12 and 14 nm from glofo/tsmc are infirior to intels 14nm+++++++++ (am i missing a + or two?) .

Denial:

I just want to point out that the nm numbers don't matter for direct comparisons - you need to look at feature size. Intel's 10nm is still overall slightly better than Samsung/TSMC/GF's 7nm nodes. I think the gap has been closed though and it will only be a matter of time before the other three are shipping better manufacturing giving AMD an advantage. Intel will either have to or compete in some other way.

Intel's 10nm and GloFlos 7nm are supposed to be pretty close in transistor densities. However there is so much more than just densities albeit that is very important. One key thing is how IBM designed GloFlos 7nm to push 5Ghz+ frequencies much like Intel's 14nm ++ does. I suspect Intel is working on that as we speak for 10nm and is likely what is holding them back as they do have low power low frequency 10nm parts out in the wild now. I don't think Intel will have to sell off it's foundry. Honestly once you hit these 10nm / 7nm nodes there isn't much more you can shrink anyhow. You really only have one more full node shrink left(5nm for Intel or 3-4nm for others) until you have to figure out something entirely new. There is going to be a long lag where everyone is on the a similar process node and those processes will get tweaked a lot over time like intel has done with 14nm.