| fatslob-:O said: It takes more than just the transistor technology itself to drive high frequencies, the logic and physical chip design matters too ... |
I'm aware.
| fatslob-:O said: Nah, I imagine AMD will only start with peanuts as I think Zen is only originally intended to get customers to trust AMD ... |
Even a few percentage points of marketshare can still result in millions of revenue.
And when AMD posts loss after loss every quarter. (January 2017 they had a -29 million dollar loss) any slice of marketshare gain is a good thing.
| fatslob-:O said: I honestly hope Zen 2 doesn't share the same socket as that limits microarchitecture improvements ... (AVX-512, TSX, PCIE 4.0, and IPC) |
You can retain socket compatability but upgrade the chipset to take advantage of newer features. Which is what AMD has historically done.
AM3+ being the crappy scenario where AMD neglected the platform for years.
Sockets should only be replaced if there is a need for it. And IMHO Intel replaces sockets way to often.
| fatslob-:O said: Plus this, GF will start mass producing chips on wafers with 7nm transistor technology by 2H 2018 so the original Zen will be history by next year so no need for AMD to invest anymore than they have now. AMD and GF should focus on getting high supply for Zen 2 to push large volume sales for the next holiday ... |
That's provided Global Foundries is actually on time with it's 7nm technology. (Likely using a 14nm FEOL or BEOL.)
Besides. early 2018 is when "Risk Production" of 7nm begins. Who knows if delays will arise.
| fatslob-:O said: AMD's golden years weren't with K7 or Zen, AMD's golden years were with the K8 and most likely Zen 2 since they mark the first time that AMD had truly competitive offerings compared to the competition, if not then superior! |
Disagree. K7 was part of AMD's "Golden Era".
We got Thunderbird, Morgan, Spitfire, Palomino, Thoroughbred, Barton, Thorton CPU's in the Athlon, Duron, Athlon XP lineup.
AMD pushed it's advantage with K8.
Before Intel moved the L2 cache-on die with Coppermine, AMD had a big CPU advantage, then when AMD did the same, AMD was ahead in the performance stakes.
Then after Coppermine we got Tualatin and Willamatte which really didn't reach the same heights as K7... AMD retained it's advantage untill Intel dropped Northwood and AMD's K7 seemed to stagnate as it couldn't scale in clocks well enough. (And is when AMD introduced it's PR rating.)
Duron smashed the Celeron at pretty much every step.
| fatslob-:O said: Intel dual cores are also subpar for the fact that it doesn't even support AVX/AVX2 ... (consoles do support AVX and Intel fusing off AVX from dual cores is another generation of missed opportunities) |
I concur.
AVX and Quad-Cores should be the minimum requirement for all CPU lineups in 2017 IMHO. Hell. My grandmother has a 6-core processor and she only uses Facebook.
| fatslob-:O said: AMD has a WSA with GF until 2020 so it's almost certain that the next AMD CPUs, APUs, and GPUs will be manufactured by GF |
I never stated/insinuated that chips wouldn't be manufactured by Global Foundries. Have you forgotten that Global Foundries 14nm process is licensed from Samsung?
| fatslob-:O said: Plus what AMD is getting out of from GF isn't all that bad when they can look forward to the next shrink in the near future and then make the switch to Samsung as their main manufacturer since they have clear plans for 5nm and below ... (I think Samsung is the only one that can truly compete with Intel soon on an even playing ground in terms of transistor technology) |
Samsung is still years behind Intel. The "nm" has turned into a marketing term that fabs are flaunting.
| fatslob-:O said: That's not what I heard ... |
Seems Anandtech isn't privvy to all the details this time around.
Global Foundries is tooling a few fabs up for "12nm". Granted it has a few caveats.
https://www.globalfoundries.com/news-events/press-releases/globalfoundries-extends-fdxtm-roadmap-12nm-fd-soi-technology
www.youtube.com/@Pemalite
