http://seekingalpha.com/article/4002864-amd-zen-likely-good-enough

>SA

Lol

>seeking alpha
Is this something like beta news?

>>56330124
>>56330124
>seeking alpha
>literaly kotaku tier

>(((market analyst))) makes absurd conclusions vs one Chink's ES bench runs in a terrible RTS game and extrapolates an entire thesis based from it

Seeking Alpha is my favorite stock manipulation site.

>>56330124
ayymemed

>>56330124
>The launch of AMD's new high end cards, Fury X and its derivatives (Fury, Fury nano) was only enough to reach the high end of the prior Nvidia

Fury X is equivalent of Titan x and 980Ti. What is he smoking?

Reason why Fury X and Titan X for that matter failed is because 980Ti was so fucking amazing.

It's pretty obvious that Zen will be a disappointment. Single thread performance will be massively improved but will end up hampered by the process tech (14lpp sucks) and the improved cache (faster cache = terrible clock speed, this is why bulldozer's caches were designed to perform terribly)

AMD has been pulling sketchy accounting, slash r&d and replacing skilled engineers with college kids on h1b visas, hyping shit products constantly, etc. He's right to be skeptical. Zen will likely be a great improvement over AMD's previous offerings but won't bring them any closer to being a viable competitor to Intel again.

>>56330124
just how retarded are you to post the link in the subj-
>SA
nevermind.

>>56330915
Samsung's 14nm LPP process is pretty outstanding. You need to stop regurgitating bullshit you read on the Anandtech forums.
AMD maintained the same power per clock shown in Excavator, and a single Zen core is larger than the entire Excavator module comparatively.
3.2ghz for 10w per core is nothing to scoff at. The process displays all characteristics that they said it did. Low leakage, increased fmax over 28nm HPP, extreme area scaling possible.

Summit Ridge isn't going to clock high, but this isn't a problem with the process, just the variant AMD chose because the dies are primarily fabbed for enterprise. High density, super low leakage, and moderate clocks.

>>56330969
AMD's Summit Ridge is competing against Haswell-E. Having IPC north of Sandy Bridge is getting pretty damn close to intel's high end.
None of this comes as a surprise. See: http://www.anandtech.com/bench/product/1684?vs=288

Take Excavator and add L3. Excluding FPU bound ops the Excavator core isn't terribly far behind Sandy Bridge, a 40% average uplift would absolutely be competitive there.
Clearly there are some best case scenarios for the Zen arch, AMD's Blender demo showed just that. On average AMD will probably have per core performance around Ivy Bridge. It'll do better in somethings, and worse in others. End of the day it still shaped up to be exactly what they said it would be.
40% IPC uplift over Excavator, no CMT penalty for multicore scaling, SMT design.

>>56331135
>Samsung's 14nm LPP process is pretty outstanding.

it's worse than both intel's 22nm and tsmc's 16ff

>AMD maintained the same power per clock shown in Excavator,

only according to roy taylor's drunken marketing antics

>3.2ghz for 10w per core is nothing to scoff at.

that's not how zen's boost system works. all core boost is 3.05ghz on the 8c ES.

>Summit Ridge isn't going to clock high, but this isn't a problem with the process,

it's just as much a fault of the process as it is of the design. 14lpp is garbage electrically and the only ICs being produced on it are because of old contracts (wafer supply agreement) and desperate businesses that are sending work to samsung/glofo when tsmc can't meet their demand.

>>56331215
>le AMD is involved so everything about it and related to it is bad
Nice to see you show your head again, autistic NEET shitposter.

>>56331278
>autistic NEET shitposter.

nice rebuttal you got there Freud

https://en.wikipedia.org/wiki/Psychological_projection

>>56331349
>instant butthurt
I know who you are, little NEET. You're the exact same NEET who shows up in every thread, talking out of your ass, showcasing your little autistic agenda against AMD.
All of your posts contain the exact same tone, and you frequently use the exact same phrases.

Intel's 22nm Trigate isn't even comparable to 14nm LPP. Mentioning them as if they're even in the same ballpark is absurd, but you don't care about facts. You're just shitposting. You can find the parametrics of intel's first iteration of their Trigate process on RWT and cross reference them with information Samsung and GloFo both provided on their collaborative 14nm node.

You do nothing but talk out of your ass and shitpost. Everything even tangibly related to AMD in anyway is terrible, the sky is falling, doom and gloom. Its laughably easy to spot you, you autistic NEET.

Let's be honest, AMD never released anything good. ATI was really good on it's own, AMD always shat everywhere.

Let's hope another company will rise with greats CPUs.

>>56331452
7970 GHz
r9 Nano
rx480

>>56331425
>autistic NEET

The things we dislike most in others are the characteristics we like least in ourselves.

>>56331452
K6-2 through K10 were all solid. Only issue was that AMD didn't continue to compete as favorably once intel debuted their P6 arch in the Core2Duo line.
ATI constantly fumbled and was involved in several small controversies of their own.

>>56331467
Careful, you're begging for shitposts. No one is allowed to say anything positive about AMD while the NEET is around.

>>56331483
No amount of deflection will change how pathetic you are, little NEET-kun.

>>56330124
amd will fail badly and nothing will be lost or amd will succeed and Intel will lower their chip prices


in both cases Intel wins
go blue team!

>>56331467
>7970 GHz

rebrand

>r9 Nano

gimped power delivery fucked the clocks and forced it to throttle horribly, also was $650 for the entire time it was relevant which made it a terrible purchase even for the niche it was in.

>rx480

cucked by the 1060, 470 is the only decent value card amd has right now

AMD has not released any competitive products in years.

>>56331425
Well AMD is pretty fucking terrible, he's not wrong

>>56331536
>Tahiti
>rebrand

You're not even trying

>Fury Nano
>gimped anything
The card has a lower power target, there is nothing wrong with its power delivery. It maintains lower power target by having more aggressive PowerTune.

>>56331536
>AMD has not released any competitive products in years

This desu senpai.

There is literally no reason to buy anything AMD anymore.

>>56331574

7970 ghz was a rebrand of the original 7970 with a price cut, anyone could overclock a 7970 to >1ghz out of the box and reap the same performance as the rebrand.

>The card has a lower power target, there is nothing wrong with its power delivery. It maintains lower power target by having more aggressive PowerTune.

that is some stalinist denialism level horseshit right there, the nano has half the power phases as the fury x and throttles aggressively even once you raise power limit (because it can't maintain higher voltages without overheating/damaging the vrms).

Really sad to read this as I really just want to upgrade from my i7 2600k but I never will need to

>>56330124
>Seeking Alpha
More like Seek Help

I seem to recall some dipshit at Seeking Alpha said my company wouldn't last the rest of the year ... That was three years ago. Still waiting to go out of business, and I imagine I'll be waiting for a very long time.

>>56331135
High core/threads always had lower clock than lower core/thread cpu given similar wattage spec(6700k is 4.0ghz clocked @ 91w 4c8t vs 6900k is 3.5ghz @ 140w 8c16t)

8c16t may only default at 3.2 but 4c8t will probably go to 3.5-4Ghz most likely at lower wattage than 8c16t variant.

>>56333115
AMD is targeting 65W for 4c8t and 95W for 8c16t. Those figures are "official," we'll see how accurate they are.

Seems like you could get an 8 core model and a good cooler and overclock the fuck out of it though, get the best of both worlds.

>>56334983
Probably gonna OC like shit. High core count processors are very hard to OC because the thermals go nuts even with small increases.

The thermals are often nuts even without overclocking. 95W isn't that bad, but it sounds pretty optimistic.

>>56331480
ewwww
he's ugly

>>56331215
>it's worse than both intel's 22nm and tsmc's 16ff
It's not, you're literally regurgitating nonsense
>that's not how zen's boost system works. all core boost is 3.05ghz on the 8c ES.
wew lad, who would had thought a ES chip wouldn't be clocked at it's final clock
>it's just as much a fault of the process as it is of the design. 14lpp is garbage electrically and the only ICs being produced on it are because of old contracts (wafer supply agreement) and desperate businesses that are sending work to samsung/glofo when tsmc can't meet their demand.
Whatever you read on anand forums senpai
>>56331653
>7970 ghz was a rebrand of the original 7970 with a price cut, anyone could overclock a 7970 to >1ghz out of the box and reap the same performance as the rebrand.
Rebrands are through genereations, not in the same gen
By your logic every recent Ti card it's a rebrand