BWHAHAHAHAHAHAHAHAHAHAH

That only pertains to a very small batch of test mobile platforms. No consumer would get 5nm, not even Samsung even though they already know how to make even 3nm stable and feasible (and they're currently the only ones who can). 2019~2021 = 7nm/7nm+ for the entire industry.

> September 21, 2018

I only feel bad for Intel if by some unimaginable curse asteroids crash on all their fabs.

>asteroids
I see what you did there.

Attached: 1543231195449.jpg (600x442, 86K)

So only TSMC and Samsung are in the transistor race.

Yep. GloFo decided to play it safe since 14 is way too profitable and mature.

rip

Hm? Who?

GloFo, are they ever going to catch up?

they have the tech, but their parent company decided that they've been bleeding money for way to long and 14nm is a safe bet
maybe when EUV becomes a thing they'll try to catch up

Playing safely for now is not going bankrupt or whatever. If anything, they're going to do just fine for a while, since they do 14nm I/O middle part for Zen 2. Yes, technically BOTH GloFo and TSMC are making Zen 2 for AMD. Chiplets are TSMC, while large I/O is GloFo. We'll see what it's going to be by Zen 2+/Zen 3, but as of right now - each and every single Zen 2 is made partially by GloFo, so they have healthy and steady income regardless if they're "behind" or not. They couldn't do 7nm even if they'd wanted to, due to lack of equipment and resources, so still working for AMD and being a part of Zen 2 product is actually a HUEG win for them.

Attached: 1541598935254.png (900x980, 103K)

Attached: 5759f915556fe422e5f3d5b07a0b32fe009221f6701bc90b7ba46d8dbdaf8f2c.png (200x248, 35K)

Wait for them to be humbled and make an honest attempt at improvement, and failing before you feel bad for this nest of vipers.

The 7nm process isn't profitable with yeilds to make large chips. The 14nm io (litterally a on socket north/southbridge) indicates higher clocks on a more refined node while keeping yeilds high.

This means AMD can quite litterally bin high frequency io chips for desktop and pairing the most solid io packages with the EPIC server chips.

Zen 2's yields are, like, 89~90%, mang. And that's counting PERFECT binning.

In all honesty Intel really let me down and I own a 6700k(died (my fault) and a 7700k. It is not because of performance or price but to a larger extent these critical security flaws that fall in the lap of design that cheats the user. Even worse they still release chips with these problems.

This shit reminds me of the old Pentium 4 days when Intel fucked everyone over and willfully lied and cheated and bribed their way to billions.

Zen2 or Zen3 will likely be my next system.

Attached: 1542025819200.jpg (638x599, 123K)

If the whole chip was a monster at the 7nm node yeilds would be less then 50% and the price would be shot. That is why AMD is going the 7nm small die chiplet route.

See youtube.com/watch?v=I6AG-woscDU

A lot of the design choices for zen in general focus on being (relatively) easy to actually build over pushing the limits of performance. This makes sense given the scale of AMD's operation and how they have to fit in with GloFo and TSMC's manufacuring availability.

At a high level chiplets are great but there are some staggering performance pitfalls that needs some clever work to get around (latency being the most obvious one).

Can anyone explain how this affects consumers and why it's important?

>2 months ago
old as fuck news kiddo

Watch the video. It litterally answers that exact question.

>how this affects consumers
2022 = 98.82% of all computer users on the planet are on AMD Zen.

>at the 7nm node yields would be less then 50% and the price would be shot
Only if you'd try doing it right now, at old tech. With EUV, 7nm will be perfected.

Doubt it. It took a long time for the 14nm node to mature above 95% yeilds. Current TSMC 7nm yeilds are just profitable. Since the chiplets are so much smaller then a regular CPU the yeilds greatly improve.

Attached: 1541721087196.jpg (492x578, 48K)

A CPU that died? Without retarded overclocks?

post Jow Forums humor I'll start
>Moore's law

He clearly tried to DELID'DIS and bricked it.

>not still using sandy bridge

You are responsible for all of this.

Ah, had a feeling it was just a meme. Thanks for clarifying.

thanks, Jim

you might be spelling "yields" incorrectly

No, risk production is not for test chips, its for companies like Apple who are willing to pay for access to a node early at the risk super low yields.
TSMC will have 5nm EUV in full production in 2020. Thats heavy EUV integration, up to 14 critical layers.

>Zen 3 with RTX 3.0

Attached: 1392769671555.gif (300x223, 1.41M)

Global Foundries is doubling down on advanced SOI, moving away from FinFET all together. They'll offer 14/12nm LPP for years to come as commodity nodes, but their bread and butter will be SOI 22FDX and 12FDX.

>advanced SOI

Attached: soyim.jpg (1200x800, 73K)

Its the only acceptable form of SOI.

...wut?

It doesn't. AMD's chipsets are like 55nm or so.

*Kvetching intensifies in the distance*

Yeah just looked it up, B350 is 55nm and it's TDP is 5.8 W. The only difference is the manufacturing cost really.

*Gevalt intensifies somewhere nearby*

>2019
>The year of Rockchip

...what?

Rockchip (Fuzhou Rockchip Electronics Co., Ltd.) is a Chinese fabless semiconductor company based in Fuzhou, Fujian province. Rockchip has been providing SoC solutions for tablets & PCs, streaming media TV boxes, AI audio & vision, IoT hardware since founded in 2001. It has offices in Shanghai, Beijing, Shenzhen, Hangzhou and Hong Kong.[4] It designs system on a chip (SoC) products, using the ARM architecture licensed from ARM Holdings for the majority of its projects.[5

>The H310C chipset is new spin on existing silicon whose headline feature is its expected support for Windows 7
WUT?

I know that, faggot. I was asking why 2019 is "year of Rockchip", not what the company is.
Rockchip's shit is literally in 98.82% of all e-ink e-book readers out there nowadays, so it's harder to find a person who doesn't know about it.

AYY LMAO
INTURD IS FINNISH & BANGRUPT

Sub-nano meter production has been around since last year in R&D.
The only thing keeping it from mass production is low yield and money.

>user thinks lower nm = faster

>not wanting a CPU for computer that runs in smartphones.

>being this fucking retarded

22nm+++++++++++++++++++++

>intel shill detected.

Good on you intel. Make Intel Great Again. As biology has taught us for the last couple of millennia, smaller does not always equal better.

Intel is just in the process of diversifying. They’ve realised that the chip market is now flooded with cheap CPUs so are now investing their resources into other products such as intelligent cloud and so on.

y-yeah
intel gpu superpower by 2020

I don't even know if this is bait because this board is so usually fucking retarded.

>low yield
See .

>14nm more than 20 later after the last attempt

Attached: 1998 i740 Intel discrete GPU.jpg (700x452, 85K)

>20 years later
slowfix

Just wait for Intel gaming GPUs in 2077

>14++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

But it's the chipset user. If you looked at any mobo on the market today, it would have parts ranging to 65nm.

Talk about click bait

But I'm not ready to join the great red botnet.

>TSMC will have 5nm EUV in full production in 2020
fucking based

14+8nm

>wise user detected.

Intel are just naturally repositioning themselves in the market. They’ve realised they no longer win the desktop CPU wars any more - does it matter as Paas, Saas etc is where future development will come from.

Intel is just shedding some fat, that’s all.

Intel will rise again with the force of 1000 shekels

2077!
trash economy!
teaching african children how to program javascript!

|
|>
|
|
|

I can tell you what they don't do: competitive consumer GPUs.

RX 590 and Vega 56 are selling like hot cakes right now, but nice try anyway GayForced ReTardXed noVideot.
Also, look here:

Attached: GIMPWORSE VS FINEWINE.png (736x736, 1.55M)

A retard-a-you.

Attached: 674858568.jpg (1265x1416, 655K)

gibbe working prim shaders nao

Attached: 1495253886505.jpg (957x621, 90K)

WEW

>no hands
>|
>|>
>|
>|3
>|
weak desu

7nm.
They need time to fix pajeet's fuck-ups, after all.

Nose is more than enough.

>one autismo alters a meme because "X is more than enough"
the nose + hand-wringing are the key elements of every abstract jewish representation but ok whatever you say lol

Attached: 1509508656629.jpg (800x800, 134K)

NOOOOOOOOOOOOO
THIS CAN'T BE HAPPENING

Attached: 1527629778452.jpg (679x758, 54K)

There are a couple other titles where GCN has always done really well. So long as that front end doesn't get taxed the cards are total power houses. Vega is great arch with only a couple caveats. Drop the vcore .1v, change the power limit in Wattman, set the core clock to 1600mhz, and you get a major reduction in power consumption while performance increases even more. It doesn't take a golden sample either, thats the average.

If prim shaders actually ever worked Vega would have been monstrous. Maybe AMD finally got it in Navi. Fingers crossed.

>Fingers crossed.

Attached: 1540609357750.gif (600x600, 203K)

If they were close to being able to do that, wouldn't we be hearing about it by now? Put the terror of the night into Nvidia since their Tensor core stuff doesn't seem particularly impressive when developers have to deal with implementation.

Who knows? It is slightly concerning that AMD haven't said or leaked a word about Navi, and CES is just two months away. If they don't detail the arch in full at CES then its sign that the GPU division has an unworkable disaster on their hands.
AMD was pretty tight lipped about the Zen core, but they did brag about the 40% IPC uplift pretty early on, and they hosted a whole HotChips presentation on the core arch. They've done similar in the past for discrete GPUs and APUs alike. They haven't done anything of the sort for Navi.
You don't just have the tech presentation right before release either. They plan these things well in advance to drum up hype for the product to drive sales. Waiting until CES 2019 to talk about Navi likely indicates 2H 2019 release date. Possibly later than that. If they don't talk about it at CES then the graphics division is fucked beyond repair. If that comes to pass then I could see them increasing orders for 7nm Vega dies from TSMC to sell in consumer cards.

With how GPU prices are right now, I could see a 7nm Vega holding them over until they have something worthy to show. I don't really see RTX having that much of a future, at least.

IT would spoil things about PS5 , guess Sony would be mad

>I don't really see RTX having that much of a future
RTX is a snake oil Ponzi scheme scam.

>If prim shaders actually ever worked Vega would have been monstrous. Maybe AMD finally got it in Navi. Fingers crossed.
Fuck primshaders. If RTG doesn't full on revise GCN's front end to eliminate the 4 triangle per clock bottleneck once and for all for Navi, then they should all be ritually crucified.

It's been the same goddamn fucking problem since fucking Fury. They need to stop kicking the can down the road and fix GCN's front end.

I hope Big Wang will spearhead that first and foremost

>Big Wang
Who?