AMD Ryzen 2018-2020 Roadmap leaked: Castle Peak, Matisse, Picasso, Vermeer and Renoir

Press F for Intel

videocardz.com/75217/amd-ryzen-2018-2020-roadmap-leaked-castle-peak-matisse-picasso-vermeer-and-renoir

Other urls found in this thread:

anandtech.com/show/12377/tsmc-starts-to-build-fab-18-5nm-in-early-2020
anandtech.com/show/12438/the-future-of-silicon-an-exclusive-interview-with-dr-gary-patton-cto-of-globalfoundries#three
twitter.com/SFWRedditImages

...

...

7nm will be the limit for all CPUs until quantum computers arrive?

So. Essentially AMD went fill tick/tock.

How many times have you fucks declared the end of Intel, only for Intel to quickly answer back with a superior product?

>several months later and no stock for 3 months
>quick

>still inferior for multithreaded performance at similar price points
>superior

Intel

B T F O
T
F
O

>mfw 7nm Threadripper coming

>inb4 intel releases something on 14nm+++

I think the theoretical limit was 4 nm

you are mistaken, but it's pretty close tho.

Not quite. It's different from intels tick/tock. Intel had alternating architecture and process nodes updates while amd has both in one step(zen and zen2 i.e.) and releases a refresh tier product a year later when the process is more mature and some initial bugs of the architecture update get fixed. Intels way promises a steady progress while amds model will give a big jump every two years while also releasing a more mature product (with a tiny performance improvement) a year later. Zen+ is just a fixed zen, with a bit higher clocking process, better boosts and removed problems(reduced latencies i.e.). Zen2 will be a jump in performance but will have it's own flaws, but you would need to wait another year to see these flaws fixed(probably not worth it).

how many times are you planning to fall for the same bait?

5-10% more multi core performance isn't something to brag about when you have 33% more cores.

t. proud r5 1600 owner

No not really.
anandtech.com/show/12377/tsmc-starts-to-build-fab-18-5nm-in-early-2020
Of course don't forget that 7/5nm is actually bunch of cleaver tricks to limit quantum problems.
>Intel can't solve 10nm
>TSMC is already building 5nm fab
TSMC won silicone race

Based threadripper-guy poster

IBM showed off a 5nm GAAFet wafer a while back.

Shut it down.

>only for Intel to quickly answer back with a superior product?
Sorry you fell for the overpriced TIM housefires

this is very unfair business practice when blessed Intel Corporation stuck on 14nm for the rest of the year and no architecture updates in sight you will be hearing from our lawyers

Oy vey better release a new socket

which manager changed the name to artists?

F

My cock is ready. I never thought that I would buy a new AMD platform anymore.

I'll switch both PCs. 2700X seems to tempting. I can probably switch the 5820K without paying much to get 30 - 40 % performance increase; 2 cores, lower consumption and higher clocks. This Shittel CPU only does 4.3 GHz at 1.3 V.

New Intel HEDT CPUs are trash. It seems their era is over.

It has only 16 lanes. Thats a joke. I won't buy any mainboard that doesn't support at least two NVMe drives.

>AMD literally has Tick-Tock now

But desperately needs it. At least until the clcks match Intel.

Replaced my 4820k @4.5ghz with a [email protected]
Single core perf is at worst equal and on average ~5% better
Double the cores
12 watts less power at full load.

Got the RAM+CPU money back from selling the old x79 gear, only had to really pay for RAM. (For which I paid dearly.)

So you're on x99/x299 then?

Ivy Bridge E? My 1700 runs at 3.8 / 1.35 V (Just quick OCd it will probably do less). My 5820K runs at 4.3 / 1.28 V, 3.6 cache.

Intel draws about 100 W even idle (Though with much RGB, sound card, 1080 Ti). My mATX PC draws around 100 W while converting a video into H265.

50 points difference, it would reach 478 on 4.2 GHz. Zen+ will deliver, actually better IPC and maybe more clock and more cores. I'm hoping to get 400 € for my CPU and X99A/3.1.

Yes. I'm thinking about a second NVMe drive. Getting one without the need of any extension cards would be nice. These are insane for video editing, but SATA SSDs will bottleneck.

>Ivy Bridge E
Yep, it was a great platform.
I tossed up the possibility of moving to a 4930K or 4960X, but the used CPUs all cost more than an x370 board + 1700.

I'm pretty happy with the move.
If a 4960X had been cheaper, I'd have stuck with x79 though.

If Intel wants to seriously throw a spanner in AMD's works - all they need to do is reintroduce the Overdrive CPU concept.
Intel could have sold me some $800 8-core Skylake-E if they had one for x79 platforms. It would have still been cheaper than Board+CPU+RAM.

Same on all X platforms. I thought of a used 5960X, but these CPUs still often cost twice as much as a brand new 1700 and mainboard.

But coming from a hexacore the 1700 isn't a worthy upgrade. The 2700X will be. So I have one energy efficient 1700 rig for converting and archiving, one high performance 2700 rig for gaming and video editing.

Skylake X is shit. The old HEDT platforms still lived longer. You could get the 5820K for 340 € in 2015, only board and memory cost a bit more, it was still great value because you had six cores long before anyone else had.

Not really, seems it is Tick+tock / smaller refresh / Tick+Tock / smaller refresh / Tock

(the last bit is assuming that in 2021, Zen 3 comes as a tock, but still on 7nm so no tick)

We have good taste.
Zen 3 chips are called Courbet and Titian.

Then why are you looking at AM4 dude? You have to go for X399 if you desire non-mainstream PCIe config.

Will we see bigger CPU's with more cores or just 2 part CPU's because the Quantum computing sector ain't going to be affordable for anyone with anything less than supercomputer type budgets.

>TSMC is already building 5nm fab
everyone building 5nm fab, nobody else except declining TSMC bothered to make first shovel dig a news topic

anandtech.com/show/12438/the-future-of-silicon-an-exclusive-interview-with-dr-gary-patton-cto-of-globalfoundries#three

5Ghz is not a rumor anymore.

What will CPUs be like in 10 years?

same as now but faster, prediction might get really freaking good, I mean otherworldly good

>tfw want dem 8 cores for compiling LLVM

They (not Intel, lel. IBM and etc) can go as far as 2nm, actually.

What will come after Zen 3? I know Intel is working on a new uArch sometime 2020-2022. And when can we expect DDR5?

There will be no mainstream desktop quantum PCs in at least 20 years more. The next step is graphene/graphite nano tubes.

DDR5 will probably emerge in 2020, and won't have any real market penetration until after that.
What comes after Zen 3 is supposed to be a new core family, though I doubt Zen will be total EoL. They'll probably have some transitional SKUs giving it a bit more life.

DDR5 looks 2021 together with Zen3

Will AMD ever fix their shitty draw call performance?

and tack another 20-30% on to the price of RAM to make the insane CPU prices less noticeable in the total build cost

>muh tick tock
since the very first integrated circuit optimizing the architecture and shrinking the size have been the two methods of improvement. just because intel gave it a catchy name and for a certain period of time applied them alternately doesn't mean they invented something.