Zen 2 expect to release by the end of 2018

youtu.be/KEU6gk8N5yc?t=1272
>64 cores leaked
hothardware.com/news/amd-epyc-2-64-cores-128-threads-and-256mb-l3-cache
>AMD Second Gen EPYC Beastly Server CPUs Could Rock 64 Cores, 128 Threads And 256MB Cache

AMD going in for the kill.

Intel on suicide watch

Other urls found in this thread:

fool.com/investing/2017/07/26/the-price-of-intel-corporations-10-nanometer-failu.aspx
phoronix.com/scan.php?page=news_item&px=AMD-EPYC-Linux-4.15-First-Test
tomshardware.com/news/new-amd-processors-asus-bios,35969.html
twitter.com/SFWRedditImages

delet

That really manages my engine

cores and threads are the new clock cycle, watch. pretty soon we will be counting cores in the thousands, hundreds of thousands, millions, etc. chips are made out of sand, it's all about design, once they learn how to deal with large numbers of cores it'll take off, like this.

OY VEY DELET

That kills Intel's datacenter roadmap.

Noooo! It isn't true!

It's not only the cores for the Zen 14+ they are pushing up over 4.4ghz. That's not even Zen 2

>That kills Intel's datacenter roadmap.
fool.com/investing/2017/07/26/the-price-of-intel-corporations-10-nanometer-failu.aspx
>As if it couldn't get any worse, by Intel's own admission, its first- and second-generation 10nm technologies -- 10nm and 10nm+, respectively -- will offer worse performance than its upcoming 14nm++ technology . Intel says the company's 10nm technology won't open up a clear performance lead over its 14nm++ technology until its third iteration -- known as 10nm++ -- which should go into production sometime in 2020.

Intel's effectively killed themselves

>Intel on suicide watch
They'll no doubt find some scummy way to get a competitive edge.

>They'll no doubt find some scummy way to get a competitive edge.
They'd probaly try and throw money at PC companies but at this point I think the industry is tired of all their shit. I mean look what they did the the Motherboard Co.'s recently. They end up screwing everyone they get involved with.

Why the larger L3?
Its so costly, in transistors and die area, also power consumption. Increasing the size of a cache generally increases its latency as well which can hurt performance if its not carefully balanced for hit rate in your typical workloads.
If this is legit then AMD would need to have faster caches, or potentially a different cache policy. I don't think they'll go back to having a clunky L3 victim cache.
They'd only do this if they were raising IPC enough to justify it. Very interesting sign.

>I think the industry is tired of all their shit
I'd like to believe that but i just don't know.

You need to feed the cores.
It's still 8 memory channels.

What is this timeline? Trump gets elected, hillary's career is over, bitcoin is money, apple is bankrupt, intel is getting BTFO by AMD..

just how?

What's the point of Zen+ in early 2018 then? Two generations in a year seems like a bit much.

Refresh with higher clocks and some tweaks to keep the market momentum going.

Have faith in mummy
Dear I say... /besttimeline/

>apple bankrupt
>~200B liquid assets
wdymbt?

You deserve the timeline and the quads, user.

DRAM is the true LLC in the memory hierarchy, and they're apparently making 3200mhz DIMMs standard for Zen 2.
Increasing DRAM speed and subsequently the IF data rate will improve performance a good degree. Why increase the L3? Its an inordinately strange thing that sticks out like a sore thumb because its such a tremendous investment.

Why is the threadripper packaging a vintage television?

What tremendous investment?
7LP offers the most dense SRAM in the market. Besides
>3200MHz DIMMs
Someone needs to make JEDEC-compliant ones first.

Its actually a display case for Brian Krzanich's meager genitals.

Quick progress, you're just used to Intel plowing you without lube, welcome this instead.

>refrain taxmoney from the government so bad that even you can't spend it
>fire senior engineers, designers in favor of hiring mentally unstable "diverse" people that literally do nothing but implement emojis
>steve jobs dies, there is nobody else to steal useful features of other products and whip the employees to implement them into one product
>even though another dictator comes in charge, there is no talented employee to whip

you won't believe how fast that 200B assets will be wasted in the hands of those mentally unstable freaks.

The investment in transistors, die area, power, yields. Doubling L3 isn't free. Memory is one of the areas where defects are most likely to appear so having more of it hurts yields in any given IC.

>64 cores
For what fucking purpose?

7LP offers hands down the most dense SRAM in the market.
The cores itself are very area efficient for anything x86.
So why not?

killing intel

To make hyperscalers jizz.

this, no one needs more than six cores in the server market

With no survivors.

>gen 1 is paper launch
>here kind sirs gen 2
wow another paper launch
>the ceo literally shorted all her stock

Because you don't just spend transistors on something that can actually hurt performance by being slower unless there is a very good reason to justify it.
Whats different between Zen 1 and Zen 2 that causes a high enough increase in L3 hits to justify doubling it, but not just hit rate, the size of the instructions as well?

I don't think they'll implement 256bit datapaths in their FPU. I also don't think they're going to have SMT4. There are no radical new instruction sets on the horizon to get adopted and justify more L3.

What I'm getting at is that it shows that Zen 2 structurally could be a very different beast from Zen 1 rather than a small generational upgrade.

kikeripping
that and servers

>this, no one needs more than six cores in the server market
>this, no one needs more than four cores in the server market

name one non-anti-semitic software suite used in the data center that utilizes more than two cores

AMD needs twice the cores of Intel to acheive IPC parity.

Any Virtualization Software ever...hold up, don't answer just yet, you said Data Center, you didn't say where in a Data Center.

this really parallelizes my threads

My dedicated streaming rig.

Glue can't beat Intel.

Despite all this, the general public still demonstrates an ever increasing propensity for shoveling all things Apple down their gullet.

...

Shoa

EPYCs are cache starved more than Ryzens, increasing L3 would let the CPUs pull more from the xGMI and die-die links while using less power.
Big L3 is crucial for off core IF communication.

Zen 2 will hit 5GHz right?

Who cares.

I care.

What if it hits 4.9GHz?
What if it has 20% more IPC than Zen1? Will you care?

>Intel needs twice the IPC of AMD to acheive core parity.
Intel's idea of innovation is just brute-forcing the Ghz to volcanic proportions

phoronix.com/scan.php?page=news_item&px=AMD-EPYC-Linux-4.15-First-Test

Meanwhile EPYC just keeps steadily increasing its performance with kernel updates.

>Despite all this, the general public still demonstrates an ever increasing propensity for shoveling all things Apple down their gullet.
This is the sad fact of the matter. Casual's don't care how shady the people they are dealing with are and how much damage they do to the environment.
They just want what the marketers tell them to want.
>Muh 500fps 480hz
>Muh 0.0002 response time on my apple

Zen 1 already does hit 5ghz and beyond when overclocking, the design isn't clock limited.
What you're asking is if SKUs will ship with factory clocks of 5ghz with sane TDP. Nobody can answer that. GloFo's target goal for their 7nm process is 5ghz for complex ICs. This doesn't necessarily mean Zen 2 will reach those clocks, though it is likely that they'll be significantly higher than current Ryzen chips.
A 25% increase in base clocks could be possible, and small momentary XFR bursts could reach 5ghz on some SKUs. Thats a more realistic expectation than a 5ghz base clock.

Installing gentoo.

>it gets better
Ohwhat

I think what we're seeing here, as highlighted by
>What I'm getting at is that it shows that Zen 2 structurally could be a very different beast from Zen 1 rather than a small generational upgrade.
Is that Zen's modular nature not only helps in the production of chips, in that you can pick and choose dies and benefit from insanely high yields. But also plays part significantly in designing future chips?
AMD now have a high performance, low power core that they can always go back to, if they can increase L3 to assist in heavier server workloads, it may well lead to increased latency across the chip, which may effect the everyday performance of the average consumer, will this performance loss be negated by an increase in clock speeds thanks to a new higher power process? Yet to be seen

>dem quads
Holy shit bois Apple gonna be bankrupt! The prophecy of kek has foretold!

>want to upgrade my old skylake i5-6400 but since I have stable OC of 4.4 Ghz, its not worth the upgrade to Zen yet
>try to wait for zen+ or cofee/whatever intel lake is next with cheaper 6c+ cores that can OC
>hear zen2 will be same year but different quater
>have to wait for zen2 to see if its worth the upgrade

mmmm.... fine wine

Either way it'll be on the same socket, and AMD were able to get motherboard vendors to drop BIOS updates to accommodate Piledriver refreshes

ASUS already dropped a new AGESA to support APUs and Zen refresh.

Datacenter likes large SRAM caches, this is gonna be a enterprise only die that will trickle down to consumers when demand doesn't outstrip supply, so probably q2 2019

So 2018 is a nothingburger in terms of hardware. 64 core Zen 2 cpus are going to be at least $5000.

Who cares, hyperscale will buy it anyway.

I find this especially surprising as ASUS couldn't do a single fucking thing right when Zen launched.

Others will get updated soon enough, at least I hope so.
Though I'm kinda interested in 2nd gen motherboards, maybe they play nicer with more powerful DRAM

ASUS is trying to be a mommy's boy.
They know where the money is.

We'll see, as per RAM speeds right now there's not much wrong, some vendors have their boards happily running 3300Mhz

Monitors fucking suck right now so what's the point in advancing.

>muh gayms


This is a processor so you can emerge @world in a few minutes.

Reminder that Threadripper has better per-core performance than equivalent Intel parts.

Also Ryzen in general is pretty much bang on Skylake X in terms of IPC and it's not like there's been much progress.

They're pretty close, but with Rome next year and Intel's only response a skylake-EP refresh(Cascade Lake)..

Intel doesn't really have much going for it, this is gonna end up with lowend EPYC2 battling out with Intel's Platnium Xeons.

Which will be hilarious, AMD isn't playing fair, we'll probably have a situation where AMD is using over 1000mm^2 of total silicon against Intel's 660mm^2
It's gonna be a slaughter.

>locked i5
>overclocked
How

Where the fuck are you seeing Zen 1 hitting 5GHz? Shit struggles to get past 3.8GHz lmao.

AMD's going for the jugular, Intel's next server line is Icelake-EP which is 10nm+

And by the looks of it won't really be out until late 2019.
And even if it comes out I really doubt Intel will get anywhere close to 64 cores on a single die without some fancy EMIB magic.

And even then, EPYC will already be on Zen3 cores when Icelake-EP comes out.

We're getting a repeat of first gen K8 opterons vs Intel at this rate.

Ryzen chips were hitting 5.1 and 5.2ghz under LN2 the very first day they launched. Their only limitation to clocks is voltage as a trait of the 14nm LPP process.

Jesus fuck, can you imagine the price tag on a 64 core monolithic monster Xeon? fucking 28 cores are already pricey as fuck nigga.

Intel chips regularly hit 5.1GHz at home an on air cooler
>poojet engineering lul

Bad thing for Intel that won't be happening, I keep hearing of a 42 or 46 core 10nm* server line, and that's nowhere close to enough to battle EPYC2 one YEAR LATTER, much less EPYC3 the same year.

Anyone know if epyc has been doing well?

It just started selling to vendors and resellers, we don't know and OEMs and hyperscalers that bought directly from AMD it won't tell you.

Right around 5.1ghz is the record for Broadwell-E, the comparable intel chip available at the time.
Try again, kiddo.

Wil x270 boards still be compatible why zen+ probs 16 32 t non tr im guessing. Alao will x299 boards work with 64 thread cpus?

Speak English, nigger.

Will x270 boards work with zen+ once bios updated. Can the board physically handle that many cores and threads. Can x299 boards support tr+?

X200 boards are Intel chipset motherboard

X370 whatever. Naming is too similar

If there's a BIOS update they will, ASUS has done preliminary support yesterday

Expect Gigabyte and ASROCK to do the same, MSI and Biostar are mostly garbage so I don't expect them to do the same unless there's some contract clause with AMD that forces them to update.

tomshardware.com/news/new-amd-processors-asus-bios,35969.html

You can't short shares you own

Are there actually people who buy MSI's junk with those "Killer!" NICs?

Oh wow, it took that long? Is it usual to take so long or did amd fuck up?

>Oh wow, it took that long?
Validation, especially for hyperscale/t1 OEMs is long.
>Is it usual to take so long or did amd fuck up?
A bit of both.

I personally think all supply went to partners (Baidu, MS, Tencent, Samsung) directly, and now they have enough stock for resellers and integrators like HPE and Dell.
There was also that new firmware came out just recently and validation time was needed so integrators preferred to wait before selling.

Some people still can't get fucking samples.

Apu

Pwople who buy linkshit gaming routers

Its been out for a while through that one reseller

Well, obviously demand is outstripping supply, that's both good and bad, and if AMD can't even fill up demand with 90% yields I don't really know what to say other than incompetence or everyone and their mother wants EPYC, even Intel wouldn't be able to have that much supply in that case.

Is it true that vega 64 and 56 might get a refresh in spring?

Probably, dunno. That, or HPC fp64 version or Polaris replacements might be more possible.