Zen 2 16 core details

Zen2 ES 16 Core
Base clock 3.3 Ghzฺ
Boost clock 4.2 Ghz
MB X570
This CPU name can't decode by decode chart

the first eng sample was at 3.8
now at 4.2ghz
twitter.com/TUM_APISAK/status/1126466183740321798

Attached: groot2.jpg (950x450, 301K)

Other urls found in this thread:

gigabyte.com/us/Motherboard/X470-AORUS-GAMING-7-WIFI-rev-10#support-cpu
cpu.userbenchmark.com/Compare/Intel-Core-i5-7500-vs-AMD-Ryzen-7-1700X/3648vs3915
cpubenchmark.net/compare/AMD-Ryzen-7-1700X-vs-Intel-i5-7500/2969vs2910
twitter.com/AnonBabble

>lets glue cores together
>IPC is same as haswell

how does it feel to be stuck with 2010 technology intelaviv fanboy?

I'm not stuck dude you are, my 8700k performs better than AMD in every benchmark.

>every benchmarks
aka only at gaming

What's the point if there is 2950X that goes to 4.4ghz ez

>muh glue
funny intel did the same way back with pentium D, and called ryzen glued together shit, and now they're doing literally the fucking same with their own high end chips. retard

dud
stop sharing this fake shit trash its just a test
8c part may run 4.9ghz ez
16c part will be able to run at least 4.6 ghz but is gonna be more expensive than people expect it as it should be

The whole core 2 quad + xeon lineup was glued and the two core 2 duo dies communicated through the northbridge

Lower price. It's an ES after all. It may go higher (or it may not).
But if someone is buying 16 cores CPU (and they weren't memed into it), single core frequency is probably not the main thing they're looking for.

selling binned chips at higher price
if they can go 4.6ghz it means they can overclock quite a lot (for a 16c that is...)

What did Jim mean by this?

Attached: Screenshot_20190510-170238.png (2048x1328, 648K)

Yes and that's the big market which matters. How many video editing faggots do you personally know and nnow compare those numbers to gaymers.

AMD is retarded if they release 16 core part together with 12c
Should go for
6c 8c 10c 12c setup

honestly not too bad

Attached: Untitled.png (391x374, 26K)

>leaked cascade lake design document

Attached: Intel_Cascade_Lake_AP_graphic_800x.png (643x395, 120K)

A chiplet is eight cores, which means any 10-core CPU needs two of them. But there's no good way to split that up. One of the chiplets would have to have five or fewer cores on it. They probably don't want to bother with that because their yields are good enough that they won't have a lot of chiplets with three bum cores, while they will have plenty with one or two (making it still useful and economical to standardize on a six-core chiplet version)

The other options are even worse, waste half a chiplet with four cores and pair it with a six?

I think that he meant that they should've gone with 6 core chiplet.
Wondering if there's going to me a single chiplet CPUs. The one at CES had only one die besides I/O. 8cores like that could be pretty nice for gaming.

new into amd ?
2 6 core = 12core at east XX% will have 1/2 cores not as fast or defective and just disable them make 10 core ez
I miss defective days
where u could turn triple core into quad
ah the good days

Are you stupid? A 6-core Zen 2 is still an 8 core chiplet with two cores disabled. Every Zen 2 processor uses the exact same chiplet.

MOAR GLUED RINGBUS POZZED HOUSEFIRES

try skylake for first gen Ryzen. we're approaching 3rd gen now with real IPC improvements
>MUH LATENCY
>MUH SINGLE COPE PERFORMANCE
piss off shekelchaser

>BGA
>j*wish glue technology
>350W per glued part
wew what a product, intel.

AMD is pathetic and should focus on single-core performance.
Nobody cares about 16 cores/32 threads on consumer hardware.

If this is real, and IPC is >= Zen+, Intel is done.

Attached: original_163738943.jpg (2048x899, 264K)

why not both at the same time without any compromises? no? well how could you know, after years and years of Stockholm syndrome with Intel. come back when you post from your Meteorcrater Lake on another NEW chipset, Chaim

AMD had been taunting >10% IPC improvement for each new generation for some time, but i didn't save the PR material on my phone to post.

>Nobody cares about 16 cores/32 threads on consumer hardware.
That's easily fixable with some marketing and memes.
>AMD is pathetic and should focus on single-core performance.
It's unironically just cheaper to separate dies and show in more cores. With the same base die for all CPUs, separated I/O and no iGPU, pretty much nothing is wasted. Performance is just sacrificed for better price here.

Don't delude yourself that they will be anywhere near skylake. Best they are doing nowadays is 6 year old haswell 4770k in even ryzen supported titles.
Also how you lose like 10-30% performance if you don't go fastest memory of 3400 or 3200.

Yes those improvements like they say is in controlled environment with selected applications. If you read any of those reports past titles they say that. Even AMD doesn't want to sugarcoat their embarassment.

you really do underestimate tech youtubers and muh gaming benchmark charts do you?
Everyone knows CPU is as good as much FPS it can get on 1080p in game noone plays

שחררו את פלסטין Free Palestine תקרית_ליברטי The USS Liberty Incident of 1967 עברית Zionism קבאל של בנקאים יהודים Cabal of Jewish Bankers כיבוש המזרח התיכון Occupation of the Middle East ארסנל גרעיני ישראלי Israeli Nuclear Arsenal השדולה הפרו-ישראלית בארה"ב Pro-Israel Lobby in the USA זכויות אדם גוי Gentile Rights פשעי מלחמה של הצבא הישראלי Israeli Army War crimes שואה מזויפת Fake Holocaust חתרנות של האירופים Subversion of Europeans סוחר שמח Happy Merchant

not really just see the microcenter graphs
the best selling cpus are always the mid tier for intel 8500/8600
same for amd also

>Nobody cares about 16 cores/32 threads on consumer hardware.

yeah 2 years ago the argument was
>nobody cares about 6 and 8 cores since 7700k exists
then it went
>nobody cares about hedt
now its
>nobody cares about cores

hmm

Not sure why you would call me a Jew when I refuse to accept mediocrity from AMD.

not really the zen+ bullshit never came from amd just like the usual bullshit we are seeing yet again from the usual falseflaggers(tweaktown and wccftech)

add his post to the doesn't matter category

So even high end x470 boards won't support the 3850X is it is real. That was to be expected to be honest.

You don't know what IPC is
You're just as dumb as the other Intel shill

if the 16c is supported then the rest will to

high end one x370/470 certainly will, the VRMs are good enough at least for stock. but mobomakers are dumb cunts and want you to buy a newer board so they don't make a new BIOS that supports those CPUs

Please explain why the average consumer should care about more cores instead of caring about more performance per core

Based but you're getting banned, I'm afraid.
We'll know for sure soon.

i don't want to JUST WAIT™ anymore

Please explain why you think the average consumer should care about 400 fps in 720p CSGO instead of the ability to multitask without any issues. Single core performance of any CPU in the past 10 years is good enough for average consumers you fucking retard

because ryzen in a clock to clock scenario is always ahead
what ryzen lacks is frequency not ipc
and as far as multicore perf goes they are far ahead

Attached: AMD-Ryzen-Vs-Kabylake-Single-THreaded-Test-IPC.png (493x280, 20K)

shut up gaymer, this is not /v/

>average user
>heavy multitasking requiring 16 cores

that 32mb L3.

AMD made 6 core and 8 core affordable and mainstream so kill yourself retard. Intel kikes really grasping for straws here.

x370 doesn't have this list. fuck you gigabyte.

>AMD made 6 core and 8 core affordable and mainstream
Yes and now they should improve performance per core with their new 7nm instead of slapping even more cores that won't get used in any common scenario for an average user

average user should buy 3600x for $200.

see You don't know what you're talking about. Fuck off back to /v/.

Too bad those cherry-picked benchmarks don't translate into real world performance...

Attached: ryzen-fail-aida.png (2160x1076, 475K)

I'm sure the average consumer would notice the latency when browsing Facebook and watching YouTube videos.
Intel really needs to give you guys a new script.

>AMD 8 cores barely pulls ahead of Incel 4 cores
How embarassing

>closed-source chinkmark using 8080 tier instructions

this test is so fucked....not only can different MB change the score on the shitty aida64 test, you can get drastically different results between two tests ON THE SAME SYSTEM

You guys are really trying hard today. Got scared because of Incel's stock prices?

When was this bios revision up?
gigabyte.com/us/Motherboard/X470-AORUS-GAMING-7-WIFI-rev-10#support-cpu

>AMD
>"if the rumors are true Intel is dead"
Sounds familiar.
Prepare for yet another disappointment, AMDrones

it's a ''leak'' from a week before CES

This, but it also suffers from latency between the dies, this is the reason why Intel performs better in scenarios like light chink games running at 200+ fps.

Latency between dies only matters when small chunks of data get jogged between cores or thread migration between cores, which is actually either a compiler problem or OS scheduler problem. Windows scheduler goes full retard as made evident by Threadripper. It's a case of Fine Wine(tm) again but people will just call me a shill.

what are the chances for computex announcement, bioses came out quite early. so, july for 3000 series is quite high?
did EPYC 2 came out or not?

They are announcing at computex and you can expect launch by june end or july.

cpu.userbenchmark.com/Compare/Intel-Core-i5-7500-vs-AMD-Ryzen-7-1700X/3648vs3915
cpubenchmark.net/compare/AMD-Ryzen-7-1700X-vs-Intel-i5-7500/2969vs2910

ohnononononononononon

keep seething intelaviv fan

Look, everyone, we have one datapoint, and that is the 8 core demo on cinebench.
Basic math tells me my 2700x would need 4.6Ghz, all cores, to do that.
And that fucker did it at half the 9900k's power.
So I wouldn't be worried about the 16 cores part's clockspeeds.

Truth is Grosoft is angry because mainstream cpus are now server tier hardware.
I heard they want to move to a per core licensing scheme.

It should already be out, to be honest.
Initial zen roadmap was a new cpu a year.

I don't necessarily believe that anything other than the 8 core Zen 2 will be reaching 9900k frequencies, but depending on the price it won't matter if it's close enough

if the current CPUs can reach 4.3GHz, it won't be hard for the new ones to reach 5.0GHz with some overclocking or even just with turbo boost

It' a total unknown, because it's a different process, on a node from another company.
As I said we have one single datapoint. So maybe it won't 5ghz, but still beat the crap out of the 9900k on IPC alone.

>Magic AMD chips may reach 5.0Ghz in 2019
>When overclockers could already get their 8 cores FX chips well over 8 Ghz years ago

>I don't know how IPC works

So it looks like zen3 will have stacked memory on top of the I/O die.
How fucked is Intel, really? because it looks like zen2 will already have them cry a river.

Attached: zen3.png (1787x963, 448K)

>It' a total unknown, because it's a different process, on a node from another company.
but we already know they won't end up like intel who had to actually use lower clocks on their new process, and not only AMD said that the clocks will get higher but TSMC also confirmed it

Disparity in height between the dies would prevent the use of any conventional cooler. In MCMs having equal Z height is super crucial.
It is possible however. All stacked memory out there tends to run pretty warm. The insulating effect of having a 95c package on top of the I/O would present a significant design challenge. Even if both have tame power consumption the bottom of the stack is always going to be a hot spot.

Inb4 you don't even need Ram anymore.

Intel are working on similar things to what AMD are, they're only really fucked if they have the same problems with their next node as 10nm.

Or, you know they make a custom heat dissipation system, aka, the thing you delid on Intel.

Higher, but how much higher?
Something like 10% higher frequences is a decent improvement, but it won't get to them to the magic 5ghz territory. Not that it really matters, people put a lot of emphasis on the 5ghz number but with +10% freq gain and 5-10% IPC they're going to be pretty much where intel is right now, likely for substantially less than what intel asks people to pay for that performance.

Think about it.
You have the I/O and memory higher than core logic, and the core logic is the vital component that needs the best contact for cooling.
Making the IHS thicker to put more material on top of the cores would lead to worse thermal transfer. You'd end up with more self heating as a result.

Ok, you don't know the first thing about heat transfer.
While what you're saying seems logical, it's not. I don't blame you.

M8, learn the basics of heat transfer and insulation. Just because a material has good potential for thermal conductance doesn't mean it has advantageous characteristics at all thicknesses.

Whatever, dude.
If what you said were true, extreme overclockers would just stick their coolers directly on-die.
Well, they don't.

People do just that, genius. The only reason why people keep the IHS in place is to prevent damage to the die.
The IHS is an inconvenience. It provides uniformity and protection, but ultimately hampers thermal transfer to the cooler. Direct die cooling provides the best thermal performance.

Its hilarious that you even attempted to talk out of your ass without knowing the basics of this.

What you don't understand is that the heat spreader is just that.
It spreads the heat on a plane area the cooler can cool.
Gains from directly applying a cooler are marginal.
But it's retarded anyway. Zen3 won't have any problem cooling itself.

What do you recommend I get if im trying to replace my 4690k @4.3ghz? Something with 8C/16T that is at least 4ghz? I don't think I'll need anymore than 16 threads at 4ghz for gaming and such.

Look like the 3700x will be the processor to go for.
I get that you don't need extra threads, but from what we've got, this will come with the extra frequency.
Going for 8 cores would be a mistake as you'd be a yield whore in case of OC attempts.

The heat spreader is an imperfect surface, its an intermediary, and adding intermediaries reduces efficiency. Again proving you're out of depth here, kiddo.
The gains from direct die cooling are substantial enough that its still done to this day. Just pulling out more nonsense from your ass won't make you any less retarded, user.

No... it's common software design for the process to spawn dozens of worker threads for processing and submit results back to a handler thread. Schedulers can only help alleviate rampant thread migration between core it wont stop fairly standard, good even, software design that isnt designed to minimize latencied for one particular architecture. Which is why ryzen still has blind spots against similar Intel processors

You're one of those guys that thinks watercooling helps cool your cpu, aren't you?

I fell for the AyyMD meme, I wont make that mistake again.

Only a fool bets against Jews in the 21st Century.

>oh shit, I made an ass of myself
>better deflect instead of walking away
lol

That meme overclocked did a test and found even when using the inferior liquid metal TIM a delidded 2700x achieves noticeably better cooling