This can't be correct ... right?

What part of "single core" do you not understand?

As I said 64 core epycs are confirmed and probably not pulling much over 200W compared to that 50% more cores is boring

12nm --> 7nm

You're not dealing with the average CPU anymore.
It has become a legend.
The legend that you fear.

Yes

Jim, Amd said 25% performance gains per watt themselves. Your made up numbers are retarded and senseless

Somebody doesn't know a good joke when they see one.

Oh. Were you ironically shilling for amd? I cant tell anymore by the shit that gets posted here unironically

It's possible. 7nm is a huge jump, but this fag is right

Why did you crop out AmdroneTV?

It most certainly looks too good to be true. Wait and see, but this will nonetheless make me wait for an upgrade until the CPUs are actually out. No hurry for me and if this is true it'll definitely be worth the wait.

OF COURSE IT'S A LIE

ROME
IS
A
LIEEEEEEEEEEEEEEEEEE

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

My view on it is that there a good chance that the numbers are complete shit, but there's also a possibility that they're not since it's not the monolithic style we're used to comparing. I'm not buying for awhile one way or another. Just here to make light of all the arguing about stuff nobody really knows about.

Its a combination of 7nm gains and a new aggressive binning of the chips. There were guys 3-4 months ago telling us that these parts were coming and the numbers they gave us were 100-200mhz lower than this so apparently AMD's end results were even better than expected.

Early engineering samples of 7nm zen@ already hit 4.5ghz so 5.1ghz is not suprising at all.

The surprising thing is that they were able to improve binning enough to allow enough 4.9-5.1ghz chips to package and sell as a standard SKU.

It's been confirmed fake several times already. AMDrones are pathetic.

Attached: tenor.gif (498x278, 1.69M)

I just don't like adored. He is the alex jones of tech consumerism. Its gross to me

ITs not a die shrink. ITs a die shrink, and completely changed architecture AND a new modular chiplet design.

They are fabbing 8core chiplets with all I/O functions removed giving them tiny 70mm2 dies with massive yields and connecting them to an I/O chip fabbed on 14nm

Basically, these SKU's are 6x0 for Ryzen3, 8x0 and 4x4 for Ryzen5, 6x6 for Ryzen7 and 8x8 for Ryzen9 this modular design allows new binning methods resulting much higher standard clocks that normal binning of monolithic dies would give you.

The big issue would have been latency but AMD solved that by improving their infinity fabric and simultaneously doubling the size of OP cache to reduce transfers from memory

Its a brilliant design and its going the rekt intel's shit.

I really hope you don't believe that

Yeah yeah until it turns out to be pretty accurate and you just want it to be untrue because you just got an 8700K last week

He prides himself on trying to predict (poorly) the future of tech releases. Makes me sick.

He said 8400 wouldnt run full turbo speed on b360. That was proven false.

Just take a look at the guy's youtube channel. It's nothing but hate for intel and nvidia and lots of ass kissing for amd.

Attached: au9O3cB.png (582x482, 325K)

maybe at 5nm it will be like this

They're still 180w parts.

He seems to be right a lot, and makes it pretty clear when he's guessing. That said, brand loyalists can still get fucked. Those guys only serve to further stagnate the tech world. I just want bang for my buck. If AMD pulls it off, usually the consumer is who wins since Intel has the money to bounce back fast.

Nah, AMD is going with single chipset + I/O chip + IGP (Probably little Vega not Navi) for their Zen2-based Ryzen line-up.
They just need to scale up clockspeed to close the performance gap with regular Intel SKUs. They aren't going to be cannibalizing their Threadripper 1/2 SKUs just yet.

Navi is being used up for AMD's semi-integration contracts (gaming consoles and Apple) until yields improve enough for upgrading IGP with desktop/laptop Zen2+/Zen3.

>/v/ tourist doesn't understand what a new process from a new fab is
Me buy 9900K for my games because it are most expensive and best. :)

9900K is pure meh-tier bullshit (Intel should just called it Core i9-9900K Extreme Edition. 9700K, 9600K and 8700K are almost as fast and doesn't cost an arm and leg and are at least obtainable.

Fuck off. Node shrink and higher clocks you can still get the same tdp With a 70% shrink.

>16 cores running at 4Ghz+

Isn't this going run hot?

Lmao, get ready to be ass blasted on release then if you think this is anything close to what you'll get in reality.

Yeah but I don't have to believe a bunch of fake leaks from a dodgy youtuber to want techology to progress. Nor do i want to encourage others to speculate and leak nonsense for ever tech product on the horizon. For ever real leak there will be a magnitude of fake ones. That doesnt help the consumer

It's most likely correct. Chiplets plus a die shrink as opposed to just one or the other can easily give these performance increases

14nm -> 7nm means the die shrinks in size by 75%. It's probably easier to cool tiny chiplets than massive monolithic dies.

All of this speculation is kind of pointless, though, because we're going to get a tech demo at CES that should provide useful data. I'd put even money on getting clock speed and TDP info at CES, and I expect a full launch sometime in Q2.

>AMD brainlets keep using terms like architecture, market share, chiplets, etc. without even knowing what it means
It's kind of cute desu

>adored
>most likely correct
>about leaks
>with no source

Attached: w352345734256.gif (250x220, 1.58M)

To each their own. I won't judge if you don't care for him.

Amd is at 12nm currently not 14nm and youre lecturing him about die shrinks.

Attached: 1528365809921.jpg (494x488, 50K)

why do indians love amd so much?

The numbers AMD is providing is from Zen1 to Zen2, not Zen+ to Zen2. Zen+ is a half-node shrink on the same uArch with minor fixes, whilst Zen2 is a full node shrink AND is a complete redesign. If we're going to have serious apples to apples comparison and not comparing oranges to baboons.

12nm is a half node, not a full node.

>They aren't going to be cannibalizing their Threadripper 1/2 SKUs just yet.

They will because they need to create a crash wave in the market

They only have one chance and a 1-2- year window of opportunity to gain the upperhand over intel with market share and mindshare. It was dumb luck that intel completely shit the bed with their 10nm node Now Intel is rushing development off their 7nm node to regain the edge. That means 2 years MAYBE 3 at the most before intel is back and the same amount of time for them to rush their own chiplet designs through R&D.

AMD needs to stick the knife right through Intel's heart and they need to do it now. A 6-8-12-16 core AM4 SKU, a 24-32-48core Threadripper SKU and a 48-64core Rome EPYC SKU is exactly what they need to crash the intel market share with no survivors.

Desktop/laptop doesn't care about getting 12/16 cores CPUs though. There's no killer mainstream app that makes quad-core CPUs woefully inadequate let alone six and eight-core units. Clockspeed is still king in this market. OEMs care about ubitiqous iGPU which is why Intel still holds onto their massive marketshare here.
AMD is going to grab mainstream mindshare by making a cheap all-in-solutions for their OEM customers that yields competitive performance and power consumption to Intel solutions.
7nm process should allow AMD to the bridge the clockspeed gap while CPUs are dirt cheap to produce unlike Intel's larger monolithic 14nm+++ chips which require nearly perfect silicon to be viable and this destroys yields = costs more to make.

AMD already has the server market in the bag with Rome. Threadripper already disrupted the fuck out of the HEDT/Workstation market.

Consumers are mindless dipshits.

All they need to do is advertise 6 cores in the price segment where intel is only selling 2-4 cores and people will buy it because of more coars.

When Intels i7 top tier is 6 cores 5 ghz and AMD is able to plaster Ryzen7 12 cores 5 ghz everywhere people will buy that even if they dont need it.

AMD needs to play the game the same way intel played against them.

>All they need to do is advertise 6 cores in the price segment where intel is only selling 2-4 cores and people will buy it because of more coars
Newfaggot detected. If this was even remotely true AMD would've cleaned up this decade.

I'd be happy with an 8c/16 3700x at 4.0 Ghz base and 4.7 Ghz turbo, with some minor IPC improvements. Would be one hell of a chip and a nice improvement over the already impressive 2700x.

9900k only has raw clockspeed at the moment.

I doubt they'd make 6c/12t the low-tier for the Ryzen 3000 series. That's just ridiculous

The 3600 will be 12 cores for 200 dollars and the 3700 16 cores for 300. :)

The more cores strategy failed because bulldozer cores were extremely weak.

Ryzen is almost at performance parity with Intel and zen2 on 7nm will match and potentially move past them in single-thread.

When the core-core performance is almost identical then yeah more cores will work as a marketing strategy.

Gaymerz will probably buy 8c/16t CPUs for a while now since current and upcoming consoles are going for 8c. I can see MOAR COARZ working for HEDT/Server, but not so much gaymen.

Wow youre idiot
Many times told but don't listen,of trying to doing dirty things here you fuker
intel is better processing with coffee lake,none need for else with it

do the neeful and buy intel ples goys

amd makes bad cpus

and here, the npc reply.

intel's must be atrocious, then.

>9900k only has raw clockspeed at the moment.
that i9 is a great CPU, what kills it is the price

Attached: lyl.png (342x356, 204K)

>as is any other overpriced cpu

Do you understand what TDP measures? It is an indication of the kind of cooling solution requires, not of power consumption per se.

This is important to understand because the chiplet design of the 3700x would give it a lower thermal density than the 2700x, meaning it could use some that more power without increasing TDP.

Next, TSMC's 7nm HP is either 1.25x performance at ISO power, or 0.5x power at ISO performance.
The 2700x is 8C/16T at 3.7ghz/4.3ghz for 105w
The 3700x is 12C/24T at 4.2ghz/5.0ghz for 105w

That's 50% more cores, 13.5% higher base clocks, 16% higher boost clocks, and the same TDP. When you consider that some proportion of the 7nm advantage was spent on reduced power consumption, and you consider the much lower thermal density of the chiplet design, it's not totally crazy that it would.be possible. That said, I'd expect the TDP to be slightly higher than 105w, but not enormously.

>Anyone who knows the history of die shrinks or adoredtv knows its a lie
So the fact that Vega's clocks increased by 20% purely from a node shrink to 7nm HP is a lie then? Because AMD announced that at New Horizon.

>Amd said 25% performance gains per watt themselves
Gee, I'm pretty sure that's now that TSMC's claim of 0.5x power at ISO performance means.

>It's been confirmed fake several times already.
When did AMD issue a statement to that effect?

>Nah, AMD is going with single chipset + I/O chip + IGP (Probably little Vega not Navi) for their Zen2-based Ryzen line-up.
Why would they do some retarded shit like using Vega chips instead of Navi when using Navi allows them to recover/bin more Navi chips? That's pissing away all kinds of money.

Fake af. I want to believe in a 6-core apu at least.

>what are memory channels

The 2600 is around £160 and 2600X slightly more. This creates a dilemma for AMD. I would expect they will price a 6/12 SKU with better clocks at a slightly higher price point so they do not eat into the previous product sales. At least until they get rid of the older line. The 8/16 core parts will be slightly above the 2700 and 2700X. Again with higher clocks. The 3700X being capable of 5Ghz with overclocking and dependent on the silicon lottery. There will probably be a 12/24 part but I doubt it will reach anywhere near 5Ghz out of the box.
16/32 is most likely just pie in the sky.

I want it to be true but I see the above as what really will happen. But it depends if AMD really lucked out on Zen 2 and how aggressive they go in against Intel. Whatever happens. The IPC increase and general clock speed uplift will still be worth it.

Just wait for next month. I personally just want an APU so that I can finally get rid of my old Core 2 Duo.

If they were to price zen 2 so advantageously it'd encourage zen/+ owners to switch too. The tides are turning away from intel and they know it. No matter where you go on the internet you'll see people praise amd for ryzen. AMD doesn't have to worry about money as much anymore so they can sacrifice profits for gaining mindshare. If anything, I'd say that the reduced cost of chiplets and massive yields on the i/o dies more than justify the pricing

Intel shitters are mad Zen has almost perfect yields. Even atrociously bad dies can still be salvaged, even if only ONE core is functional AMD could use two Zen chiplets with only one core functional each and sell 2c/4t cpus. Yields are the least of their problems.

Attached: 1536694146733.png (1200x800, 164K)

>t. dumb shit
glofo 12nm is rebranded samsung 14nm, not a die shrink

CES can't come soon enough. Tears from shills are going to be glorious

why would literal shills be emotionally invested?

>It's been confirmed fake several times already.
Except it hasn't. The spreadsheet "leak" was confirmed fake, which was completely different to this one. How shocking that an animeposter can't comprehend a simple situation.

It is also a blast furnace at maximum load.

Navi has barely taped out. Anything produced is being allocated towards semi-integrated customers ala Apple, Sony and Microsoft via contracts. The remainder will be used for AMD RTG's discrete line-up.

OTOH, there's a crapload of little Vega chips that haven't been sold yet (stuff that Apple didn't want). AMD needs to clear that inventory out and it is still using GF's 14nm process (keeps the happy). It is just cheaper and quicker to implement then waiting until Navi production ramps up and yields improve. Navi-based IGP will likely be used in Zen2+ or Zen3 when 7nm production improves and yields are good enough.

What kills it is its own EVERYTHING.

Attached: 1544307635963.png (1327x1222, 69K)

Already debunked by tech experts. But please get hyped up for it, it will be much funnier when the inevitable news hits

jesus

>Consumers are mindless dipshits.
>All they need to do is advertise 6 cores in the price segment where intel is only selling 2-4 cores and people will buy it because of more coars.
>What is the FX8350 VS 2500K phenomenon

>some aussie fag on jewtube
>tech experts

I'm glad that at least one user here understands that TDP is not power draw (though obviously there is a correlation). That said I suspect those turbo figures are single core which makes the leaks a lot more believable when you consider in such a scenario (and this applies to Intel as well) that when one core is going so high clockwise the rest of the cpu is going to be sitting close to idle.

2.81W per core then, Multiply that by 12 and you get 33.6W. It isn't really that doubtful that you'd pull ~3.1x the power for ~2.5x the frequency.

I don't find this to be unlikely at all.

>I don't find this to be unlikely at all.
Congrats, this just shows your own ignorance on the subject matter.

Wait for ANYTHING official before starting the circle jerk, starting this early is 100% going to lead to disappointed AMDrones on launch.

5.1GHz might be pushing it, but I really don't think 4.9GHz is unlikely. We've got engineering samples at 4.5GHz already, and we all know that ES tend to have lower frequency than the final product.

>just wait for 10nm

Attached: 1514937277489.jpg (626x657, 81K)

They are skipping 10nm on all but NUC's. Going straight to 7nm in 2020.

Most intel shills have given up on 10nm and are focusing on 7nm, in all likelihood, there will only be a single generation of 10nm next year, and by 2020/21 they'll be on 7nm.

no no no. it's 10nm DOESN'T MATTER!
pic related
>anandtech.com/show/13683/intel-euvenabled-7nm-process-tech-is-on-track

Attached: 7nm.png (692x546, 569K)

inb4 in 2021 7nm is still going to be on track

Attached: on track.jpg (660x491, 46K)

To be fair, intel probably has the most pull when it comes to actually getting the EUV lithography machines in their fabs. Intel has money for days and fabs out the ass ready to go, ASML (who manufacturers the EUV machines) is probably willing to ship as many units as intel wants, whereas a smaller customer may get told to wait a few months since intel is taking up the machines available.