Zen "3" won't be ACTUAL 3, but Zen 2+. Also, AMD already has Zen 5 added to the full pipeline

Only Zen 4 was known to be existing on the roadmap thus far, now we got confirmed info that Zen pipeline is going to continue even further, at least to 5.

Samsung hit 3nm? Damn I didn't know that. I thought at 5nm quantum tunnelling became a prominent issue, and I'll admit I don't know anything about AMD's plans for Zen 3. I guess I'm just used to intel's refresh it again mindset so I assumed it was a refresh.

They can't just hop off the Jim Keller Zen train and hope to get success

michael clark spilled the beans on zen 5 months ago too.

We already new Zen 3 will be 7 nm.

>Samsung hit 3nm?
Not "hit", but they've already figured out how to make it work feasibly enough. The thing is - 3nm is LITERALLY just 3 atoms and you CAN'T go smaller than that even with EUV, so this is where it ends for ALL, EVERYONE, regardless of quality of fabs, resources poured in, and machinery available. At 3nm, EVERYONE will hit the ABSOLUTE wall which could be alleviated ONLY with new materials and tech. Gladly, 3nm won't be coming onto the scene anytime earlier than 2028 at the very least, so entire semiconductor making industry OF THE PLANET has at least 9~10 years more to go. After that, if solution won't be found out...humanity in general is FUCKED.

Quantum tunneling being an issue is something only reddit tier popsci trash tech blogs opine about.
Samsung continuing their plans with 3nm GAA isn't even some big breakthrough. Processes are always designed to cope with current leakage, most notably this is why FinFETs came to be after decades of planar gates, and why GAA will replace fins in years to come.

Attached: different-transistor-topologies.jpg (1195x894, 479K)

There is probably a solution out there but it's being held back. Like a lot of innovations.

Keller only created design papers enough to last up until Zen 3 at the most - everything else beyond that is a completely new thing AMD's R&D developing by themselves, without relying on Keller's works. And THREADRIPPER even in it's very first iteration was NOT foreseen by Keller, it was coined and developed STRICTLY in-house as a SIDE project in R&D's spare time, lul.

That's not the point, read more carefully. The point is that "3" won't be a big jump, like Zen 2 is to Zen/Zen+, but going to be incremental update for Zen 2, Zen 2+ essentially. It's literally tick-tock, but from AMD and much better than what Inturd tried to do, since even such an "incremental, insignificant" update will yield at least 10% more overall gains than the previous CPU hierarchy before it (Inturd's shit was stagnating in the "2~5%" territory for years).

Attached: 1540220717862.png (1384x725, 138K)

Graphene nanotubes seem as most feasible solution thus far, but shit's hard as fuck to make stable (Graphene is SUPER-FUCKING-REACTIVE on the EDGES of it's structural net's layer/stratum) AND yields aren't going to be any good at all (less than 50%), making shit build with that tech VERY expensive even at farther end of adoption cycle. And 100% Quantum Computers in non-scientific consumerist market segments are absolutely out of question until 2048 at the very earliest.

can i get some source on all of this?

It's not 12%, though. Just 2~8% (if perfect lottery and LN2 OC to the roof).

did they even manage to fabricate some transistor out of graphene nanotubes to begin with?

wccftech should be filtered. Nothing here is new, and when "new information" comes from that site it's very frequently false.

theinquirer.net/inquirer/news/3016760/amd-threadripper-was-developed-by-engineers-in-their-spare-time

Zen 3 is Zen 3.
What you mean is the Ryzen generation, don't mess things up you fucking brainlet.
Fucking 70 IQ.

8700k and 2700x cost roughly the same though. 8700k can be had on sale for 280 for like most of last year.

just let intelfags have that one, it also implies the 9900k is way cheaper than it actually is

>8700k and 2700x cost roughly the same
...the fuck are you smoking, nigger?

>Fucking 70 IQ
My deepest condolences to your family/relatives.

On the upside I don't have to feel bad about buying Zen 2 next year with Zen 3 already announced if it's only going to be a modest refresh anyway. Not that worried about Zen 2 efficiency tbqh, the jump to 7 nm alone will decrease power demand considerably. I mean, look at what happened to the TDP of Intel processors when they went from 20 nm to 14 nm.

microcenter is selling the 2700x for 265 right now, and as i've said 8700k is around 280 on sale most of the year. what is hard to understand?

What really will fuck with people is that zen won't be the 2xxx series but the 3xxx series. Who's dumb idea was that shit?

>I don't have to feel bad about buying Zen 2
You shouldn't any if you buy Zen+ OR even just gen1 Zen, because PERFECT BACKWARDS AND FORWARD COMPATIBILITY. EZ A DN COMFY AS FUCK.

>if it's only going to be a modest refresh anyway
I hardly can call "at 10%" a "modest" and a "refresh". It's just very obvious that at this point AMD is just sooo CONFIDENT in themselves that they're SANDBAGGING THE FLYING FUCKS OUT of it all. Again - "FX to gen1 Zen" was supposed to be ~40%...and they managed to get 65%, lel.

>currytech
kys

>microcenter is selling the 2700x for 265 right now
...Amazon had a Black Friday deal going with Full Retail Box packaged 2700X for ~160$ just a couple days ago, mang.

>look what happened to the TDP of Intel processors when they went to 14 nm

Attached: 1543241975159.png (1327x1222, 69K)

>Gen 1 Zen essentially clusters of mobile ARM cores
What the fuck are you going on about you teenage piece of trash
>when in reality turned out to be more than 65%...turned out to be 12-14%
What the fuck are you going on about you retarded small town hick
>graphene nanotube transistors
Oh, you're just retarded.

everyone knew, you idiot
we're about to get ryzen 3 tho, which is zen 2
while we currently have zen 1+ (ryzen 2)

>because PERFECT BACKWARDS AND FORWARD COMPATIBILITY
What's the point though? It's not like I'm going to downgrade anything, and I'm likely not going to upgrade a CPU or motherboard without the other because you usually match the performance potential of the two, and if your CPU is outdated so is your chipset.

|
|>
|
|
|

>soldering is for fagits

Why is that the only response you little shithead niggers have when someone calls you out ?

You don't know half the shit you're talking about.

Attached: 1513824452944.jpg (544x400, 51K)

>if your CPU is outdated so is your chipset
Typical brainwashed Intelaviv milking cow, ladies and gentlemen. This creatura never knew the feeling of FREEDOM which AMD's approach to platforms was providing for more than a decade (AM2, 3/3+, and now AM4 + THREADRIPPER).

>So, Zen "3" is actually Zen 2+
That was sort of expected after Zen+, though. It's obvious they'll want 7nm node optimizations first before doing an architectural overhaul again.

>The thing is - 3nm is LITERALLY just 3 atoms and you CAN'T go smaller
A silicon bond is 0.235nm
Imagine the kind of bs your post has next

YES, YESSS, GOOD GOY.
A SOCKET CHANGE A YEAR KEEPS GOYIM IN FEAR.

Attached: 3we4ugyuit87.jpg (491x491, 40K)

>"FX to gen1 Zen" was supposed to be ~40%...and they managed to get 65%
Jesus Christ, I'm gonna get assraped by performance once I finally migrate from my old trusty FX to Zen 2/2+/3 (probably 2+ though)

Jim keller didn't touch the march of zen. He worked on different aspects of the cpu.
The zen design team is still working for amd

Why wait? You can bend over and fully take it all down to the hilt with no lube whatsoever at all if you build on 2600X/2700X or low/mid tier THREADRIPPER+ right now.

>Jim keller didn't touch the march of zen. He worked on different aspects of the cpu.
Ye, pretty sure he worked on Infinity Fabric aka Glue that intel now also uses since they hired him

>aka Glue that Intel now also uses
Inturd doesn't have and can't Infinity Fabric. Theirs is JUST™ GLUE with...DOUBLE THE BINGBUSES - .

>Why wait?
Because I have no need for an upgrade as of now. I'm not even sure I will need it at Zen 2+/3 time so I'm speaking purely hypothetically really.
>not a gamer
>FX 6300 absolutely satisfactory as my daily driver for web browsing, programming (mainly android studio) and really, really casual gaming
But can't know how the world looks by 2-3 years, so I may very well need it by then.

Every four years. And after four years compatibility really doesn't matter anymore, you're going to need a new chipset for that new processor.

Even if keller manages to make something for intel, intel won't be able to make it.
Amd has almost 10 years of r&d in packaging, interposers and interconnects that are used on zen,fury and vega.
They used to have their own packaging line iirc.
Intel will have to do giant leaps to make a threadripperripper

Yay it's so great intel has kept 1151 so long... too bad each new cpu generation release requires a new iteration of it and existing motherboards are as good as trash.

so when the fuck is it coming out, who cares about the name

You should expect official product reveals in February and store launch in March just like the first Zen series.

Anyone knows if 3000 series APUs will be 7nm or 12nm refresh? I'd really fancy a 6-8 core APU with something akin to Vega 11...

They're 12nm refreshes. The 7nm APU comes after the 7nm desktop chips.

FUCK OFF RETARD

>inb4 that desperate faggot shitposter shows up again

Attached: 1534326485713.png (1824x1026, 431K)

Aww shiet. HODLing until 4000 APUs it is, then.

I spent all night looking for that deal at my local best buy and microcenter for that deal but settled on a 2600 for $149 instead

Attached: Michael-Scott-Sad-Face-Funny-Picture.jpg (982x1024, 67K)

Dunning Kruger effect my dude.

1 atom is 1 nm
Wow, how convenient. I imagine your brain is much smaller than that.

1nm is not 1 atom, however at 3nm lithography 1 gate is 3 atoms only in size, you dumb fuck.

Just wait for ice lake with moar cache and intel's new uarch, in like 3 years finally they will have a response to Threadripper

Attached: 1542027477262.png (807x745, 205K)

If you buy AMD you're anti-semitic and a totally bad person, you should support feminist frequency and white genocide.

Attached: 1492139297054.jpg (752x548, 282K)

Burning your house down is a good thing!

wtf

Attached: aaahhhh.webm (439x371, 203K)

Zen 3 will be 7nm+, Zen4 will most likely be on 5nm, but that should be somewhere in 2021 or so.

Attached: 2018-11-06_23-20-09.png (2559x621, 243K)

>wtf
Nothing, really. Just this:
youtube.com/watch?v=s1Ww2vNAjN0
and this
youtube.com/watch?v=RbuQ4SK7wRA

Man, is zen 2 going to fix pic related? I know the 2700X OC'd to 4.3 GHz gets like 2% close to i7-8700K performance but only if it uses cream of the crop CL14 3400MHz RAM.

Attached: 6ZSDax7.png (767x441, 45K)

Zen 2+ = 2020, Zen 3 = 2022, Zen 4 = 2024. There also might be Zen 3+ in-between as well.

>2700X OC'd to 4.3 GHz gets like 2% close to i7-8700K
It's 2% at stock + Turbo, you fucking cuckshit.

n-nnoo...it was OUR turn to hit 5ghz!

Attached: 1497878953615.jpg (552x661, 71K)

see

>Tick-tock and Moore's Law lives still, in AMD,
To some extent, but AMD is effectively doing a tick and tock all at once with Zen2.

No I mean literally forcing all cores to operate at 4.3 GHz for the 2700X. Forcing all cores on the i7-8700K to operate at 4.7 GHz causes MASSIVE frame stuttering even with high end cooling and deliding because most motherboards don't support its "95W" TDP. The problem is exponentially worse with the i9-9900K even with Z motherboards. Realistically you're only going to be able to run the i7-8700K with the stock settings or locking all cores to 4.4 GHz.

I'm an intelfag myself but I fucking hate how everyone here keeps pretending the i7-8700K can hit 5 GHz on air.

Attached: 1511396778699.jpg (900x900, 160K)

>is Zen 2 going to fix latency
20~70. Gen1 Zen was 90~170, Zen+ is 70~140. Inturd is 40~70 on general.

>I mean literally forcing all cores to operate at 4.3 GHz for the 2700X
...why would even need that any, to begin with? Like, there's no point so far. "MUH GHz" is a meme race since Zen came into the scene.

>Zen "3" won't be ACTUAL 3, but Zen 2+.
How surprising, OP is a faggot. The article makes no such claim. The reason why the architecture on the Ryzen 2XXX series was named Zen+ rather than Zen 2 was because the microarchitecture itself saw very little changes. What this article talks about is the way they'll be using the benefits from the fabrication updates, it doesn't say anything at all about the microarchitecture. It could be a small change, it could be a big upgrade, we don't know.

Attached: faggot-17.jpg (705x435, 230K)

>The article makes no such claim
It's easily deductible by the incremental, minor update in performance/power envelope, you dumb shit.

Ryzens 3xxx better be cheap or else ...

Attached: arthur[1].jpg (800x450, 82K)

Because turbo by itself will not perfectly maintain consistent minimum frame rates which is more important than average especially when some scenes make the cpu throttle to different frequencies because more load is placed on the CPU. And yeah I know the GHz war is coming a close as more games start utilizing 8 cores since consoles now use 8 cores as well but it's still gonna be a thing for a while as there will always be pubg-grade games.

They'll probably be about the same price while amd price cuts the 2xxx series

WTF is going on in prey and the witcher 3? That makes no sense desu senpai.

Attached: 1519078327960.png (379x205, 19K)

No, it explicitly says that they're going to use *the benefits of EUV* mostly for power rather than performance. It makes no claims as to what microarchitectural performance boosts it may or may not get.

Just set stable 4.0~4.2 on all cores and stop the retardation.

>That makes no sense

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

>WTF is going on in Prey and the Witcher 3?
MELTDOWN (((MITIGATIONS)))
SPECTRE """PATCHES"""

zen+ is already dirt cheap you filthy jew, 4 years ago a good 8-core processor from intel was like $1,000 and had a max turbo of 3.5 GHz. Now you have a good 8-core processor from AMD with a max turbo of 4.1 GHz going for like $200.

>max turbo of 4.1 GHz
That's gen1, not Zen+.

I dont have much moneys sir AMD needs to be cheaper

See .

That would honestly be better and still have like 95% of the i7-8700K stock performance. Anyway I'm rooting for AMD now desu, if zen 2 can deliver that sweet sweet IPC uplift we can finally have 4.2 GHz on all core chips with the performance of 5 GHz coffee lake intel chips being cooled by a hyper 212 evo.

Attached: 1496950975767.jpg (325x326, 46K)

Lowest I saw the 2700X was like $260, I spotted the 2700 for $199 at frys I think.

FUCK

No, that's not right. It'll be Zen2 @ 7nm, Zen3 @ 7nm EUV, Zen4 @ 5nm EUV, Zen5 @ 5nm EUV, Zen5+ @ 5nm EUV.

Zen6 or Next-gen will end up on 3nm EUV.

7nm brings us 64c/128t + I/O die via chiplets; which are here to stay. 7nm EUV improves on the process by reducing area in print anywhere from 15-20%. This translates as one of two things:

1. 15-20% drop in power/heat for same clocks and a major improvement in latency

or

2. Same power/heat as existing Zen2 with 10% higher clocks and a major improvement in latency.

Translating that to numbers, if Zen1 (14nm) did 74W @ 8c/16t @ ~4GHz, then Zen2 @ 7nm @ 8c/16t @ ~4GHz would draw 37W (with EYPC keynote having a slide that showed 50% power reduction). But as I understand it, AMD is targeting 95W TDP for their 8c/16t R7 3700/X parts, so I'd guess that we'd be looking at around a 4.4-4.5GHz base clock. Let's assume that it's 4.4GHz (and ignore all-core boost + XFR for the moment).

A 15% reduction in area, with a 1:1 assumption in power drop, gives us a 15% reduction in power. So this means that @ 4.4GHz, Zen2 8c/16t would at full load draw 80.75W, a core and half watt usage above Zen1 full load @ 4GHz. 400MHz across increase across all 8 cores leading to only about a 6.75W increase in power is insanely efficient. So then let's take the other scenario, the 10% increase in clocks at same power/heat. So if our baseline for the R7 chips is 4.4GHz, then 10% of that is 400MHz, we factor that back into the CPU and now we have an R7 3700/X that sits between 4.7 and 4.8GHz baseclock, with an all-core boost (w/ or w/o XFR) at 5.0-5.1GHz while drawing only 95W underload.

Above assumes a pure linear scaling between reduction in area & perf/heat w/a focus on clocks. The 15% reduction in area also will reduce the t-dist of the L1/2/3 caches; leading to major latency uplift, ~3-5ns. We could be theoretically looking at 4.7-4.8GHz 8c/16t w/ all-core boost to 5.0-5.1GHz @ 95W w/ ~8700k latencies on Hyper212+ cooler (~$40).