Do you think zen+ will bring amd up to at least parity of the latest and greatest generation from intel...

do you think zen+ will bring amd up to at least parity of the latest and greatest generation from intel? or be a generation below like it is now or even a generation above?

Other urls found in this thread:

wccftech.com/intel-coffee-lake-desktop-6-core-4-core-cpu-leaked/
overclocking.guide/the-truth-about-cpu-soldering/
anandtech.com/show/10968/the-intel-core-i7-7700k-91w-review-the-new-stock-performance-champion/11
tomshardware.com/reviews/intel-kaby-lake-core-i7-7700k-i7-7700-i5-7600k-i5-7600,4870-8.html
overclock.net/t/1342268/pcw-trinity-has-the-ib-problem-thermal-paste-between-the-chip-and-ihs-delidding-results
smallformfactor.net/other/amd-am4-bristol-ridge-apu-delidded
twitter.com/SFWRedditGifs

Intel's next gen being 10nm has less performance than their current CPU's.

Too bad it wont! Intel is already ahead by over 40% in some games!

It would be nice if they increase single core performance but I bet they will add more cores.

Keeping the more cores meme alive.

Because games are dev for intel platforms

No they won't, you're clearly clueless on how Zen is designed if you actually think that.

IPC increase of at least 10%. Probably 15%.
Go away IMG_xxxx

Zen + was originally slated for +15% IPC.
But because Ryzen came in at 50% instead of 40. I'm not sure what the go is. There's two-three other features that weren't included in Ryzen that will be on Zen+, touch wood.

Ryzen already wins in pretty much everything but heavy single threaded workloads (i.e. shitty Javascript-laden websites), poorly optimized games (ARMA 3), really obscure games (DF) and buggy alpha quality emulators (CEmu). Intel's next generation of mainstream CPUs is going 6 cores anyway, so the single threaded performance wars are officially over.

It's gonna be sad times if Zen+ comes out and is only 5% IPC increase, but I don't think they would pull an Intel like that.

>There's two-three other features that weren't included in Ryzen that will be on Zen+, touch wood.
i bought a 1800x a month ago. what features are missing in zen that will be included in zen+? did i get jewed?

You can easily upgrade to Zen+ and sell your old CPU. AM4 isn't going to be instantly obselete.

It already is

you are retarded if you think amd will provide that much increase. they have learned from the jewest of the jew companies.

if the architecture allows them to improve 15%, they would improve only 5% at best, this would allow them to have 3 generations of processors based on the same architecture.

We got another Intel stockholm syndrome victim over here.

AMD is supposed to be putting Zen+ on 7nm before Intel even gets 10nm to market. Intel already had to crank up the clocks on Kaby Lake to beat Ryzen, which is why they have the thermal throttling issues. My guess is Zen+ will match or beat anything Intel brings to the table for single threaded applications and there will be no real comparison in mutlithread workloads, and especially not in price/performance.

Zen+ is not 7nm, Zen++ probably is.

>Intel's next generation of mainstream CPUs is going 6 cores
That'll chew through power like no tomorrow. I'm in Aus, so that's a major concern.
Sorry. I keep forgetting the names of the stuff. So much aberrations to keep track off.
>did i get jewed?
Don't think so. From bit's I've seen it's marginal gains, but gains none the less.
Your next step would be Xeon. Or a 7700K you're not allowed to overclock, and Delidding is a gamble.
I'll be looking forward to see what XFR can do on Zen+.

I guess kike conditioning really fucking works. Like, people are thinking the only way to increase single core performance is by ramping the clocks to housefire levels.

Zen+ is 14nm, GloFo's 7nm DUV is H2 2017 for volume production.

Didn't they say 7nm would come end of next year? When is Zen+ supposed to release?

>Intel's next generation of mainstream CPUs is going 6 cores anyway
i doubt intel will go 6 core mainstream before 2020. even if they currently state they plan to they will probably backtrack.

>That'll chew through power like no tomorrow
It will also reduce it's single core advantage and OC potential significantly.

Q1 2018, so probably March

>2017
2018. Risk production in Q3-Q4 2017.
Only risk production, not a volume one.

It was already spotted in the wild. wccftech.com/intel-coffee-lake-desktop-6-core-4-core-cpu-leaked/

>if the architecture allows them to improve 15%, they would improve only 5% at best, this would allow them to have 3 generations of processors based on the same architecture.
Holy fuck my sides. KEK.
You seem to be suffering from brand confusion. Just because AMD is now the best. Soesn't make them like the former in the elast.

Whoa whoa hold your horses, H2 2018 for volume, Q4 2017 risk production.
Not that it matters much when Intel's 10nm desktop CPUs won't be coming out until Q3 2018 anyway.

OC potential is questionable, but at stock it's gonna be a 3.5GHz shitshow unless they plan to send a 150W CPU on mainstream LGA sockets

But on multi should be competitive. Should the clock not tank too hard.
Still the juice it'll require is absurd. And on second thoughts the heat. Get's fucking 40c here. I don't need that costly shit.
Gonna be fun explaining to shills that my power bill would go up $200+ a quarter because of a six core intel and the air conditioning to not die of exhaustion.

People will try to OC it, find it's basically a poorly binned i7-6850K and give up after their house catches on fire.

No, AMD's SMT is strictly superior. And 6 Intel cores means much lower clocks.

that makes no sense though. why would intel release skylake-x this year along with a kaby lake-x if they will be releasing coffee lake with 6 cores at the same time? why would any buy skylake-x when they can just get coffee?

at this point intel should just abandon hedt because it makes little sense and just focus on xeon again for those who need 12 - 16 cores.

I mean, there's no reason they *can't* add more cores.
But that's a purely SoC/chip level change independent of updates to the Zen core itself.

Personally I'd like to see a socket with a 4 channel memory controller and a healthy number of PCIe Gen3 lanes like X99.

Core level changes used to be the meat and potatoes of ArsTechnica and Anandtech way back in the day. They'd cover changes to branch prediction, number of reservation stations, L1/L2 size and latency, ROB depth, etc. Hopefully we get that kind of coverage for Zen vs Zen+1, because otherwise we are going to have nothing to go by trying to look beyond the benchmark numbers to possible causes for better or worse performance on certain workloads.

I don't know. i7-7740K is just a desperation panic move to see if they can get the last of the MUH SINGLE CORE idiots to drop more cash on it. It will probably still have jizz TIM.

Because a socket twice a year keeps the goy in fear

4 channels on Zeppelin-like die means 16 fucking channels on Naples. I mean whoa, it's cool but someone like Supermicro will literally kill themselves due to total instanity of wiring 16 memory channels per socket.

unless intel fixed their voltage problem with 14nm. which does seem like intel has been slowly refining it. kaby lake kept voltages similar to skylake but bump up the clocks.

i could see intel releasing a 6 core coffee with the same ipc as skylake and kaby, but come stock at 4ghz, 4.2ghz all core, with a 4.6ghz turbo boost 3. they could easily integrate their turbo boost 3 that's been in use with broadwell-e on x99.

tossing in six cores, four extra threads, can easily give them their "hurrr 15% increase over kaby" claim. we know now since kaby intel isn't necessary talking about IPC increase. kaby was 100% identical to skylake in IPC, but offered much higher clocks and that's where their boost came from. coffee lake can be from core increase and further higher clocks. we could see quad coffees in the 5ghz range.

Coffee Lake-S will probably still be on LGA1151, but will require a new chipset .. so RIP Z270 motherboards.

agner

>stock at 4ghz, 4.2ghz all core, with a 4.6ghz turbo boost 3
Their engineering sample was clocked at 3.5 GHz base. I don't think you're going to see those kind of numbers except under an extreme OC.

So presshot all over again? Nice, MOAR NIGGUTHURTZ, BURN THEIR HOUSES DOWN.

Sorry, you've confused me for a intel fanboy. By competitive I mean from the eyes of an AMD consumer. Ball park figures. Not pissing contests.
Still be interesting to see. Especially the price. Not too long ago on this board, the faggots were wanting intel to drop the Igpu for moar cores. expecting it to cost the same.

Intelshill detected?
Now we have Ryzen and Core processors. Just Wait(TM) for new games with optimized code

Could just do 1 DPC rather than 2 DPC, would increase bandwidth but not capacity

yeah and we saw ryzen engineering samples at 2.8ghz. that doesn't mean those are final clocks.

Board complexity will skyrocket anyway. I mean sounds cool, let's wait for 7nm Zen.

You seem more versed in this than the idiots up in the thread, so lets talk properly:

They CAN add more cores, but they won't for mostly cost reasons.

-A CCX would have to be 6 core, so that would be 12 core per chip
-The reason for 12 cores instead of 5+5=10 is because of the APUs, they're made of one CCX and these things are always power of two
-That also means there's nowhere near enough pins on AM4 for that
-Now you also have a problem with feeding the cores, you need three or quad channel memory, which is too expensive for mainstream platform
-But now you've also completely invalidated your 12-16 core HEDT quad channel memory line
-But now you also got a 350mm^2+ CPU die for mainstream which you're gonna sell at bulk under $200, which is stupid.

Ergo, more cores are gonna be kept for 7nm
Pinnacle ridge(aka zen+) is adding in the features that Zen didn't get due to time constraints

i7-6850K is already 3.6 GHz base, 3.8 GHz boost. And that has a TDP of 140w. You're telling me they're going to improve by 400 MHz and keep TDP under control? It's not even a die shrink.

We knew nothing about Zen's efficiency. We know that Intel's sucks. They can't just add more coarz without lowering the clocks. Unless you want another presshot.

>Unless you want another presshot.
I do, but only because it would be hilarious to watch.

I want it just to see THG shill it. AGAIN.

>6 cores at 5.0GHz
>literally pulling 300W from the socket

Anyone think this is a bit of a TERRIBLE idea without very expensive motherboards?

>things are always power of two
excuse me did you just assume my gender
I sexually identify as a tricore and you will regard me as such

You never go full retard.

>You never go full retard.
Oh, Intel fucking LOVES to go full retard.

I mean really way to be a numerist

6850k is based off of xeon for one. so its under a slightly different design even though it shares the same architecture as regular broadwell. it also has quad channel memory and pci-express lane support upwards of 40. that = more power consumption. also, a 6850k in practice, at stock clocks, rarely will use anywhere close to 100 watts. i should know, as pic related, i had a 6850k.

gut out the quad channel, gut out the extra pci-express lanes, incorporate some of the newness they added to broadwell-e, like down-clocking in AVX workloads and incorporate turbo boost 3, more maturing of their 14nm node, and i could easily seem them releasing a 95 watt 6 core with coffee in the 4ghz range.

zen is very energy efficient, but intel also isn't horrible. if they do the above and get a hold of their voltage problem, they could very well easily offer a 6 core coffee in the 4ghz range.

>-A CCX would have to be 6 core, so that would be 12 core per chip
More than likely, the basic CCX design intrinsically takes four cores. I don't think they'll be able to scale the local interconnect beyond that. It's more likely they'll add CCXs instead if they want more cores.

Slap shitty jew cum TIM on top of it and the recipe of housefire is yours.

They could do all that and even then if they took it to 5ghz it'd use 300w, Kaby Lake starts pulling hilarious amounts of power, moreso than prior processes, when you start cranking it up.

If it's true that Intel only used jizz TIM for Kaby Lake because of the smaller die size, the larger die size of 6 core Coffee Lake-S should allow them to solder it. We'll see if they actually do so.

Good thought, but considering their own roadmap puts Pinnacle ridge at 8 cores, we can expect 3xCCX or 2xCCX with 6 cores on 7nm, if even that, I'd actually be happier if they spent transistors on making the IMC and cores faster, possibly some more lanes as well.

For moarcorz, there's the HEDT line.

Haha no they've been using jizz since Ivy. Microcracks are meme.

If they can get the voltage hump moved out to around 4.5 and get a little ipc increase. If not

>gut out the quad channel, gut out the extra pci-express lanes, incorporate some of the newness they added to broadwell-e, like down-clocking in AVX workloads and incorporate turbo boost 3, more maturing of their 14nm node, and i could easily seem them releasing a 95 watt 6 core with coffee in the 4ghz range.

You're forgetting their mainstream CPUs have integrated GPUs. Unless they leave those out, which I don't think they would do, their job of hitting 4 GHz stock at 95 watts just got a lot harder.

the only reason why zen is soldered is because of its larger die. intel wasn't bullshitting about the reasons why they adopted tim. don't get me wrong, it also has the nicer benefit of being cheaper, but there is a scientific reason why. if amd could get away with using tim, believe me, they would too. don't misunderstand me though. fuck them and profits. i prefer solder.
unless, again, they get a hold of their voltage. yes kaby does get out of control, but kaby did bring improvements with voltage over skylake. with being able to run at 4.4ghz all cores at the same voltage skylake required to run 4.2ghz on a single core. amd has a voltage problem as well. zen is amazingly efficient until you break the 3.6ghz barrier. take amd's pinnacle in binning, the 1800x. it needs jack shit voltage for 2ghz all the way to 3.6ghz. a low, gradual increase. but the moment you try 3.8ghz? 4ghz? 4.1ghz+? voltage goes through the roof.

>implying the iGPU is even needed for the 7700k at STOCK to blow 150W of power.
What a fucking housefire, can you imagine this at 6 cores at even higher clocks?

LIterally exploding mosfets everywhere

>considering their own roadmap puts Pinnacle ridge at 8 cores, we can expect 3xCCX or 2xCCX with 6 cores on 7nm
I don't see why the former would imply the latter.

I've generally noticed that a very great many CPU designs except Intel's seem to focus on 4 cores per some kind of cluster. Probably the reason for this is that they're using a crossbar or some other kind of broadcast-based design per cluster, and that 4 cores is the highest number for such a design before it either starts dropping like a rock in bus conflicts, or the crossbar starts scaling away like a rocket.

I'm too stupid to read this chart, where does it show 150 watts at stock?

No, they were. They stopped soldering since Ivy. When the dies were big.

Top of the chart under "default"

unless intel doesn't include the gpu as part of the cpu's tpd. there is nothing that prevents them from doing so.

What is delta next to the power column?

You don't think Intel would tell a big fat lie like that, do you?

read what i wrote. they stopped because ivy die size became to small. ivy is smaller than sandy.
if you want to understand why you can't solder with dies that become a certain, small size.
overclocking.guide/the-truth-about-cpu-soldering/

oh goy intel would never do such a thing!

Fucking kike get in the oven. Intel's 22nm quadcores are about as big as Zen is right fucking now and they were not soldered. Fucking gas yourself schlomo.

I looked up this chart and it's from here, they claim delta is power draw. I'm not sure what power actually is, then. anandtech.com/show/10968/the-intel-core-i7-7700k-91w-review-the-new-stock-performance-champion/11

>it took amd only 7 years to get to sandy bridge levels of performance
>1 year from now and we will get ivy bridge

BASED AMD AM I RIGHT FOLKS?

How confusing.

What a surprise, he's getting BTFO'd in the comments.

>Intel Clarksdale 2c is soldered and has a 81mm die size which is twice as small as Ivy Bridge 4c being 160mm which uses TIM. Sandybridge 2c is 131mm and 149mm which are both also soldered.
Skylake uses TIM and is 122mm in size compared to Clarksdale Soldered 81mm.

glad to see you read the link /s

>ivy
160mm
>ryzen
~210mm

Tom's is shit, but at least they have good power testing.

tomshardware.com/reviews/intel-kaby-lake-core-i7-7700k-i7-7700-i5-7600k-i5-7600,4870-8.html

And yup, 140W at stock under max load

>The micro cracks will also decrease the thermal conductivity but will especially increase the thermal resistance at the corner of the DIE. Without the gold layer between diffusion barrier and solder preform, delamination of the solder preform would occure after few thermal cycles. Micro cracks occur after about 200 to 300 thermal cycles. A thermal cycle is performed by going from -55 °C to 125 °C while each temperature is hold for 15 minutes. The micro cracks will grow over time and can damage the CPU permanently if the thermal resistance increases too much or the solder preform cracks completely.

>Void and micro crack occurrence is mainly affected by the solder area – thus the DIE size. Small DIE size (below 130 mm2) e. g. Skylake will facilitate the void occurence significantly. However, CPUs with a medium to large DIE size (above 270 mm2) e. g. Haswell-E show no significant increase of micro cracking during thermal cycling (Figure 12).

Wow, he says you need to pour some fucking LN2 FOR fucking cracks to occur. Gas yourself you fucking kike.

That only seems to be happening using the Intel Power Thermal Utility. Which looks like some sort of ultimate stress test.

well it makes sense, when you put so much voltage on such small die you end up wasting it somewhere

Holy fucking shit there's some legit Intel shills out there.

you are a real idiot. that's to speed up the micro cracks to showcase them without having to wait a few years. if you take a processor, and doing constant thermal fluctuations, such as going from 30c to 80c, and back and forth, over a few years, from normal day to day, you will gain enough cracks to not only hurt thermals, but damage the processor as well.

Zen got to mostly nearly Haswell IPC (theoretically fundamentally higher but maybe a little slowed down by branch prediction, internal bandwidth, etc.).

Zen+/2/whatever should get to somewhere between Broadwell and Skylake for integer perf while still skimping on FP and again not clocking as high.

Intel will still have higher single-thread performance, and AMD will still have the only affordable 8c desktop and 12c/16c HEDT chips, so everybody can be happy with whichever one they want.

>up to at least parity

My gut feelings is that when the dust settles this year, AMD will be about one-half of a generation behind Intel.

I think AMD's poor single-thread performance will hurt them badly. You gotta remember that a vast majority of users don't make much use of those extra cores. They'll feel that poor single-thread performance whenever they view slow JavaScript web pages, which will be all the fucking time. That's a much bigger use case than you enthusiasts who run handbrake or whatever and enjoy every last one of those 16 threads.

That's why I said max load, the FPU stress test only tests half the execution units.

A more realistic tests would be some heavy AVX2 code

overclock.net/t/1342268/pcw-trinity-has-the-ib-problem-thermal-paste-between-the-chip-and-ihs-delidding-results

>amd uses tim
that's right folks. amd uses tim when they can. they do it for the same reasons as jewtel.

You fucking kike, read the fucking article you posted, the actually CPU silicon is shielded by several metal layers. Holy fuck schlomo Intel used to solder even the smallest of dies and no one, fucking no one complained about microcracks you filthy kike.

They stopped using fucking TIM right after that.

It would be realistic if Intel didn't throttle clocks during AVX.

It overheats like crazy without AVX offset.

smallformfactor.net/other/amd-am4-bristol-ridge-apu-delidded
>AMD Socket AM4 "Bristol Ridge" APU De-lidded
>AMD is using high-quality TIM between the die and the IHS
though i will give them credit to at least using liquid metal now.

>four high-end cpus
>8 mb llc
>entire northbridge
>12x12 mm
We truly are living in the future.