Mid 2016

>Mid 2016
>AMD says Zen is on par with an Intel 6900k

>Late 2016
>Reporters say Zen barely matches a 6 core 6850k

>2017
>???

The hype is already going downhill, AMDrones

enjoy your Sandy Bridge IPC in 2017

Other urls found in this thread:

youtube.com/watch?v=xzZT2xH3zBk
twitter.com/SFWRedditImages

another good reason to stay on Sandy Bridge for the next 5 years

Oh shit, nobody really cares. But keep shitposting, you probably have nothing to do.

>>AMD says Zen is on par with an Intel 6900k
AMD literally never said this.
It was consumers overhyping once more. an unfortunate byproduct of AMD keeping tight lips.

All AMD ever confirmed was that Zen had hit the targetted 40% IPC increase over Excavator. Which would put it at around Devil's Canyon levels depending on clocks.

>Im real butthurt so Im gonna speak for everyone else
this is a technology board you attention whoring angry faggot

>8 core, 16 thread, proper SMT
>~Ivy Bridge/Broadwell IPC
>unlocked multiplier
>recent process
>$250-$300?!
>motherboards with USB 3.1 and DDR4, prepped for one big graphics card

I'll consider that. Gains since Sandy Bridge are pretty much line noise except AVX2, and Zen has that and does okay with it too.

It's sounding like a worthwhile prospect at this point. We'll need to wait for final prices and compare it to Intel's Kaby Lake desktop lineup properly, but it seems Intel have some serious competition on their hands for 2017.

Fucking finally. Intel might get their heads out their asses, stop the ridiculous product differentiation for no reason they've been doing for years, and drop prices to have a proper tussle - which is good news all round for us.

Also Intel can't GPU their way out of a paper bag compared to AMD or nVIDIA but they insist on shoving them in there where they're not wanted by either the enthusiast or professional workstation markets.

The 32-core Zen Opteron sounds like it's going to be a fucking beast, too, and is likely to be not as stupidly priced as Intel's high core count Xeon offerings. How popular it'll be, remains to be seen.

Also this means SGX won't be usable for DRM due to the lack of a monopoly. We do not want a complete monopoly. It's good to see the underdog still kicking.

AMD literally put up Zen against a 6900k in one of their tech demos, you cant just shift the blame fanboy faggot

>Broadwell IPC
8 core Zen competes with 6 core Broadwell
This thing has 75% of Broadwell's IPC, sounds awful

The only Xeon SKUs with iGPUs are the E3-1225, 1235, and 1245.

Faildozer all over again.

sooo business as usual at AMD

Whats worse is that they lie about this shit, leave it for the fanboys to make up even more dumb shit, and this board still treats them as some kind of savior against that Jewish merchants which have a stronghold on technology, when they're just as bad.

You forgot the pricing leak showing that the 2-300$ cpu that's on par with a 6800k is a 6c/12t variant.

Post it.

>on par with Broadwell-E at 95w
Im calling bullshit, AMD is the housefire company

>30W more thermal headroom
>only 100MHz gain in base frequency
REPEAT AFTER ME:
POO IN LOO

its the OP pic, good job figuring out shit for yourself

And twice the threads

>twice the threads.
Look at the thermal margin differences between an i5 and an I7. I'm not confirming nor endorsing anything without any proof. Just pointing out that current leaks puts the cpu curry tech mentions as a 6c Sku. Not to mention that even if they only managed ivy bridge levels of IPC, you would be out of your damn mind to expect an 8c/16t cpu for 300$ or less.

i7 6700 is 65w and 6700k is 91w

the 6700 is locked and underclocked

you're looking at the same thing with those 2 Zen cpus, its the same situation. Has nothing to do with the threads.

>Reporters say Zen barely matches a 6 core 6850k

AMD literally never said that

>blender demo
so.. this is the usual AMD stuff , showing off multi threaded performance, not talking about single thread stuff

Did AMD learn nothing from its recent GPU launch ?
Don't hype shit up if you can't deliver.

AMD usually leaves it to their fanboys to overhype it for them instead

I bet it wont even match an i5

You know AMD won when all these intel shills are doing maximum damage controll in this thread.

I mean theyre both 8 core so you can just search up benchmarks for 6900k and find out for yourself

but some people are saying its closer to the 6 core Intel, which means single core will be trash

There is no actual citation for "Zen matches a 6 core 6850k"

It's a single articles that click bated the shit out of everyone that then changed title in couple of hours.

>ITT: [CITATION NEEDED]

in the OP pic, both processors were clocked to 3.0ghz.

it was a demo of IPC, nothing more.

I mean if you want to be a retard and buy into overhyped trash thats your problem

Why not just make a single core CPU with insane clock speeds

thats funny because AMD provides little to no info and sells purely on hype

I would rather lose my dick than give any money to Shilltell or its mouth breathing diaper wearing fans

My old ass junkrat 8350 can still deliver 60fps with a 980ti at 1440p

They delivered years ago and will do so again, intelcuck.

Double checked older i7s expecting to see hyperthreading increasing tdp. I stand corrected on that point. Though I would not be surprised if amds smt solution increases the tdp of the chip. All the cpus are supposed to be unlocked for zen.

And it was a success in that regard. They were 100% transparent in what they were doing. Then shills flipped shit how amd "crippled" the Intel chip and that the turbo would make it go even higher than 3ghz. No mention that they would only increase by 1-200mhz on an 8 core turbo. Assuming the cooler could support it.

Fucking finally someone said it. Also why the HELL are they trying to make CPUs so small? I'd take the extra heat/power for a BIG ASS CPU that just wrecks everything

are you serious?

>AMD provides little to no info
Intel doesn't provide much info on future products either. There's no reason to until the product is nearly released.

Their engineering samples clock at like 3.2GHz. They downclocked the 6900k to 3.2 GHz to show that their IPC is in the same ballpark.

Running 40 single threaded applications on a single core is less efficient than running 40 single threaded applications on 4 cores. Adding more cores improves performance in general until you run out of applications to run on those cores.

In the blender demo it was 3ghz with no turbo. Later leaks showed engineering samples with a 3.2ghz clock rate, assumed the turbo. Current leaks show 3ghz with a max 3.5ghz turbo. How that max turbo is stepped and achieved is unknown. Going off historical amd turbo methodology, it'll likely be shutting off half the cores. Maybe disabling smt.

>Fundamental misunderstanding of what IPC is
>Conflating multi threaded performance with IPC

Who is this retard spreading his autism?

You're just a retard that can't read or listen.

youtube.com/watch?v=xzZT2xH3zBk

Strong emphasis on on identical clock speeds

>I don't know what IPC is
>calculate IPC by comparing cores instead of frequency of a single core vs total single core performance

If this is true then AMD has won.

it will be shit

trust me, we dont even know all the other problems it has

You're riddled with problems, yet the world still lets you live.

Are you clinically retarded? When they use an older manufacturing process they obviously have chips that produce more heat and consume more power but costs less. This has nothing to do with the competence of the company itself.

Zen is closing the gap because they stopped using an ancient manufacturing process that requires religious incantations.

>intel cpu vs amd cpu
>same core count
>same frequency
What more do you want?

What single core workloads are faster when the other cores are not utilized?

desu I think those programmers that can't parallelise their compute bound workloads onto multiple cores deserve to be shot. Especially when it comes to games.

Because power consuption scales quadratically. Four cores with 3ghz each consume less power than a single 12ghz core. This is why servers are mostly below 3ghz.

They sell on price and theoretical performance. In practice muh games run on a single core because of incompetence.

Conveniently much of the work servers do can be parallelized.

how about benchmarks that haven't always favored AMD's multi core solutions ?

>95w 8 core
there better be some mini itx motherboards for this, all Lga2011 has is a single shitty Asrock

You won't see any x370 motherboards, but the one step down that has overclocking, or even just base models without overclocking I'm sure.

/thread
i5-2500k master race

Correction. I suppose you could call a need for the added pcie lanes given m.2 and Sata express. But I'm doubtful they'd be able to cram all that shit on an itx board.

E5-1650 v2 baby. Zen won't even be a significant upgrade for me. But I'm excited. More looking forward to zen+

if Applel can cram a workstation into a dust bin why havent mobo companies followed this yet

>barely matches Skywell E
????
Are you memeing here that is great fucking news. I was expecting not even haswell IPC. If it's close to skylake ipc a true 8core on a mainstream platform for like $300 will btfo intel on so many levels.

Broadwell E is way below skylake IPC

>$169 for an unlocked 4 physical-core Zen processor with a 512-core GPU @ 1GHz that can use 51.2GB/s DDR4 RAM (2x 3200MHz)
Is this real? Wouldn't this basically have the performance of a $250 i5-4690K + a bitching iGPU?

Does it support ECC?

ZEN WHEN

Broadwell is a die shrink of Haswell.
Skylake is Broadwell with "optimization".
Kaby Lake is Skylake with "optimization".

Intel has not changed it's architecture. They will all scale linearly in IPC as clock speeds go up. Any gaymen performance Min/Avg increases you see are a direct result of platform change to DDR4 ram.

I hate Sup Forums. Im going back to my cave.

Right now

The key difference between the x370 chipset and the chip one step down is 2x16 crossfire. Unless you're going to design your own 100% proprietary system(like apple) there's no way to cram 32 pcie lanes worth of shit into an itx form factor, as well as being built up for overclocking. There would be no point in using that particular chipset when there is an overclocking ready one better suited to the itx form factor.

Unconfirmed leak, and couldn't find source to the image myself. The igpu would be effectively 67% of an rx 460 running off of slower memory. Though the hbm support is intriguing. Also, that Sku is 4c/8t

Noone knows.

No its not you idiot.

So, so many reasons at this point:

1. It's not cost-effective; it requires more R&D than multicore.

2. It's not energy-efficient; it requires more TDP to ramp up to higher ghz. The ideal chip would probably be one sitting at 3ghz with insane IPC.

3. It's not heat-efficient; the added TDP drives up heat and makes difficult issues like electron transmigration and system cooling even more pernicious.

4. It's not as easy as designing multicore; there are few designers even around in the world today that are capable of drastically improving clockspeed/IPC at the drop of a hat.

5. Clock speed doesn't scale well. Why do you think every series of production chip ever is considered mature by Intel and AMD alike when it is clocking somewhere between 3-4ghz? IPC is where the real improvements are in any case. The 1000 extra cycles a 5ghz chip has over a 4ghz chip don't mean nearly as much as the 1000 enjoyed by a 2ghz over a 1ghz; aside from even the technical aspects of handling that many instructions per second and making the rest of the system actually use those cycles, it's diminishing returns for lots more electricity; that fifth thousand is only 20% increase in raw speed compared to the 50% increase that the second thousand was.

6. It's easier to rewrite software to use multicore for close to 100% scaling than it is to redesign hardware to handle a 6ghz CPU. Most software these days is multithreaded to one extent or another.

7. Software that NEEDS to use multiple threads may experience issues trying to stack those threads on single cores; witness the severe stuttering GTA5 had with overclocked Pentium G3258 systems. It's a documented problem and it can be surmised the problem would get worse in scientific compute where software must run in multiple threads for more serious reasons.

8. Decreasing process sizes naturally lead designers to make use of die space that would otherwise go underutilized.

you dont need 32 pcie lanes on mini itx, at least I dont think
I mean, Asrock could do it for X99. Im just hoping someone does it for Zen is all.

Then why don't AMDs 8 core processors wreck 4 core intel chips in gay man performance?

Because their IPC is too low, you subhuman.

Well I use apple so...

Yeah, all I was saying is you likely won't get an x370 board if that's what you were looking for. zen has a much smaller socket than lga 2011, and is more soc-like. So it would actually be easier to downsize.

Zen is really a single high-end CPU? That's weird, I was expecting it to be an architecture with multiple tiers like Skylake or Haswell.

>muh IPC
stop it. CPU performance isn't measured in IPC. There is no objective measurement of IPC anyway.

Trolling is an art, and you've got no paintbrush.

This user is right. CPU performance is measured in bananas peeled per second.

I've not really looked into Zen. I'm considering putting together a system for around Q1 of 2017. Generally speaking I've been set on going with an E5-1650 v4 (or whatever is the equivalent at the time)

May Zen be able to offer something competitive against the E5-1650, particularly for virtualisation on ESXi? Kind of curious now, especially since the fall of GBP making them now cost circa £600 as opposed to the previous £500...

I get the impression that historically AMD has been quite friendly with virtualisation in relation to what is and isn't supported/locked down... I'd hope this too is the case for Zen?

The real question is, how long will it take them to beat my Evga sr-2 based 12core with a mild overclock.

I'm guessing 2020 or so.

theres no time travelers here you faggot
wait and see.

>instructions per clock
>an actual set value

>this Intel chip has millions of IPC
>this toaster has four
>durr u can't beasure performance using IPC

Then why is everyone speculating so fucking much about it?

Has there been anything officially confirmed aside from the usual >Zen will beat 'x', honest!

Alright so based on the information we have now is it looking like Zen is going to be a worthwhile upgrade coming from a Deneb chip?

by the way, you're a loony coon, mate.

>Fucking finally. Intel might get their heads out their asses
Intel has been wanting to start production with EUVL and will finally do so.

>EUVL

It should be. Hold up until it comes out though.

What about efficiency?
Would it be any good for mobile?

amd completely restructured their cpus with zen to be more power efficient and less hot

No word on mobile raven ridge SoCs yet. I imagine they'll have VP9 and HEVC HW decoding like kaby lake though.

AMD says it has better efficiency than Carrizo and it's supposed to scale from small low power chips to large chips in multi-chip configurations

>Intel has been wanting to start production with EUVL and will finally do so.
And will fail at it for a few years just like their 14nm node?
It's ridiculously impressive the energy efficiency they achieved on 32nm node and on a really generic 28nm node
The Zen APU's are coming after their HEDT offerings, I don't expect them before 2017H2
It's a heads to head comparison, exact same cores and exact same clocks
The only thing that would pull Intel ahead are completely skewed benchmarks on their favor, or their stuff being clocked higher (it might be possible that Zen doesn't clock as high as current Intel offerings)
>May Zen be able to offer something competitive against the E5-1650, particularly for virtualisation on ESXi? Kind of curious now, especially since the fall of GBP making them now cost circa £600 as opposed to the previous £500...
Zen is aimed mostly at the hueg server market, Naples shows this, they want their Opteron market back
>I get the impression that historically AMD has been quite friendly with virtualisation in relation to what is and isn't supported/locked down... I'd hope this too is the case for Zen?
Yeah, there's no reason to think they will segment AMD-V
Zen is aiming at everything with an emphasis on everything, but it should be able to compete well in ULV applications, I don't expect to see it outside of laptops and maybe some meme dev boards
They have implemented a lot of power saving features, like really aggressive clock gating
They will, Polaris already has it

Polaris hw decoder for hevc is awful i hope they fix it somehow with a driver update but right now it has problems with hevc @ 60fps

>>Reporters say Zen barely matches a 6 core 6850k
We don't know the frequency the zen was running at :^)

>AMD official press releases never said that

>Wccftech does not have real reporters, they write articles based on forum posts/rumors.

Internet trolls baiting with false figures hyping it up to stupid levels so AMD gets blamed for poor performance when it lands, even though they are hitting the targets they promised. Fucking hype niggas man. Fucking hype niggas.

At what bitrate and resolution?

>>Reporters say Zen barely matches a 6 core 6850k

*Citation needed

And their new APU's that is what I care about

They 3.2 GHz dual-channel DDR4 now nigga. intel on suicide watch imminent.

They are saying it will be as good as a Rx 460
That's huge

If Zen isn't worth it in 2018, I'm going to light myself on fire and run screaming through AMD's head office.

Carrizo/Bristol Ridge only support 2400mhz without increasing the north bridge, base clock, and that fucks everything else in the system.
They have the exact same Cadence IMC used in Kaveri, and some of the highest memory overclocks with it were 2600mhz~

Nothing confirmed for Raven Ridge yet. We still have no official word on whether or not it'll use HBM. All we know is that its 12CU, and the fastest DDR4 possible still wouldn't provide enough bandwidth for it.

2017*